Automotive Informatics

and Communicative
Systems:
Principles in Vehicular
Networks and Data Exchange

Huaqun Guo
Institute for Infocomm Research, A*STAR, Singapore

Information science reference

Hershey • New York
Director of Editorial Content: Kristin Klinger
Senior Managing Editor: Jamie Snavely
Managing Editor: Jeff Ash
Assistant Managing Editor: Carole Coulson
Typesetter: Michael Brehm
Cover Design: Lisa Tosheff
Printed at: Yurchak Printing Inc.

Published in the United States of America by

Information Science Reference (an imprint of IGI Global)
701 E. Chocolate Avenue, Suite 200
Hershey PA 17033
Tel: 717-533-8845
Fax: 717-533-8661
E-mail: cust@igi-global.com
Web site: http://www.igi-global.com/reference

and in the United Kingdom by

Information Science Reference (an imprint of IGI Global)
3 Henrietta Street
Covent Garden
London WC2E 8LU
Tel: 44 20 7240 0856
Fax: 44 20 7379 0609
Web site: http://www.eurospanbookstore.com

Copyright © 2009 by IGI Global. All rights reserved. No part of this publication may be reproduced, stored or distributed in any form or by
any means, electronic or mechanical, including photocopying, without written permission from the publisher.
Product or company names used in this set are for identi.cation purposes only. Inclusion of the names of the products or companies does
not indicate a claim of ownership by IGI Global of the trademark or registered trademark.

Library of Congress Cataloging-in-Publication Data

Automotive informatics and communicative systems : principles in vehicular

networks and data exchange / Huaqun Guo, editor.
p. cm.
Includes bibliographical references and index.
Summary: "This book advances the understanding of management methods, information technology, and their joint application in business
processes"--Provided by publisher.
ISBN 978-1-60566-338-8 (hardcover) -- ISBN 978-1-60566-367-8 (ebook) 1. Automobile industry and trade--Management. 2.
Information technology. I. Guo, Huaqun, 1967- HD9710.A2A8725 2009 388.3'12--dc22

200805016

British Cataloguing in Publication Data

A Cataloguing in Publication record for this book is available from the British Library.

All work contributed to this book is new, previously-unpublished material. The views expressed in this book are those of the authors, but not
necessarily of the publisher.
Editorial Advisory Board
Lawrence Wai-Choong Wong, National University of Singapore, Singapore
Weihua Zhuang, University of Waterloo, Canada
Ozan K. Tonguz, Carnegie Mellon University, USA
Javier Ibañez-Guzmán, RENAULT S.A.S., France
Todd Hubing, Clemson University, USA
Nicholas F. Maxemchuk, Columbia University, USA
Farid Naït-Abdesselam, University of Sciences and Technologies of Lille, France
Yul Chu, University of Texas–Pan American, USA

List of Reviewers
Sohel Anwar, Indiana University Purdue University Indianapolis, USA
Raúl Aquino-Santos, Universidad de Colima, México
Teck Yoong Chai, Institute for Infocomm Research, A*STAR, Singapore
Yul Chu, University of Texas–Pan American, USA
Gavin Holland, HRL Laboratories, LLC., USA
Todd Hubing, Clemson University, USA
Javier Ibañez-Guzmán, RENAULT S.A., France
Tie Yan Li, Institute for Infocomm Research, A*STAR, Singapore
Nicholas F. Maxemchuk, Columbia University, USA
Farid Naït-Abdesselam, University of Sciences and Technologies of Lille, France
Lek Heng Ngoh, Institute for Infocomm Research, A*STAR, Singapore
Fabienne Nouvel, Laboratory IETR/INSA, France
Stephan Olariu, Old Dominion University, USA
Vasudha Ramnath, Institute for Infocomm Research, A*STAR, Singapore
Biplab Sikdar, Rensselaer Polytechnic Institute, USA
Joseph Chee Ming Teo, Institute for Infocomm Research, A*STAR, Singapore
Satish Ukkusuri, Rensselaer Polytechnic Institute, USA
Ziyuan Wang, University of Melbourne, Australia
Lawrence Wai-Choong Wong, National University of Singapore, Singapore
Yew Fai Wong, National University of Singapore, Singapore
Zonghua Zhang, National Institute of Information and Communication Technology, Japan
Weihua Zhuang, University of Waterloo, Canada
Table of Contents

Foreword..............................................................................................................................................xiii

Preface................................................................................................................................................... xv

Acknowledgment................................................................................................................................. xxi

Chapter I
Introduction: An Emerging Area of Vehicular Networks and Data Exchange........................................ 1
Huaqun Guo, Institute for Infocomm Research, A*STAR, Singapore

Chapter II
Drive by Wire Systems: Impact on Vehicle Safety and Performance.................................................... 12
Sohel Anwar, Indiana University-Purdue University Indianapolis, USA

Chapter III
Electromagnetic Compatibility Issues in Automotive Communications............................................... 48
Todd H. Hubing, Clemson University International Center for Automotive Research, USA

Chapter IV
Automotive Network Architecture for ECUs Communications............................................................ 69
Fabienne Nouvel, Laboratory IETR-UMR, INSA, France
Wilfried Gouret, Laboratory IETR-UMR, INSA, France
Patrice Mazério, Laboratory IETR-UMR, INSA, France
Ghais El Zein, Laboratory IETR-UMR, INSA, France

Chapter V
Enabling Secure Wireless Real-Time Vehicle Monitoring and Control................................................ 91
Lek Heng Ngoh, Institute for Infocomm Research, A*STAR, Singapore

Chapter VI
MAC and Routing Protocols for Vehicle to Vehicle Communications................................................ 105
Xiaobo Long, Rensselaer Polytechnic Institute, USA
Biplab Sikdar, Rensselaer Polytechnic Institute, USA
Chapter VII
Inter-Vehicular Communications Using Wireless Ad Hoc Networks.................................................. 120
Raúl Aquino-Santos, University of Colima, Mexico
Víctor Rangel-Licea, National Autonomous University of Mexico, Mexico
Miguel A. García-Ruiz, University of Colima, Mexico
Apolinar González-Potes, University of Colima, Mexico
Omar Álvarez-Cardenas, University of Colima, Mexico
Arthur Edwards-Block, University of Colima, Mexico
Margarita G. Mayoral-Baldivia, University of Colima, Mexico
Sara Sandoval-Carrillo, University of Colima, Mexico

Chapter VIII
The Role of Communications in Cyber-Physical Vehicle Applications.............................................. 139
Nicholas F. Maxemchuk, Columbia University, USA & IMDEA Networks, Spain
Patcharinee Tientrakool, Columbia University, USA
Theodore L. Willke, Columbia University, USA & Intel Corporation, USA

Chapter IX
Integrating Traffic Flow Features to Characterize the Interference in Vehicular Ad Hoc
Networks.............................................................................................................................................. 162
Lili Du, Purdue University, USA
Satish Ukkusuri, Rensselaer Polytechnic Institute, USA
Shivkumar Kalyanaraman, Rensselaer Polytechnic Institute, USA

Chapter X
Proactive Traffic Merging Strategies for Sensor-Enabled Cars........................................................... 180
Ziyuan Wang, University of Melbourne, Australia
Lars Kulik, University of Melbourne, Australia
Kotagiri Ramamohanarao, University of Melbourne, Australia

Chapter XI
The Localisation Problem in Cooperative Vehicle Applications......................................................... 200
Javier Ibañez-Guzmán, RENAULT S.A.S., France

Chapter XII
An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle
Navigation Systems............................................................................................................................. 219
Denis Gingras, Université de Sherbrooke, Canada

Chapter XIII
Efficient and Reliable Pseudonymous Authentication......................................................................... 247
Giorgio Calandriello, Politecnico di Torino, Italy
Antonio Lioy, Politecnico di Torino, Italy
Chapter XIV
Simulation of VANET Applications.................................................................................................... 264
Valentin Cristea, University Politehnica of Bucharest, Romania
Victor Gradinescu, University Politehnica of Bucharest, Romania
Cristian Gorgorin, University Politehnica of Bucharest, Romania
Raluca Diaconescu, University Politehnica of Bucharest, Romania
Liviu Iftode, Rutgers University, USA

Chapter XV
In-Vehicle Network Architecture for the Next-Generation Vehicles................................................... 283
Syed Masud Mahmud, Wayne State University, USA

Compilation of References ............................................................................................................... 303

About the Contributors .................................................................................................................... 330

Index.................................................................................................................................................... 338
Detailed Table of Contents

Foreword..............................................................................................................................................xiii

Preface................................................................................................................................................... xv

Acknowledgment................................................................................................................................. xxi

Chapter I
Introduction: An Emerging Area of Vehicular Networks and Data Exchange........................................ 1
Huaqun Guo, Institute for Infocomm Research, A*STAR, Singapore

This chapter gives an overview of this emerging area of vehicular networks, its potential applications,
its potential wireless technologies for data exchange, and its research activities in the Europe, the United
States (U.S.), Japan, and Singapore.

Chapter II
Drive by Wire Systems: Impact on Vehicle Safety and Performance.................................................... 12
Sohel Anwar, Indiana University-Purdue University Indianapolis, USA

An overview of the drive-by-wire technology is presented along with in-depth coverage of salient drive
by systems such as throttle-by-wire, brake-by-wire, and steer-by-wire systems, and hybrid-electric propul-
sion. A review of drive-by-wire system benefits in performance enhancements and vehicle active safety is
then discussed. This is followed by in-depth coverage of technological challenges that must be overcome
before drive-by-wire systems can be production ready. Current state of the art of possible solutions to
these technological hurdles is then discussed. Future trends in the drive-by-wire systems and economic
and commercialization aspects of these system are presented at the conclusion of the chapter.

Chapter III
Electromagnetic Compatibility Issues in Automotive Communications............................................... 48
Todd H. Hubing, Clemson University International Center for Automotive Research, USA
This chapter reviews automotive EMC requirements and discusses the design of automotive electron-
ics for EMC. The objective of the chapter is to provide non-EMC engineers and engineering managers
with basic information that will help them recognize the importance of designing for electromagnetic
compatibility, rather than addressing electronic noise problems as they arise.

Chapter IV
Automotive Network Architecture for ECUs Communications............................................................ 69
Fabienne Nouvel, Laboratory IETR-UMR, INSA, France
Wilfried Gouret, Laboratory IETR-UMR, INSA, France
Patrice Mazério, Laboratory IETR-UMR, INSA, France
Ghais El Zein, Laboratory IETR-UMR, INSA, France

This chapter introduces the most widely used automotive networks like LIN (Local Interconnect Net-
work), CAN (Controller Area Network), MOST (Media-Oriented Systems Transport), and FlexRay.
To fulfill the increasing demand of intra-vehicle communications, a new technique based on power
line communication (PLC) is then proposed. This allows the transmission of both power and messages
without functional barriers. On the other hand, there are several infotainment applications (like mobile
phones, laptop computers) pushing for the adoption of intra-vehicle wireless communications. Thus,
some potential wireless technologies used in the automotive domain, namely Bluetooth, IEEE 802.11
b/g wireless technology – WiFi, and Zigbee are covered here. Finally, the chapter highlights the chal-
lenges of these wired or wireless alternative solutions in automotive networks.

Chapter V
Enabling Secure Wireless Real-Time Vehicle Monitoring and Control................................................ 91
Lek Heng Ngoh, Institute for Infocomm Research, A*STAR, Singapore

In this chapter, the author extends the use of these embedded vehicular networks by proposing to remotely
monitor and control the vehicles through them, in order to realize safety and driver assistance related
applications. To accomplish this task, additional technologies such as real-time wireless communications
and data security are required, and each of them is introduced and described in this chapter.

Chapter VI
MAC and Routing Protocols for Vehicle to Vehicle Communications................................................ 105
Xiaobo Long, Rensselaer Polytechnic Institute, USA
Biplab Sikdar, Rensselaer Polytechnic Institute, USA

Numerous efforts are currently under progress to enhance the safety and efficiency of vehicular traffic
through intelligent transportation systems. In addition, the growing demand for access to data and infor-
mation from human users on the go has created the need for advanced vehicle-to-vehicle and vehicle-
to-roadside communication systems capable of high data rates and amenable to high degrees of node
mobility. Vehicular communications and networks are expected to be used for a number of purposes such
as for enabling mobile users to transfer data and information from other networks such as the Internet
and also for implementing services such as Intersection Decision Systems (IDS), Automated Highway
Systems (AHS), and Advanced Vehicle Safety Systems (AVS). In this chapter the authors describe me-
dium access control (MAC) and routing protocols for vehicular networks and the various factors that
affect their design and performance.

Chapter VII
Inter-Vehicular Communications Using Wireless Ad Hoc Networks.................................................. 120
Raúl Aquino-Santos, University of Colima, Mexico
Víctor Rangel-Licea, National Autonomous University of Mexico, Mexico
Miguel A. García-Ruiz, University of Colima, Mexico
Apolinar González-Potes, University of Colima, Mexico
Omar Álvarez-Cardenas, University of Colima, Mexico
Arthur Edwards-Block, University of Colima, Mexico
Margarita G. Mayoral-Baldivia, University of Colima, Mexico
Sara Sandoval-Carrillo, University of Colima, Mexico

This chapter proposes a new routing algorithm that allows communication in vehicular ad hoc networks.
In vehicular ad hoc networks, the transmitter node cannot determine the immediate future position of the
receiving node beforehand. Furthermore, rapid topological changes and limited bandwidth compound
the difficulties nodes experience when attempting to exchange position information.

Chapter VIII
The Role of Communications in Cyber-Physical Vehicle Applications.............................................. 139
Nicholas F. Maxemchuk, Columbia University, USA & IMDEA Networks, Spain
Patcharinee Tientrakool, Columbia University, USA
Theodore L. Willke, Columbia University, USA & Intel Corporation, USA

The authors describe applications that improve the operation of automobiles, control traffic lights, and
distribute the load on roadways. The requirements on the communications protocols that implement the ap-
plications are determined and a new communications paradigm, neighborcast, is described. Neighborcast
communicates between nearby entities, and is particularly well suited to transportation applications.

Chapter IX
Integrating Traffic Flow Features to Characterize the Interference in Vehicular Ad Hoc
Networks.............................................................................................................................................. 162
Lili Du, Purdue University, USA
Satish Ukkusuri, Rensselaer Polytechnic Institute, USA
Shivkumar Kalyanaraman, Rensselaer Polytechnic Institute, USA

The research in this chapter investigates several fundamental issues, such as the connectivity, the reach-
ability, the interference, and the capacity, with respect to information propagation in VANETs. The authors’
work is distinguished with previous efforts, since they incorporate the characteristics of traffic into these
issues in the communication layer of VANETs; this mainly address the issue of the interference. Previous
efforts to solve this problem only consider static network topologies. However, high node mobility and
dynamic traffic features make the interference problem in VANETs quite different.
Chapter X
Proactive Traffic Merging Strategies for Sensor-Enabled Cars........................................................... 180
Ziyuan Wang, University of Melbourne, Australia
Lars Kulik, University of Melbourne, Australia
Kotagiri Ramamohanarao, University of Melbourne, Australia

This chapter surveys traffic control strategies for optimizing traffic flow on highways, with a focus on
more adaptive and flexible strategies facilitated by current advancements in sensor-enabled cars and
vehicular ad hoc networks (VANETs). The authors investigate proactive merging strategies assuming
that sensor-enabled cars can detect the distance to neighboring cars and communicate their velocity and
acceleration among each other. Proactive merging strategies can significantly improve traffic flow by
increasing it up to 100% and reduce the overall travel delay by 30%.

Chapter XI
The Localisation Problem in Cooperative Vehicle Applications......................................................... 200
Javier Ibañez-Guzmán, RENAULT S.A.S., France

In this chapter, V2V and V2I applications are considered as a spatio-temporal problem. The tenet is that
sharing information can be made only if this is time stamped and related to a spatial description of the in-
formation sources. The chapter formulates the spatio-temporal problem having as constraint the precision
of the pose estimates of the vehicles involved. It regards the localisation problem and accuracy of digital
road maps as a combined issue that needs to be addressed for the successful deployment of cooperative
vehicle applications. Two case studies, intersection safely and an overtaking manoeuvre are included.
Recommendations on the precision limits of the vehicle pose estimations and the potential uncertainties
that need to be considered when designing V2V and V2I applications complete the chapter.

Chapter XII
An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle
Navigation Systems............................................................................................................................. 219
Denis Gingras, Université de Sherbrooke, Canada

In this chapter, the authors will review the problem of estimating in real-time the position of a vehicle
for use in land navigation systems. After describing the application context and giving a definition of the
problem, they will look at the mathematical framework and technologies involved to design positioning
systems. The authors will compare the performance of some of the most popular data fusion approaches
and provide some insights on their limitations and capabilities.

Chapter XIII
Efficient and Reliable Pseudonymous Authentication......................................................................... 247
Giorgio Calandriello, Politecnico di Torino, Italy
Antonio Lioy, Politecnico di Torino, Italy

Privacy, security, and reliability are key requirements in deploying vehicular ad-hoc networks (VANET).
Without those the VANET technology will not be suitable for market diffusion. In this chapter, the au-
thors are concerned with how to fulfill these requirements by using pseudonym-based authentication,
designing security schemes that do not endanger transport safety while maintaining low overhead. At
the same time the design improves the system usability by allowing nodes to self-generate their own
pseudonyms.

Chapter XIV
Simulation of VANET Applications.................................................................................................... 264
Valentin Cristea, University Politehnica of Bucharest, Romania
Victor Gradinescu, University Politehnica of Bucharest, Romania
Cristian Gorgorin, University Politehnica of Bucharest, Romania
Raluca Diaconescu, University Politehnica of Bucharest, Romania
Liviu Iftode, Rutgers University, USA

This chapter systematically presents actual issues regarding the simulation of VANET applications.
Some of them refer to challenges in developing VANET simulators. The chapter discusses simulator
architectures, models used for representing the communication among vehicles, vehicles mobility fea-
tures, and simulation tool implementation methods. A critical analysis of the solutions adopted in some
well-known actual simulators is also included.

Chapter XV
In-Vehicle Network Architecture for the Next-Generation Vehicles................................................... 283
Syed Masud Mahmud, Wayne State University, USA

This book chapter describes a number of ways using which the networks of future vehicles could be
designed and implemented in a cost-effective manner. The book chapter also shows how simulation
models can be developed to evaluate the performance of various types of in-vehicle network topologies
and select the most appropriate topology for given requirements and specifications.

Compilation of References ............................................................................................................... 303

About the Contributors .................................................................................................................... 330

Index.................................................................................................................................................... 338
 xiii

Foreword

Huaqun Guo has introduced the emerging areas of vehicular networks in the forms of Intra-Vehicle, Vehicle-
to-Vehicle, and Vehicle-to-Infrastructure communications and edited this new book to reflect the advance
information technologies that shape the modern automobiles. These new technologies on automotive infor-
matics and communicative systems will enable a variety of applications for safety, traffic efficiency, driver
assistance, as well as infotainment to be incorporated into modern automobile designs.
Over the last century, the design, manufacture and operation of the automobile have grown into complex
system integration paradigms cutting across applications of traditional disciplines in physical sciences, en-
gineering, social and behavioral sciences and business. Today, this complexity is compounded and acceler-
ated by the advent of enabling technologies in advanced materials, sensing, actuation, computing, controls,
diagnostics, electronics and software, all amid myriad – and often conflicting – policy changes. This creates
new possibilities and challenges in simultaneously providing effective means of transportation - with a high
degree of driver and occupant safety - along with reduced energy use and environmental impact.
Informatics, telematics, electronics and communication systems play an ever increasing role in the ad-
vancement of the automobile and are critical from a number of perspectives. The advances that are most
easily noticed by a consumer are vehicle options such as infotainment systems, navigation systems, and con-
nectivity such as Bluetooth. However, other onboard systems such as active stability control, engine control,
and the several supporting in-vehicle communication networks and protocols are the real technologies that
are propelling the automobile into the 21st Century. Such systems are key elements in achieving the desired
operational characteristics of the vehicle such as performance, emissions, safety and fuel efficiency. To achieve
these ever more stringent desired characteristics in a cost effective manner, the amount of information and
processing that occurs on a typical vehicle is staggering. Most vehicles today have well over 50 processors
on board, and the number continues to grow. Indeed, electronics can account for over 40% of the vehicle’s
cost and this percentage will continue to grow.
Onboard systems are only part of the explosion of automotive informatics and commutations. Infrastruc-
ture -to-vehicle and vehicle-to-vehicle communicants are enabling a host of new frontiers related to safety,
traffic control and maintenance. Onboard navigation systems can now route an individual vehicle through
significant traffic jams or disruptions. However, in the near future, coordinated efforts between the traffic
infrastructure and multiple vehicles may distribute the traffic load to minimize congestion or the effects of
construction or a traffic accident. Furthermore, information from adjacent vehicles may be used to avoid
collisions. For example, vehicles that are rapidly decelerating on a highway might warn subsequent cars of
an impending “stopped traffic hazard.” From a maintenance perspective, connectivity has already enabled
the vehicle to communicate its health status and potential failures to service personnel. Such information is
not only critical to keep a vehicle functioning properly, but also enables fleet manufacturers to track potential
problems, and address them as rapidly as possible. Furthermore, this information can easily and rapidly be
utilized in improving next generation vehicles.
xiv

For both onboard systems and supporting infrastructure systems, the acceleration of technological change
is driving vehicle designers, manufacturers and consumers to rethink how the automobile is developed from
conceptualization to production to service to end of life. The rapidly changing electronics and informatics
sector has pushed vehicle system design and integration to a new level of agility. The consumer desires
state-of-the-art capabilities, and automobile producers no longer have several years to incorporate the latest
technology into their products. This is fostering a change in the way vehicles are designed and perceived.
Indeed, if one looks at the automobile, it is changing rapidly and the pace of change is ever increasing. The
car of today is vastly different from its predecessors of 30 or 40 years ago, and next generation vehicles will
continue to change dramatically driven by multiple issues of which many are related to informatics and com-
munication systems.
This book has provided fundamental principles, as well as practice, and new research/trend for vehicular
networks and advanced information technologies applied in the automotive area. First, this book presents the
impact of drive-by-wire systems on vehicle safety and performance, and electromagnetic compatibility issues
affecting automotive communications. It then introduces Intra-vehicle networks like LIN (Local Interconnect
Network), CAN (Controller Area Network), MOST (Media-Oriented Systems Transport), Flexray, power-
line communication, and so forth. It also describes in-vehicle network architecture for the next-generation
vehicles and elaborates the potential applications and related technical challenges in achieving secure remote
monitoring and control of vehicles via CAN.
Second, this book presents the technologies related to Vehicle-to-Vehicle, and Vehicle-to-Infrastructure
communications by describing the current medium access control (MAC) and routing protocols for vehicular
networks, and the role of communications in cyber-physical vehicle applications. Furthermore, it incorporates
the characteristics of traffic flow into the interference issue in the communication layer of VANETs (Vehicular
Ad Hoc Networks), and presents new research into proactive traffic merging algorithms and the potential
benefits of applying sensor-enabled cars. The book has also captured the state-of-the-art in the area of traffic
control with the assistance of VANETs, and reviewed the problem of estimating in real-time the position of
a vehicle for use in land navigation system.
Last but not least, privacy, security and reliability as key requirements in deploying VANETs are addressed,
as well as simulation architectures and simulation tools implementation methods with the aim to improve the
traffic safety and control. Through all chapters, this book has discussed the future trends for the automotive
informatics and communicative systems in each individual domain.
I highly recommend Dr. Guo’s timely book. I believe it will benefit many readers and be a good refer-
ence.

Thomas R. Kurfess
International Center for Automotive Research, Clemson University, USA

Thomas R. Kurfess received his SB, SM and PhD degrees in mechanical engineering from M.I.T. in 1986, 1987 and 1989,
respectively. He also received an SM degree from MIT in electrical engineering and computer science in 1988. Following graduation,
he joined Carnegie Mellon University where he rose to the rank of associate professor. In 1994 he moved to the Georgia Institute of
Technology where he rose to the rank of Professor in the George W. Woodruff School of Mechanical Engineering. In 2005 he was
named Professor and BMW Chair of Manufacturing in the Department of Mechanical Engineering at Clemson University. He is also
the Director of the Campbell Graduate Engineering Center at Clemson University’s International Center for Automotive Research.
He has served as a special consultant of the United Nations to the Government of Malaysia in the area of applied mechatronics and
manufacturing, and as a participating guest at the Lawrence Livermore National Laboratory in their Precision Engineering
Program. His research focuses on the design and development of advanced systems targeting the automotive sector (OEM
and supplier) including vehicle and production systems. He has signi.cant experience in high precision manufacturing and
metrology systems. He has received numerous awards including a National Science Foundation (NSF) Young Investigator
Award, an NSF Presidential Faculty Fellowship Award, the ASME Pi Tau Sigma Award, SME Young Manufacturing Engineer
of the Year Award, the ASME Blackall Machine Tool and Gage Award, the ASME Gustus L. Larson Award. He is a Fellow of
the SME and of the ASME.
 xv

Preface

The automotive industry is undergoing a continuous transformation; vehicles are no longer thermo me-
chanical systems with some electronic components used to start engines and lighting. Today’s vehicles
are complex systems, with networks of computers controlling their most important functions. Increas-
ing fuel costs, as well as increasing awareness of vehicular pollution and noise affecting large human
agglomerations and unacceptable numbers of traffic accidents and road congestion are exerting much
pressure for change on the automotive industry. What kind of change is expected?
Within a short period, mobile communications have changed our lifestyles allowing us to exchange
information, almost anywhere at anytime. The introduction of such mobile communications systems
in motor vehicles should be therefore only a matter of time. This should bring a new paradigm, that of
sharing information amongst vehicles and infrastructure, and lead to numerous applications for safety,
traffic efficiency as well as infotainment.
The main purpose of this book is to provide an overview of the information and communications
technologies that are to be deployed in the new generations of vehicles – to provide valuable insights
into the technologies for vehicular networks and data exchange, from both theoretical and practical
perspectives. We hope that the contents can be used in graduate level courses as a reference and by the
automotive industry as training material. The book should provide a concise background and a good
foundation to students entering the field of automotive information and communications technologies.
We also hope that it would serve as a reference to researchers/scientists and practitioners by enabling
them to offer exciting and novel technologies and applications that would, in the future, transform our
land transportation systems.
Information technology is the driving force behind innovations in the automotive industry. In the
past years, control systems of cars have moved from the analog to the digital domain. In particular, x-
by-wire systems began to appear, and have driven research efforts of the whole automotive industry in
the last decade. Networked Electronic Control Units (ECUs) are increasingly being deployed in cars to
realize diverse functions such as engine management, air-bag deployment, and even in intelligent brake
systems. At the same time, emerging vehicular networks in the forms of intra-vehicle, vehicle-to-vehicle
and vehicle-to-infrastructure communications are fast becoming a reality. They will enable a variety of
applications for safety, traffic efficiency, driver assistance, as well as infotainment to be incorporated
into modern automobile designs.
This book introduces the advanced information technologies that shape the ultra-modern automo-
tive industry today. Contributions to this publication are made by professors, researchers, scientists
and practitioners throughout the world, bringing together their rich expertise and results of their cur-
rent endeavors. The authors have several years of expertise in their respective domains, and have good
publication records.
xvi

As can be seen from the table of contents, the book comprises of 15 chapters. It spreads across many
technical areas with car communications as the central theme. The first chapter is an introductory chap-
ter on the emerging area of vehicular networks in the forms of Intra-Vehicle (InV), Vehicle-to-Vehicle
(V2V), and Vehicle-to-Infrastructure (V2I) communications. Chapters II to III and a large part of Chapter
V cover the technologies related to InV networks. The rest of Chapter V and Chapters VI to XII present
the technologies related to V2V and V2I communications which will enable a variety of applications
for safety, traffic efficiency, driver assistance and infotainment. Privacy, security, and reliability as key
requirements in deploying VANETs (Vehicular Ad Hoc Networks) are addresses in Chapter XIII. Chapter
XIV discusses simulation architectures, models used for representing the communication among vehicles,
vehicle mobility features, and ways to implement simulation tools with the aim to improve traffic safety
and control. Chapter XV describes in-vehicle network architectures for the next-generation vehicles.
Chapter-wise details are presented bellow.
The introductory chapter presents the emerging area of vehicular networks in the forms of Intra-Ve-
hicle (InV), Vehicle-to-Vehicle (V2V), and Vehicle-to-Infrastructure (V2I) communications. This will
enable a variety of applications for safety, traffic efficiency, driver assistance, as well as infotainment, to
be incorporated into modern automotive designs. Critical data is being exchanged within a vehicle and
with outside the vehicle via vehicular networks. Thus, this chapter first introduces car communications,
potential vehicular applications and wireless technologies, as well as specially designed technologies
DSRC (Dedicated Short Range Communications) standards and communication stack for data exchange.
As the emerging area of vehicular networks is attracting widespread interest from research groups around
the world, this chapter next introduces the consortiums and initiatives working on advanced automotive
technologies in Europe, the United States, Japan, and Singapore. Finally, in the future trend, vehicular
networks still plays a vital role in enhancing the automotive industry for safety, security and entertain-
ment.
Chapter II presents the impact of drive by wire systems on vehicle safety and performance. An
overview of the drive-by-wire technology is presented along with in-depth coverage of salient drive by
systems such as throttle-by-wire, brake-by-wire, and steer-by-wire systems, and hybrid-electric propul-
sion. This is followed by in-depth coverage of technological challenges and the current state-of-the-art
solutions to these technological hurdles. For example, an analytical redundancy/model-based fault-
tolerant control can not only reduce the overall system cost by reducing the total number of redundant
components, but also further improve overall reliability of the system through the usage of a diverse
array of sensory information. Future trends in the drive-by-wire systems include various drive-by-wire
systems in the same vehicle sharing a diversity of sensors and actuators via data fusion methodologies,
integrated control of various drive by wire systems and future communication bus for x-by-wire systems,
for example FlexRay,
Chapter III provides a basic overview of the electromagnetic compatibility (EMC) issues affecting
automotive communications. As the number of electronic systems in automobiles rises, the potential
for electromagnetic interference increases. Designing for electromagnetic compatibility is important to
devote proper attention to electromagnetic compatibility at every stage of an automobile’s development.
Problems discovered late in the design cycle can seriously impact development schedules and product
cost. This chapter provides basic information of electromagnetic compatibility issues affecting automo-
tive communications for non-EMC engineers and engineering managers who work with automotive
networks.
Chapter IV introduces the most widely used automotive networks like LIN (Local Interconnect
Network), CAN (Controller Area Network), MOST (Media-Oriented Systems Transport), and FlexRay.
To fulfill the increasing demand of intra-vehicle communications, a new technique based on power
 xvii

line communication (PLC) is then proposed. This allows the transmission of both power and messages
without functional barriers. On the other hand, there are several infotainment applications (like mobile
phones, laptop computers) pushing for the adoption of intra-vehicle wireless communications. Thus,
some potential wireless technologies used in the automotive domain, namely Bluetooth, IEEE 802.11
b/g wireless technology – WiFi, and Zigbee are covered here. Finally, the chapter highlights the chal-
lenges of these wired or wireless alternative solutions in automotive networks.
Chapter V elaborates one of the most popular in-vehicle networking technologies called Controller
Area Network (CAN). The chapter begins with an overview of the basis and the general technology of
CAN in automotive industry and the deployment of in-vehicle CAN networks. It then presents the vari-
ous existing and future potential applications that make use of the CAN data, and the related technical
challenges in achieving secure remote monitoring and control of vehicles via CAN. Furthermore, the
chapter elaborates two key components in achieving remote vehicle monitoring and control, namely, the
wireless communication component and data security component. It stresses the importance of secure
data and information flow between vehicles and an application server. Finally, the chapter presents an
overall architecture for secure wireless real-time vehicle monitoring and control environment. In future
rends, the author foresees the area of real-time monitoring and control being a fertile ground for future
automotive innovations and services.
Chapter VI describes the current medium access control (MAC) and routing protocols for vehicular
networks, and the various factors that affect their design and performance. The mobility and speed of the
communicating nodes in vehicular networks add extra dimensions to the challenges faced by the MAC
protocols, in addition to the existing requirements of reliability and efficiency. This chapter reviews
some of the existing MAC protocols for vehicular network. For example, basic MAC protocols, the
IEEE 802.11 MAC Extension for Vehicular Networks, and other MAC Protocols for Vehicular Networks
(ADHOC-MAC, the Directional MAC (D-MAC) protocol). Future Intelligent Transportation Systems
require fast and reliable communication between cars (vehicle-to-vehicle) or between a car and a road
side unit (vehicle-to-infrastructure). Ad hoc unicast routing schemes can be divided into two categories:
topology-based routing and position-based routing. Topology-based schemes use a variety of proac-
tive routing schemes (DSDV ((Destination Sequenced Distance Vector routing), Optimized Link State
Routing (OLSR), Fisheye State Routing (FSR)) or reactive approaches (AODV (Ad Hoc On-Demand
Distance Vector Routing), DSR ((Dynamic Source Routing), Temporally Ordered Routing Algorithm
(TORA), Associativity Based Routing Algorithm (ABR), or hierarchical protocols (Cluster Based
Routing Protocol (CBRP), Core Extraction Distributed Ad-hoc Routing (CEDAR) and Zone Routing
Protocol (ZRP)) to create routes.
Popular location services in position-based routing protocols are Distance Routing Effect Algorithm
for Mobility (DREAM) and Grid Location Service (GLS). Context Assisted Routing (CAR) and Spatially
Aware Routing (SAR) are proposed routing algorithms to overcome the problem of topology holes in
position-based routing. Current multicast protocols that can be used in V2V networks include: Posi-
tion-based Multicast (LBM), GeoGRID, Unicast Routing with Area Delivery and Inter-Vehicle Geocast.
Overall, routing of communications for vehicular safety applications remains a challenging topic.
Chapter VII presents a new reactive algorithm based on location information in the context of
vehicular ad-hoc networks. It proposes a Location Routing Algorithm with Cluster-Based Flooding
(LORA-CBF), which is formed with one cluster head, zero or more members in every cluster, and one
or more gateways to communicate with other cluster heads. It first validates the model at one, two, and
three hops by comparing the results of the test bed with the results of the model developed in OPNET.
For more than three hops, it validates the model by comparing with two non-position-based routing
algorithms (AODV and DSR) and one position-based routing algorithm (GPSR (Greedy Perimeter
xviii

Stateless Routing)). Results show that mobility and network size affects the performance of AODV and
DSR more significantly than LORA_CBF and GPSR. It is also observed that GPSR and LORA-CBF
behave similarly in terms of the end-to-end delay, and LORA_CBF is more robust in terms of delivery
ratio, routing overhead, route discovery time, and routing load compared with GPSR.
Chapter VIII presents the role of communications in cyber-physical vehicle applications. Cyber-physical
systems use sensing, communications and computing to control the operation of physical devices. The
embedded computers and sensors both within the vehicles and in the infrastructure will be networked
into cyber-physical systems to reduce accidents, improve fuel efficiency, increase the capacity of the
transportation infrastructure, and reduce commute time. Communications between nearby vehicles will
enable cooperative control paradigms that reduce accidents more than computing and sensors alone, and
communications between vehicles and the infrastructure will improve the scheduling of traffic signals
and route planning. The chapter describes applications that improve the operation of automobiles, control
traffic lights and distribute the load on roadways. The requirements on the communications protocols
that implement the applications are determined and a new communications paradigm, neighborcast,
is described. Neighborcast communicates between nearby entities, and is particularly well suited to
transportation applications.
Chapter IX incorporates the characteristics of traffic flow into the interference issue at the com-
munication layer of VANETs. There are several fundamental issues, such as connectivity, reachability,
interference and capacity, with respect to information propagation in VANETs. This chapter mainly
addresses the issue of interference, by incorporating the characteristics of traffic into this issue at the
communication layer of VANETs. High node mobility and dynamic traffic features make the interfer-
ence problem in VANETs quite different. As compared with previous efforts to solve this problem
which only considered static network topologies, this work is (to the best of our knowledge), the first to
demonstrate the interference features in VANETs by incorporating realistic traffic flow characteristics
based on a validated simulation model. Analytical expressions are developed to evaluate the interfer-
ence in VANETs taking account of both the macroscopic and the microscopic traffic flow characteristics.
These analytical expressions are validated within the simulation framework. The results show that the
analytical characterization performs very well to capture the interference in VANETs. The results from
this work can facilitate the development of better algorithms for maximizing throughput in VANETs,
and the research efforts bridging the features of both the communication layer and the transportation
layer will help to build more efficient systems.
Chapter X first captures the state-of-the-art in the area of traffic control with the assistance of
VANETs in terms of vehicular traffic models, vehicular traffic theories, flow control strategies, and
performance measurement methodologies. It surveys traffic control strategies for optimizing traffic flow
on highways, with a focus on more adaptive and flexible strategies facilitated by current advancements
in sensor-enabled cars and VANETs. It provides an overview of new ideas and approaches in the area
of traffic flow control with the assistance of VANETs. This chapter then presents new research into
proactive traffic merging strategies and the potential benefits of applying sensor-enabled cars. It shows
how sensor-enabled cars can assist in improving merging algorithms, and compares proactive merg-
ing algorithms against a conventional merging strategy: priority-based merging. Assisted by advanced
sensing and communication technologies, traffic control strategies and merging algorithms will lead
to more efficient use of the current road networks and ultimately help to alleviate traffic congestion. It
has shown that the significant improvement in traffic flow and the decrease in travel time mainly result
from the decoupling of the merging point and the decision point, and multilane optimizations, such as
pre-lane-changing. Proactive merging strategies can significantly improve traffic flow by increasing it
by up to 100% and reduce overall travel delay by 30%.
 xix

Chapter XI presents the localization problem in cooperative vehicle applications by focusing on the
constraints imposed by the need for precise vehicle localization estimates. Such accurate estimates are
a pre-requisite to deploying vehicles with adequate communications capabilities in real traffic condi-
tions. V2V and V2I safety applications are complex, and the problem does not limit to communications
capabilities. It involves spatial information with respect to the vehicles’ locations relative to each other
and the infrastructure. Thus, V2V and V2I applications are considered as a spatio-temporal problem. The
tenet is that information can be shared only if this is time stamped and related to a spatial description
of the information sources. The chapter formulates the spatio-temporal problem having as constraint
the precision of the pose estimates of the vehicles involved. It formulates the localization problem and
accuracy of digital road maps as a combined issue that needs to be addressed for the successful deploy-
ment of cooperative vehicle applications. The problem formulation is completed by two case studies,
the use of V2V or V2I communications to traverse safely an intersection and an overtaking manoeuvre.
The chapter concludes by including comments and recommendations on the precision limits of the ve-
hicle pose estimations and the potential uncertainties that need to be considered when designing V2V
and V2I applications.
Chapter XII reviews the problem of estimating (in real-time) the position of a vehicle for use in land
navigation systems. After describing the application context and giving a definition of the problem, it
looks at the mathematical framework and technologies involved in the design of positioning systems.
Through a review of some of the various sensor fusion techniques usually encountered in such sys-
tems, it compares the performance of some of the most popular data fusion approaches, and provides
some insights on their limitations and capabilities. The extended Kalman filter (EKF) in data fusion
centralized architectures remains a design of choice for most applications. The chapter then describes
how to make positioning systems more robust and adaptive by detecting and identifying sensor faults.
Finally, it explores possible architectures for collaborative positioning systems, where many vehicles
are interacting and exchanging data to improve their own position estimate using a collaborative and
geometric data fusion approach. One major trend seen in the field of dense sensor networks is in the use
of multilateration techniques for location accuracy. Despite significant errors in range estimates between
sensors, multilateration is able to render more accurate location estimates, thus making it suitable for use
in vehicle navigation. With the current evolution of automotive technologies, all vehicles are becoming
networked and equipped with wireless communication capabilities, thus allowing the use of distributed
and collaborative techniques for navigation and positioning. Wireless communications networks are
becoming attractive to localize vehicles using various radio-based range technologies such as received
signal strength indicators (RSSI), power signal attenuation or time-of-arrival (TOA) techniques.
Privacy, security, and reliability as key requirements in deploying VANETs are addressed in Chapter
XIII. Without these strengths, the VANET technology will not be suitable for market diffusion. This
chapter concerns with how to fulfill these requirements by using pseudonym-based authentication,
and designing security schemes that do not endanger transport safety while maintaining low overhead.
At the same time, the design improves system usability by allowing nodes to self-generate their own
pseudonyms. It manages security credentials in VANET through self-generation and self-certification
of pseudonyms, which greatly simplifies the security management and makes a step towards a usable
system. It employs group signatures to generate certificates which satisfy the requirements of anonymity
and liability attribution, and results show that the computational cost and the overhead are comparable
to the baseline approach. Next, it analyzes the costs imposed by security on the transportation systems
by analyzing data link performance to obtain packet reception probability curves for the pseudonym-
based security systems, and analyzing the impact of safety messaging, security and privacy-enabling
technologies on transportation safety to show that secure communication schemes achieve safety levels
xx

comparable to those with no security at all. This chapter performs a detailed investigation of pseudonym-
based authentication by analyzing several system issues and showing how these security mechanisms
can be applied in practice.
Chapter XIV systematically presents actual issues faced by developers and engineers in the simulation
of VANET applications, some of which are related to the challenges in developing VANET simulators.
It discusses simulation architectures, models used for representing the communication among vehicles,
vehicles mobility features, and simulation tools implementation methods. The focus is on the new trends
in communication protocols and traffic models, and on new facilities incorporated in simulation tools.
Advances in VANET technology and protocols support the adoption and use of more complex mobil-
ity models and of more flexible and adaptable traffic controls. VANETs’ rapid topology changes or the
changes in the vehicles mobility as reaction to traffic changes are captured by the simulation models,
which become more or less complicated and include more elements that constrain vehicle mobility: maps,
real traffic conditions (congestion), driver behavior, fuel consumption, pollutant emissions, and so forth.
It also includes a critical analysis of the solutions adopted in some well-known actual simulators. Other
issues related to the use of simulation in the evaluation of applications that aim at improving traffic safety
and control are discussed. Representative city and highway application scenarios are analyzed, and results
obtained by simulation, along with ways these results can be exploited by VANET developers and users,
are highlighted. Future trends in the development of simulators that produce more accurate results, and
their use for the evaluation of more sophisticated traffic control solutions, are also included.
Chapter XV takes a more futuristic look at various types of topologies and protocols that could be
used specifically in in-vehicle networks. Varying functionalities of vehicles will require different types
of communication networks and networking protocols. As the size and complexity of the network grows,
integration, maintenance and troubleshooting will become a major challenge. To facilitate integration and
troubleshooting of various nodes and networks, it would be desirable that networks of future vehicles
be partitioned, and the partitions be interconnected by a hierarchical or multi-layer physical network.
These partitions must be appropriately interconnected to handle functional dependencies and for better
diagnostics. A number of network topologies have been presented and analyzed for cost, bandwidth and
message latencies. This chapter describes a number of ways using which the networks of future vehicles
could be designed and implemented in a cost-effective manner. Since future vehicles will also be com-
municating with external entities for various reasons, the chapter also addresses the issues of security,
safety and privacy which should be taken into consideration at the time of designing the in-vehicle net-
work components. Finally, some ideas have been presented in developing simulation models to analyze
various types of networks which will ultimately help in selecting the most appropriate network topology
and various network components for a given set of requirements and specifications.
Thus, we have walked through the world of new information and communication technologies be-
ing developed for vehicular systems. We hope that the book is of interest to academia and industry. We
earnestly hope that the insights provided by this book, on the specific information and communication
technologies used in vehicles, will help inspire and spawn a multitude of novel applications and in-
novations.

This book is dedicated to my parents Lanying Guo and Tianfu Guo.

Huaqun Guo
Singapore,October 2008
 xxi

Acknowledgment

I would like to extend my utmost gratitude to the people who have, in one way or other, inspired, aided
and contributed to the successful completion of this book.
First of all, I would like to express my sincere gratitude to Deputy Executive Director (Research)
Professor Lye Kin Mun at the Institute for Infocomm Research (I2R) for his strong encouragement of
my editing works.
I would also like to thank the A*CAR (A*STAR Capabilities for Automotive Research) taskforce,
and the management of Institute for Infocomm Research (I2R) for providing me the opportunity to carry
out research into the exciting area of vehicular networks. In particular, I would like to thank Dr. Feng
Bao and Dr. Yongdong Wu at I2R for their support.
Special thanks to all reviewers for their expertise, time, effort and timely response throughout the
peer evaluation process. In particular, I wish to express my appreciation to the members of the Editorial
Advisory Board for their guidance, support and constant encouragement.
I would also like to take this opportunity to thank Tyler Heath and Heather A. Probst for their as-
sistance with this book.
Last but not least, the heartiest gratitude is given to my family for their love and encouragement.

Huaqun Guo
Institute for Infocomm Research, A*STAR, Singapore
 

Chapter I
Introduction:
An Emerging Area of Vehicular Networks
and Data Exchange

Huaqun Guo
Institute for Infocomm Research, A*STAR, Singapore

ABSTRtract

Emerging vehicular networks in the forms of Intra-Vehicle (InV), Vehicle-to-Vehicle (V2V), and Vehicle-
to-Infrastructure (V2I) communications will enable a variety of applications for safety, traffic efficiency,
driver assistance, as well as infotainment to be incorporated into modern automobile designs. At the
same time, networked Electronic Control Units (ECUs) are increasingly being deployed in automobiles to
realize functions such as engine management, air-bag deployment, and even in intelligent brake systems.
In addition, users now expect to sit in an automobile and have their brought-in devices, and beamed-in
services harmoniously integrated with the built-in interfaces inside the automobile. Thus, widespread
adoption of vehicular networks is fast becoming a reality and critical data is being exchanged with-inside
and with-outside vehicle via vehicular networks. This chapter gives an overview of this emerging area
of vehicular networks, its potential applications, its potential wireless technologies for data exchange,
and its research activities in the Europe, the United States (U.S.), Japan, and Singapore.

INTRODUCTION increasingly being deployed in automobiles to

In recent years, control systems of automobiles
have moved from the analog to the digital do-
main. In particular, x-by-wire systems are ap-
pearing and have driven research efforts of the
whole automotive industry for the recent decade.
Networked Electronic Control Units (ECUs) are
realize diverse functions such as engine manage-
ment, air-bag deployment, and even in intelligent
brake systems. For example, at least 70 networked
ECUs are employed in a Mercedes S-Class car
(Heffernan & Leen, 2008; Vasilash, 2005). At the
same time, emerging vehicular networks in the
forms of Intra-Vehicle (InV), Vehicle-to-Vehicle

Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.

(V2V), and Vehicle-to-Infrastructure (V2I) com-
munications are fast becoming a reality and will
enable a variety of applications for safety, traffic
efficiency, driver assistance, as well as infotain-
ment to be incorporated into modern automobile
designs.
There are currently a number of study groups
working on car communications and defin-
ing the standards for various applications. InV
Communications, such as CAN (Controller Area
Network, 2008; CiA, 2008), LIN (Local Inter-
connect Network, 2008), FlexRay (2008), are
used for interconnecting in-car ECUs, sensors,
and so on. V2V Communications, such as IEEE
802.11p (IEEE 802.11p, 2008; Jiang & Delgrossi,
2008), Dedicated Short Range Communications
(DSRC) (Dedicated Short Range Communica-
tions, 2008), may be used for safety applications.
V2I communications, e.g. IEEE 802.11p and IEEE
1609 Family of Standards for Wireless Access in
Vehicular Environments (WAVE, 2008) may be
used for traffic information.
In addition, users now expect to sit in an au-
tomobile and have their brought-in devices and
beamed-in services harmoniously integrated with
the built-in interfaces inside the automobile. To
integrate mobile phones and digital music players,
Ford designs Ford Sync that integrates voice-ac-
tivated in-car communication and entertainment
system (Ford Sync, 2008). RM MICHAELIDES
provides wireless CAN interfaces to transmit
CAN between different networks using Bluetooth,
RFID (radio-frequency identification), Infrared,
UHF (ultra high frequency), etc (Michaelides,
2008). A Controller Area Network Gateway to
ZigBee was described in (Kuban, 2007). There
are some wireless CAN products, such as CANRF
(Dammeyer, 2008) and CAN Bridge (Matric,
2008). The performance of wireless CAN in terms
of latency and throughput was studied in (Dridi,
Gouissem, Hasnaoui, & Rezig, 2006).
Thus, with vehicular networks fast becoming
commonplace, critical data is being exchanged
with-inside and with-outside vehicle via vehicu-

Introduction
lar networks, and new technologies have been
developed for vehicular networks. This chapter is
meant to introduce the emerging area of vehicular
networks and data exchange, give an overview of
the new technologies for car communications,
and present automotive research activities in the
Europe, the United States (the U.S.), and Japan
as well as in Singapore.
CAR COMMUNICATIONS
New technologies are being developed for ve-
hicular networks and these networks provide an
efficient method for today’s complex car com-
munications. Figure 1 shows the example of InV,
V2V and V2I communications.
InV provides communication among ECUs/
sensors in a vehicle while V2V and V2I provide
communications among nearby vehicles and be-
tween vehicles and nearby fixed roadside equip-
ments. Vehicular networks are a cornerstone of
the envisioned Intelligent Transportation Systems
(ITS). By enabling vehicles to communicate with
its function systems via InV communication,
with other vehicles via V2V communication as
well as with roadside base stations via V2I com-
munication, vehicular networks will contribute
to safer and more efficient roads by providing
timely information to drivers and concerned
authorities.
Potential Applications
The emerging vehicular networks will enable a
variety of applications for safety, traffic efficiency,
driver assistance and infotainment:
1. Safety: Vehicular network technologies will
be applied to reduce accidents so as to save
lives and reduce injuries. Examples of such
applications include vehicle breakdown and
obstacle detection, lane departure warn-
ing, accident warnings, collision warning,
Introduction
Figure 1. Example of InV, V2V, and V2I communications
ECU
ECU
ECU
ECU
ECU
ECU
CAN BUS
ECU
over-height / over-width warning, turnover
warning, work zone warnings, black box,
and so on.
2. Traffic Efficiency: Vehicular network
technologies will be applied to improve
the flow of traffic and reduce congestion,
for example, cooperative adaptive cruise
control, highway/rail intersection traffic
management, congestion information for
traffic control, electronic toll collection,
etc.
3. Driver Assistance: Vehicular networks can
also provide accurate information and data,
as well as good communications for drivers
to improve safety and security, e.g., digital
road maps downloading, advanced navi-
gation system, parking information, real-
time traffic information, various warning
information, driver’s daily blog, automatic
emergency call, etc.
4. Infotainment: Vehicles in future are fore-
seen to be dominated with feature-rich au-
dio-video infotainment. Users can expect to
sit in an automobile and have their brought-in
devices and beamed-in services harmoni-
ously integrated with the built-in interfaces
inside the automobile. Furthermore, users
can also wirelessly purchase and synchro-
nize the latest movies and songs when they
top up fuel at the kiosk. Last but not least,
wide range of applications for entertainment
through Internet will be brought to automo-
tive passengers.
Potential Wireless Technologies
With the rapid development of information tech-
nologies, there are a number of wireless technolo-
gies which are potential for wireless InV, V2V and
V2I communications, and listed in Table 1 and
Table 2. These new technologies could be used
for data exchange between users’ devices and
vehicles, among vehicles, and between vehicles
and infrastructure. The details of application
scenarios of data exchange with each technology
are listed in two tables as well.
DSRC is the recent technological trends to
provide real time traffic information for effective
implementation of ITS. Thus, in the following
subsection, we focus on introducing the new
technologies of DSRC for vehicular networks.

 Introduction

Table 1. Wireless technologies for InV communications

ZigBee UWB Bluetooth Wireless USB Wireless CAN
(ultra-wide band) (Universal Serial Port)
Standard/ Ratified in Decem- Transmitting infor- First launched (1998) Short-range, high- CANRF (CAN over
Technology ber 2004 mation spread over bandwidth based RF)/
a large bandwidth on the WiMedia CAN Bridge
(>500 MHz) Alliance’s UWB
Coverage 10 and 75 meters < 60 cm for a 500 1 meter, 10 meters, 480 Mbit/s at up to 3 /
MHz wide pulse, < 100 meters meters and 110 Mbit/ 500 feet (152.4M)
23 cm for a 1.3 GHz s at up to 10 meters
bandwidth pulse
Bit Rate 20-250 kbit/s per extremely high data 3 Mbit/s (Version 2.0 480 Mbit/s at dis- 20kbps/
channel rates + EDR) tances up to 3 meters 52.8kbps-164.4kbps
1000+ Mbps 53-480 Mbps and 110 Mbit/s at up
(WiMedia Alliance to 10 meters
(proposed)
Applications Entertainment, Used at very low Connect and exchange Game controllers, Communication
smart Lighting energy levels for information between digital cameras, MP3 among sensors and
control, advanced short-range high- devices such as mobile players, hard disks ECUs
temperature control, bandwidth commu- phones, laptops, and flash drives.
safety & security, nications by using a personal computers, Also suitable for
sensors, etc larger portion of the video game consoles, transferring parallel
radio spectrum etc video streams.

Table 2. Wireless Technologies for V2V & V2I communications

GSM/3G WiFi WiMax DSRC
(Wi-Fi Alliance Version of
802.11n)
Standard/ Third generation cellular New Wi-Fi technology with Broadband technology in A short to medium range
Technology technology in 2001 MIMO standard in 2009, 2007 communications
802.11n standard in 2009
Coverage kilometers 500m 5 km 1000m
Bit Rate 2-3 Mbps 600Mbps using MIMO 75Mbps 6 to 27 Mbps
Applications Between Vehicle and Roadside to vehicle and Internet access, Email, VoIP Roadside to vehicle and
mobile phone communi- vehicle to vehicle com- (Voice over IP) vehicle to vehicle com-
cation munication munication

Dedicated Short Range communication zones are important (Armstrong,

Communications
DSRC is a short to medium range wireless pro-
tocol specifically designed for automotive use. It
supports both public safety and private operations
for V2V and V2I communication environments.
DSRC is a complement to cellular communica-
tions by providing very high data transfer rates in
circumstances where minimizing latency in the
communication link and isolating relatively small
2008). This technology for ITS applications is
working in the 5.9 GHz band (the U.S.) or 5.8
GHz band (Japan & Europe). DSRC standards
and communication stack are shown in Figure 2
(Jiang & Delgrossi, 2008).
IEEE 802.11p
IEEE 802.11p is a draft amendment to the IEEE
802.11 standard to add wireless access in the


Introduction
vehicular environment. It defines enhancements
to 802.11 required to support ITS applications.
This includes data exchange between high-speed
vehicles and between the vehicles and the roadside
infrastructure in the licensed ITS band of 5.9 GHz
(5.850-5.925 GHz) (IEEE 802.11p, 2008). IEEE
802.11p supports physical layer management
entity (PLME), lower MAC (Medium Access
Control) layer management entity (L_MLME),
Wireless Access in Vehicular Environments
physical layer (WAVE PHY) as well as WAVE
lower MAC in DSRC technology. The 802.11p
Task Group is still active, resolving comments
on Draft 3.0, and the final approval is expected
in March 2009.
IEEE 1609—Family of Standards for
Wireless Access in Vehicular
Environments (WAVE)
The WAVE standards define an architecture and
a complementary, standardized set of services
Figure 2. DSRC standards and communication stack
Safety Applications
WME
IEEE P1609.3
IEEE P1609.2
Security
U_MLME
IEEE P1609.4
L_MLME
IEEE 802.11p
PLME
IEEE 802.11p
and interfaces that collectively enable secure
V2V and V2I wireless communications. To-
gether these standards provide the foundation
for a broad range of applications in the trans-
portation environment, including vehicle safety,
automated tolling, enhanced navigation, traffic
management and many others. The IEEE 1609
Family of Standards for Wireless Access in Ve-
hicular Environments (WAVE) consists of IEEE
P1609.1—Standard for Wireless Access in Vehicu-
lar Environments (WAVE)—Resource Manager,
IEEE P1609.2—Standard for Wireless Access in
Vehicular Environments (WAVE)—Security Ser-
vices for Applications and Management Messages,
IEEE P1609.3—Standard for Wireless Access in
Vehicular Environments (WAVE)—Network-
ing Services, and IEEE P1609.4—Standard for
Wireless Access in Vehicular Environments
(WAVE)—Multi-Channel Operations (WAVE,
2008).
Non- Safety Applications
WSMP UDP TCP
IEEE 1609.3
IP
LLC
WAVE upper MAC
IEEE P1609.4
WAVE lower MAC
IEEE 802.11p
WAVE PHY
IEEE 802.11p


CONonortium and INITIAnitiati
Vehicular networks, as an emerging area, attract
a number of R&D groups in the Europe, the
U.S., Japan and Singapore to work on the new
technologies.
Car 2 Car Communication Consortium
The Car 2 Car Communication Consortium (2008)
is initiated by European vehicle manufacturers
and partners include Audi, BMW, Daimler, Fiat,
Honda, Opel, Renault, and Volkswagen. Its objec-
tive is to further increase road traffic safety and
efficiency by means of inter-vehicle communica-
tions. The schedule for the official agreements of
consortium is from July 2005 for the basic concept
till December 2010 for frequency allocation.
Network on Wheels, Germany
Network on Wheels (NoW) (2008) was founded
in 2004 and the current partners include Daimler
AG, BMW AG, Volkswagen AG, Fraunhofer In-
stitute for Open Communication Systems, NEC
Deutschland GmbH, IMST GmbH and embedded
wireless GmbH. Besides the partners, the Uni-
versities of Mannheim, Karlsruhe and Munich
and the Carmeq GmbH also cooperate within
NoW. NoW is a German research project which
is supported by Federal Ministry of Education
and Research. Its main objectives are to solve
technical problems on communication protocols
and data security for car-to-car communica-
tions and to submit the results to the Car 2 Car
Communication Consortium. On May 8, 2008,
NoW presented its results in a final workshop at
the Daimler Research & Development Center in
Ulm (Germany).
SAFESPOT
SAFESPOT (2008) integrated research project
was co-funded by the European Commission
Information Society Technologies under the

Introduction
initiatives of the 6th Framework Program. The
objective is to understand how intelligent vehicles
and intelligent roads can cooperate to produce a
breakthrough for road safety based on V2V and
V2I communications.
eSafety
eSafety (2008), the first pillar of the Intelligent
Car Initiative (i2010 Intelligent Car Initiative,
2008), brings together the European Commission,
public authorities, industry and other stakehold-
ers with an aim to accelerate the development,
deployment and use of Intelligent Vehicle Safety
Systems that use information and communication
technologies. The main target is to contribute to
the European Commission’s 2001 goal of reducing
the road fatalities by 50% by 2010 (from 54 000
to 27 000 – between 2001 and 2010) (European
Commission, 2008).
PReVET
PReVENT (2008) is a European automotive in-
dustry activity co-funded by the European Com-
mission to contribute to road safety by developing
and demonstrating preventive safety applications
and technologies. Membership in its Core Group
consists of seven vehicle manufacturers (Daim-
lerChrysler, BMW, Renault, PSA Peugeot Citroen,
Ford, CRF, Volvo Technical Development), four
automotive suppliers (Siemens VDO, Delphi,
SAGEM and Bosch) and one research institute
(INRETS). One goal of PReVENT is also to con-
tribute to the European Commission’s 2001 goal
of halving the number of fatalities on Europe’s
roads by 2010 as specified in eSafety for Road
and Air Transport. IP PReVENT Final Report
can be found in PReVENT web site.
EASIS
The EASIS (Electronic Architecture and Sys-
tem Engineering for Integrated Safety Systems)
Introduction
(2007), which was part of the European Com-
mission’s 6th Framework Programme launched
in 2004, is a partnership of 22 European vehicle
manufacturers, automotive suppliers, tool sup-
pliers and research institutes with the aim to
develop technologies for the realization of future
ISS (Integrated Safety Systems).
SEVEOM
SEVECOM (Secure Vehicular Communica-
tion) (2008), an EU-funded project launched in
2006, focuses on providing a full definition and
implementation of security requirements for ve-
hicular communications. A liaison with security
activities in EASIS supported the activities of
SEVECOM. Its members include TRIALOG,
Bosch, Budapest University of Technology &
Economics, Daimler, EPFL, CRF-Fiat Research
Center, Katholieke Universiteit Leuven, and Ulm
University. The SEVECOM project will end by
January 1, 2009.
Vehicle Safety Communications
Consortium
The Vehicle Safety Communications (VSC) Proj-
ect was a 2.5 year program started in May 2002.
Vehicle Safety Communications Consortium
(2008) members, including BMW, DaimlerChrys-
ler, Ford, GM, Nissan, Toyota and Volkswagen,
participated with the U.S. Department of Trans-
portation in this cooperative program. The objec-
tive was to identify vehicle safety applications
enhanced or enabled by external communications,
determine their respective communication re-
quirements, evaluate the emerging 5.9 GHz DSRC
vehicle communications technology and align the
proposed DSRC communications protocols to
meet the needs of vehicle safety applications.
US DOT
The U.S. Department of Transportation’s (US-
DOT, 2008) ITS program focuses on intelligent
vehicles, intelligent infrastructure and the creation
of an intelligent transportation system through
the integration of these two components. The
Federal ITS Program Initiatives in 2004 included
Vehicle Infrastructure Integration, Cooperative
Intersection Collision Avoidance Systems, and In-
tegrated Vehicle Based Safety Systems. Through
the Integrated Vehicle-Based Safety Systems
initiative, the USDOT is seeking to establish a
partnership with the automotive and commercial
vehicle industries to accelerate the introduction
of integrated vehicle-based safety systems into
the Nation’s vehicle fleet.
VII Consortium (VIIC)
VII (Vehicle and Infrastructure Integration)
Consortium (VIIC) (VIIC and VII Program Over-
view, 2005; Robinson, 2006) was incorporated
in November 2004, and cooperative agreement
was signed in December 2005. Current member
participation includes Ford, DCX, Nissan, Honda,
VW and BMW. The objective is to create an
enabling communication infrastructure to save
lives using intelligent warning systems, improve
mobility and congestion, enhance driving expe-
rience with new services and enhance roadway
maintenance and planning. VIIC provides single
voice to USDOT and joint pre-competitive tech-
nology development environment.
California PATH
California Partners for Advanced Transit and
Highways (California PATH, 2008) was founded
in 1986 and is administered by the Institute of
Transportation Studies, and University of Cali-
fornia, Berkeley, in collaboration with Caltrans
(California Department of Transportation). PATH
is a multi-disciplinary and cooperative program
for staffs, faculties and students from universities
statewide to engage in cooperative projects with
private industry, state and local agencies, and
non-profit institutions. Its mission is to develop


solutions to the problems of California’s surface
transportation systems. Current PATH research
projects are divided into four program areas:
Policy and Behavioral Research, Transportation
Safety Research, Traffic Operations Research,
and Transit Operations Research.
CALM Continuous Communications
for Vehicles
CALM (2008) is under ISO TC 204 Working
Group 16 Wide Area Communications-Protocols
and Interfaces. The scope of CALM is to provide
a standardized set of air interface protocols and
parameters for medium and long range, high
speed ITS communication including V2V and
V2I communications.
AUTOSAR
AUTomotive Open System ARchitecture (AU-
TOSAR, 2008) is a partnership of automotive
manufacturers and suppliers working together to
develop and establish an open and standard auto-
motive software architectures. Its core partners
are BMW, Bosch, Continental, Daimler, Ford,
OPEL, GM, PSA Peugeot Citroën, TOYOTA, and
Volkswagen. The AUTOSAR project plan was
released in May 2003 and the first AUTOSAR
Open Conference held in October 2008.
JasPar
JasPar (2008) is formed by the Japanese automo-
tive companies and its board members include
TOYOTA MOTOR CORPORATION, Nissan
Motor Co., Ltd, Toyota Tsusho Electronics, Honda
R&D Co., Ltd and DENSO CORPORATION. The
aim of JasPar is to reduce technology develop-
ment costs and promote technology development
by encouraging Japanese companies to collab-
oratively develop pre-competitive technologies
such as automotive LAN enabling technology,

Introduction
middleware and software platform, and contribute
to development of global standards.
Internet ITS Consortium (Japan)
Internet ITS Consortium (2008) is developing
a common Internet ITS platform and aims for
promotion of a standardized global Internet ITS
specification. It is formed by Japanese companies
and has 11 companies as Executive Members, 12
companies as Regular Members and 69 companies
as Supporting Members.
A*CAR in Singapore
With the rapid growth in the global automotive
population, the A*STAR (Agency for Science,
Technology and Research) Capabilities for
Automotive Research (A*CAR) began as a task-
force with the aim of establishing an initiative
to address technical challenges in the automo-
tive area and provide technical leadership to
the automotive supplier industry in Singapore
through R&D (Yong, 2008). A*CAR taskforce
has launched an automotive consortium to bring
together automotive OEMs, suppliers and R&D
community to work hand in hand in addressing
key research areas in automotive technology.
The consortium is driven by 7 A*STAR research
institutes namely, Data Storage Institute (DSI),
Institute for Infocomm Research (I²R), Institute
of Chemical and Engineering Sciences (ICES),
Institute of High Performance Computing (IHPC),
Institute of Materials Research and Engineering
(IMRE), Institute of Microelectronics (IME), and
Singapore Institute of Manufacturing Technology
(SIMTech) (A*CAR, 2008).
FUTURE TRENDS
The global automotive industry is the world’s
largest manufacturing industry and most industry
Introduction
analysts predict that there will be a year-on-year
growth till 2012 (PwC Automotive Institute,
2008). The overall market for automotive network-
ing will grow at a 10% annual rate through 2011
and the components for the automobile networking
market will nearly triple between 2005 and 2011.
Driving this trend is the introduction of more
electronics in every vehicle, and that accounts
for the much faster annual growth rate than the
auto industry’s 2 to 3% (Regt, 2007).
In the future, more and more control modules
will be attached into the high-speed data network-
ing backbones embedded in a vehicle. Vehicular
networks will simplify the operation of control
systems and allow more features to be deployed
within automobile. For example, information and
direction displays will be embedded directly into
windshields. Another example is that instead of
self-parking being a service that is offered only
on luxury cars, it will soon become a standard
feature on every car. With the advances in sen-
sors, control, GPS (Global Position System) and
networking technologies, “self-driving cars” will
become possible in the future (Uldrich, 2008).
Neurotechnology that studies drivers’ brain
patterns and brain computers that read drivers’
intentions from their brainwaves through electro-
encephalogram (EEG) will help keep them alert
and monitor sleepy driver syndrome.
Vehicular networks still plays a vital role in
enhancing the automotive industry for safety,
security and entertainment. In the future, as
more and more information is streamed onto
the Internet, vehicular networks will allow driv-
ers and passengers to enjoy their journey more
than ever before with entertainment and area
information.
CONcLUSION
This introduction chapter presents the emerging
area of vehicular networks in the forms of Intra-
Vehicle (InV), Vehicle-to-Vehicle (V2V), and Ve-
hicle-to-Infrastructure (V2I) communications. It
briefly surveys the car communications, potential
applications, potential wireless technologies, and
specially designed technologies DSRC standards
and communication stack for data exchange. As
the emerging area of vehicular networks has at-
tracted a number of R&D groups in the world,
this chapter then introduces the consortiums and
initiatives working on advanced automotive tech-
nologies in Europe, the U.S., Japan and Singapore.
In the future, vehicular networks certainly play
a vital role in enhancing the automotive industry
for safety, security and entertainment.
RE
Armstrong, L. (2008). What is DSRC? Retrieved
September 2, 2008, from http://www.leearm-
strong.com/DSRC/DSRCHomeset.htm
A*CAR. (2008). Automotive @ A*STAR. Bro-
chure from Agency for Science, Technology and
Research (A*STAR).
AUTOSAR. (2008). Retrieved September 2, 2008,
from http://www.autosar.org
California PATH. Retrieved September 2, 2008,
from http://www.path.berkeley.edu
CALM. (2008). CALM Continuous Communica-
tions for Vehicles. Retrieved September 2, 2008,
from http://www.calm.hu/
Car 2 Car Communication Consortium. (2008).
Retrieved September 2, 2008, from http://
www.car-to-car.org
CiA. (2008). CAN-based in-vehicle networks.
Retrieved September 2, 2008, from http://www.
can-cia.de/index.php?id=228
Controller Area Network. (2008). Retrieved Sep-
tember 2, 2008, from http://www.can-cia.org/
Dammeyer, J. (2008). Wireless Controller Area
Network. Retrieved September 2, 2008, from


http://www.autoartisans.com/documents/canrf_
prod_announcement.pdf
Dedicated Short Range Communications. (2008).
Retrieved September 2, 2008, from http://grouper.
ieee.org/groups/scc32/dsrc/
Dridi, S., Gouissem, B., Hasnaoui, S., & Rezig, H.
(2006). Coupling Latency Time to the Throughput
Performance Analysis on Wireless CAN Net-
works. In Prof. of the International Multi-Con-
ference on Computing in the Global Information
Technology (ICCGI’06).
EASIS. (2007). Retrieved September 2, 2008,
from http://www.easis-online.org
eSafety. (2008). Retrieved September 2, 2008,
from http://ec.europa.eu/information_society/ac-
tivities/esafety/index_en.htm
European Commission. (2008). European Road
Safety Day on 13 October 2008. Retrieved
September 2, 2008, from http://ec.europa.
eu/transport/roadsafety/road_safety_days/in-
dex_2008_en.htm
FlexRay. (2008). Retrieved September 2, 2008,
from http://www.flexray.com/
Ford Sync. (2008). Today’s Drivers Demand
Staying Connected. Retrieved September 2,
2008, from http://media.ford.com/article_display.
cfm?article_id=25169
Heffernan, D., & Leen, G. (2008). ICT based
research at Limerick contributes to automotive
‘drive-by-wire’ technology. Retrieved Septem-
ber 2, 2008, from http://www.irishscientist.
ie/2002/contents.asp?contentxml=02p237b.
xml&contentxsl=is02pages.xsl
i2010 Intelligent Car Initiative. (2008). Retrieved
September 2, 2008, from http://ec.europa.eu/
information_society/activities/intelligentcar/in-
dex_en.htm
10
Introduction
IEEE 802.11p. (2008). Retrieved September 2,
2008, from http://en.wikipedia.org/wiki/IEEE_
802.11p
Internet ITS Consortium. (2008). Retrieved
September 2, 2008, from http://www.internetits.
org/
JasPar. (2008). Retrieved September 2, 2008, from
https://www.jaspar.jp/english/index_e.php
Jiang, D., & Delgrossi, L. (2008). IEEE 802.11p:
Towards an International Standard for wireless
Access in Vehicular Environments. Retrieved
September 2, 2008, from http://ieeexplore.ieee.
org/stamp/stamp.jsp?arnumber=04526014
Kuban, P. A. (2007). A Controller Area Net-
work Gateway to ZigBee- A Proposition of an
Architecture to Extend CAN. VDM Verlag Dr.
Mueller e.K.
Local Interconnect Network. (2008). Retrieved
September 2, 2008, from http://www.lin-sub-
bus.org/
Matric. (2008). The Controller Area Network
(CAN) has gone wireless! Retrieved September
2, 2008, from http://www.matric.com/resources/
Canbridge_brochureV2.pdf
Michaelides, R. (2008). Wireless CAN Interface.
Retrieved September 2, 2008, from http://www.
rmcan.com/index.php?id=17&L=1
Network on Wheels (NoW). (2008). Retrieved
September 2, 2008, from http://www.network-
on-wheels.de/
PReVENT. (2008). Retrieved September 2, 2008,
from http://www.prevent-ip.org/
PwC Automotive Institute. (2008). Analyst Note,
August 1, 2008.
Regt, d. H. (2007). Back To Basics & Future
Trends: Automotive Networking. December
7. 2007. Retrieved September 8, 2008, from
Introduction
http://www.automotivedesignline.com/how-
to/204702733
Robinson, R. (2006). The VII Consortium and the
Cooperative Agreement with USDOT. Retrieved
September 2, 2008, from http://www.itsmichigan.
org/AnnualConf2006/presentations/Ralph-ITS-
M%20Meeting-%20VIIC%20Overview_a.ppt
SAFESPOT. (2008). Retrieved September 2,
2008, from http://www.safespot-eu.org/pages/
page.php
SEVECOM. (2008). Retrieved September 2, 2008,
from http://www.sevecom.org/
Uldrich, J. (2008). The Future of the Automobile.
March 10, 2008. Retrieved September 8, 2008,
from http://www.jumpthecurve.net/index.php/re-
cent_posts/the_future_of_the_automobile/
USDOT. (2008). Retrieved September 2, 2008,
from http://www.its.dot.gov/its_overview.htm
Vasilash, G. S. (2005). Hardware-in-the-loop: As-
suring performance & quality by combining the
real & the virtual testing. Automotive Design &
Production. Gardner Publications, Inc. Retrieved
October 14, 2008, from http://findarticles.com/p/
articles/mi_m0KJI/is_/ai_n13784131
Vehicle Safety Communications Consortium.
(2008). Retrieved September 2, 2008, from http://
www-nrd.nhtsa.dot.gov/pdf/nrd-12/CAMP3/
pages/VSCC.htm
VIIC and VII Program Overview. (2005). Retrieved
September 2, 2008, from http://www.leearmstrong.
com/DSRC%20Home/Standards%20Programs/
North%20American/Previous%20Meetings/
March%2006/VIIC%20Overview.ppt#1
WAVE. (2008). IEEE 1609—Family of Standards
for Wireless Access in Vehicular Environments.
Retrieved September 2, 2008, from http://www.
standards.its.dot.gov/fact_sheet.asp?f=80
Yong, M. S. (2008). A*CAR. A*STAR Program
Proposal.
11
12

Chapter II
Drive by Wire Systems:
Impact on Vehicle Safety
and Performance

Sohel Anwar
Indiana University-Purdue University Indianapolis, USA

ABSRACT

An overview of the drive by wire technology is presented along with in-depth coverage of salient drive by
systems such as throttle-by-wire, brake-by-wire, and steer-by-wire systems, and hybrid-electric propul-
sion. A review of drive by wire system benefits in performance enhancements and vehicle active safety is
then discussed. This is followed by in-depth coverage of technological challenges that must be overcome
before drive-by-wire systems can be production ready. Current state of the art of possible solutions to
these technological hurdles is then discussed. Future trends in the drive-by-wire systems and economic
and commercialization aspects of these system are presented at the conclusion of the chapter.

INTRODUCTION are no mechanical or hydraulic connections be-

Drive by wire (DBW) systems are relatively new
technology that are increasingly finding their
place in modern automobiles. A drive by wire
system is an automotive system that interprets
driver’s inputs and executes the commands to
produce desired vehicle behavior, typically via a
microprocessor-based control system. A typical
drive-by-wire system comprises of redundant
sensors, actuators, microprocessors, and com-
munication channels for fault tolerance. There
tween driver’s input interface (e.g. throttle, brake,
steering) and vehicle system (e.g. engine/traction
motor, brake/steering actuators) in a drive by wire
equipped vehicle.
A broadened definition of drive by wire sys-
tems will include other microprocessor based
automotive control systems such as anti-lock
braking system (ABS), traction control system
(TCS), yaw stability control (YSC), etc. These
systems are designed to enhance the safety of
the vehicle by continuously monitoring various

Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.
Drive by Wire Systems
vehicle states and taking corrective action and /
or warning the driver upon detection of an im-
pending unsafe vehicle condition. The first such
system came into commercialization is ABS in
early 1970’s. It was followed by traction control
system and electronic stability control in the 1980’s
and 1990’s (Margolin, 1997; Stanton & Marsden,
1997; Wagstaff, 1999; Davis, 2001; Higgins &
Koucky, 2002; Anon, 2003; Fowler, 2003; Ross,
2003; Lee, 2003; Daniels, 2005; Kendall, 2005).
The first true drive by wire system to come to
the market was Throttle By Wire (TBW) which
was incorporated in high end vehicles such as Audi
A6, Mercedes Benz, Lexus, and BMW models
in the late 1990’s and early 2000’s. The TBW
systems were advantageous in stability control
applications where the throttle deactivation may
be needed in order to improve the traction so that
sufficient brake torque can be generated.
Electro-hydraulic brake (EHB) system, a form
of brake by wire (BBW), was first introduced in
Mercedes Benz SL series in 2001-02 (Higgins &
Koucky, 2002). Although hydraulically actuated,
these brakes operate on commands from sensors
at the brake pedal and generate the necessary
brake pressure at the wheel cylinders via a set
of electronically controlled valves and a pump.
However, the brake by wire system was decom-
missioned and removed from the vehicle due to a
number of field problems a few years later. Work
on the electro-mechanical brakes (EMB), another
form of brake by wire system that does not use
hydraulic fluid, was done in the late 1990’s by
number of automotive companies such as Bosch,
Continental, and TRW. However, issues related
to their reliability and fault tolerance still remain
which must be addressed before these system can
be used in an automobile.
Steer by wire (SBW) system is by far the most
complex drive by wire system which is also the
most safety critical by-wire system in an automo-
bile. In a pure steer by wire system, the steering
column is eliminated. Sensors mounted on the
steering wheel are interpreted by the controller
to generate the correct amount of road wheel
angle using electric motors based on the vehicle
velocity. If a sensor stops functioning properly, the
controller will not be able to actuate the motors to
generate the correct road wheel angle, potentially
causing hazardous situation.
Figure 1 shows a brief chronology of the drive
by wire system introduction into the modern
automobile with the broadened definition (Iser-
mann et al, 2002). As shown, the steer by wire
system will likely be the last of the drive by wire
system to be introduced in the automobile due to
its complexity and safety criticality.
In an even broader definition, hybrid electric
vehicles, electric vehicles, and plug-in hybrid
electric vehicles can also be classified as drive by
wire equipped automobiles due to the electronic
control of various subsystems in these vehicles.
Electric vehicles (EV) by their very nature are
drive by wire that is propelled by electronic control
of the electric traction motor based on the sensor
information from the throttle pedal. However, the
steering and brakes of an EV may still be hydro-
mechanically operated. In case of hybrid electric
vehicle (HEV), a sophisticated microprocessor
based control system channels the power flow
between the internal combustion (IC) engine,
the battery, the electric motor / generator, and
the vehicle wheels (Lu & Hedrick, 2005). All of
these functions are done via a central controller
for optimal performance. Plug-in hybrid electric
vehicles are very similar to hybrid electric vehicle,
except that a more powerful battery extends the
vehicle range in pure electric mode.
This chapter is organized as follows: A more
detailed coverage on drive by wire system is
covered in the next section. The performance and
safety benefits of the drive by wire systems are il-
lustrated in the following section. This is followed
by the section on technological challenges and
possible solutions associated with DBW system.
Future trends for the DBW system is presented in
the next section. Lastly, some final thought will
be presented in the conclusion section.
13

Figure 1. Hazard severity of failures in advanced automotive control and drive by wire systems
DRIVE WIRE SYSTEMS:
CURRET STATE OF THE ART
Figure 2 shows a pictorial view of a number of
drive by wire systems in a concept automobile. In
addition to standard drive by wire systems, this
concept vehicle also includes a 42V converter
which is used to power the by-wire systems. With
all the drive by wire systems in the vehicle which
use electrical power for actuation, the electric
power demand must be met using higher voltage
systems. 42V power source is thought to be a
compromise between high voltage requirement
and safety. However, the power demand for hy-
brid electric vehicles is significantly higher due
to propulsion need and hence uses a 240-300V
DC power bus. An extra layer of safety in the
design of such systems must be incorporated
to eliminate the possibility of electrocution. In
addition, for fault tolerant architecture, at least
two or more such power sources are required for
a DBW equipped vehicle.
14
Drive by Wire Systems
Figure 3 illustrates the first commercialized
brake by wire system by Daimler Benz (Higgins
& Koucky, 2002). This is an electro-hydraulic
brake (EHB) system with mechanical backup that
was installed in SL 500 model. The brake pedal
displacement sensor output is used to determine
the desired wheel cylinder pressure which is
generated via a closed loop control system that
includes a set of electro-hydraulic valves and an
electric motor driven pump. This system was
later recalled due to reliability issues. Since then
no automakers have incorporated brake by wire
systems in any of their vehicles.
Figure 4 shows another concept vehicle that
incorporates drive by wire systems on a hybrid
electric vehicle. This concept is based on the
synergy of combining a DBW system with HEV.
The DBW systems can easily be powered by the
HEV battery pack or the high voltage power bus.
The IC engine along with the motor generator
will ensure that power is always available for the
DBW systems.
Drive by Wire Systems

Figure 2. Concept drive by wire equipped vehicle

Figure 3. Brake by wire equipped Mercedes Benz model SL 500 (source: Daimler Benz)

15
 Drive by Wire Systems

Figure 4. A drive by wire equipped hybrid electric concept vehicle

Figure 5. General Motors’ HyWire concept vehicle (source: GM)

16
Drive by Wire Systems
Another concept vehicle that incorporates drive
by wire system in a Fuel Cell HEV (FCHEV) is
illustrated in Figure 5. The concept vehicle was
designed by General Motors and was codenamed
“HyWire” (“Hy” for hydrogen fuel cell and “Wire”
for drive by wire). GM later renamed this vehicle
“FX-3” and built a prototype of this vehicle in
2006. This vehicle has fuel cell propulsion system
with four wheel electric traction motors. Vehicle
steering, braking, and suspension are all controlled
electronically (drive by wire).
Drive-By-Wire systems offer a number of
benefits when incorporated on a vehicle. Some
of the benefits are as follows:
• DBW systems can easily be configured
(via software updates) for added or tunable
features such as brake pedal feel / enhanced
safety via stability control.
• DBW systems can enhance vehicle perfor-
mance by prepositioning brake calipers for
fast brake actuation or allowing for over-
steering to enhance maneuverability.
• DBW systems can improve fuel economy via
better engine / motor / powertrain control
and via regenerative braking.
• DBW systems can offer better ergonomics
such as adjustable feel at driver’s interface
(steering wheel, brake / accelerator ped-
als).
• DBW systems offer better fault detection
and warning to the driver which enhances
the safety, reliability, and maintenance of
the vehicle.
• DBW systems are able to incorporate multi-
functionality in a single system thereby mak-
ing today’s advanced features (e.g. stability
control systems via active steering) more
cost effective in these automobiles.
• SBW systems can not only provide these
additional features, but also can free up
premium packaging space by eliminating
the steering column thereby enabling easy
assembly of the instrument panel.
• SBW systems also offer variable steering
ratio at different vehicle speeds for additional
safety.
• Automatic line keeping is possible for SBW
equipped vehicle on the automated highways
of the future. This feature will further en-
hance safety and comfort of the driver.
• Haptic steering for improved ergonomics and
operator safety & performance is possible
with SBW vehicles.
• DBW systems will offer reduced overall
production cost for steering systems via
standardized modules and software.
• BBW systems offers improved brake noise,
vibration, and harshness (NVH) since there
is no direct mechanical or hydraulic link
from the pedal to the wheels. ABS pulsation
in normal hydraulic brake will disappear in
a BBW system.
• BBW systems also allows for Better Comfort
Level
○ Anti-Dive Algorithm
• Better fuel economy through regenerative
braking
• Easier maintenance
○ Less wear and tear
○ No need to turn those rotors
• Extended brake life
○ No need to replace brake shoes, pads,
or rotors
• Better cold performance than wet brake
systems
• Better braking-in-a-turn performance
• Enhanced safety
○ Stability control algorithms
○ Reduction of brake fade
• Enhanced cruise control on a down hill
However, there are a number of challenges that
a DBW system must address before full commer-
cialization. These challenges are discussed in the
“Technological Challenges” section.
In this section, the following drive by wire
systems are presented in more detail:
17

1. Throttle By Wire Systems
2. Brake By Wire Systems
3. Steer By wire Systems
Throttle B y Wire (TBW) Systems
For traction control system based on engine inter-
vention, it is necessary to control engine torque.
By electronically controlling the mass air flow rate
via a throttle by wire (TBW) system, it is possible
to achieve this functionality. For yaw stability
control system, it may be necessary to control the
engine torque to the wheels to maximize available
traction. TBW systems allow this functionality
through their very design. For hybrid electric
vehicle propulsion, it is necessary to mechanically
disconnect the pedal from the throttle to allow for
torque blending between the engine and electric
motor which can easily be achieved via a TBW
system. A TBW system also provides an easier
Figure 6a. Schematic of a throttle by wire (TBW) system
Figure 6b. Schematic of an electronic throttle body in a TBW system
18
Drive by Wire Systems
way to implement engine braking for commercial
vehicle (Jake brake, Exhaust gas brake, etc.). For
autonomous vehicles, it is necessary to have a
TBW system for vehicle propulsion.
Figure 6a illustrates the basic operation of a
TBW system. The pedal sensor information along
with throttle angle feedback is used to actuate the
TBW motor to maintain the commanded throttle
angle. Figure 6b shows an electronic throttle body
(ETB) used in a TBW system.
Figure 7 illustrates a possible interface between
ABS / TCS / YSC systems and a TBW system. As
indicated earlier, a TBW system is a natural fit in
an active safety system or a combination of active
systems (as in this case) where upon the ABS /
TCS / YSC microprocessors share information
with a TBW microprocessor and a supervisory
microprocessor. The TBW system microprocessor
gives preference to the active safety throttle com-
mand over its own pedal command and controls
the throttle angle accordingly.
Drive by Wire Systems
Figure 7. TBW system interface with ABS / TCS / YSC active safety systems
Brake B y Wire (BBW) Systems
Brake by wire systems can be classified in three
main categories, based on the type of actuation
system, as stated below:
1. Electro-hydraulic (EH) BBW systems
2. Electromechanical (EM) BBW systems
3. Electromagnetic (EMg) BBW systems (Re-
generative braking via electric generator or
Eddy Current braking)
The electro-hydraulic brake by wire system is
composed of an electronically controlled pump/
motor assembly, a set electronically controlled
valves, pressure transducers, brake pedal sensors,
driver modules, and microcontrollers. Figure 8a
illustrates the actuation system of an EH brake by
wire system. The brake pedal sensor information
is sent to microcontroller which interprets this
in terms of a desired wheel cylinder pressure
command. The microcontroller also receives
the wheel cylinder pressure information from
the pressure transducer. Based on the desired
and actual wheel cylinder pressure values, the
microcontroller calculates appropriate pump and
valve commands (via a control algorithm) which
are then used to either build or dump the pressure
in order to minimize the pressure error.
The electromechanical (EM) brake by wire
system is a dry system and no hydraulic fluid
is used. It still uses the friction pads or shoes
for braking the vehicle as in EH BBW system.
However, the actuation mechanism is based on
an electric motor and a mechanism that applies
force onto the friction pads. Figure 8b illustrates
the operation of this type of brake system (Hose-
innezhad, 2006). In addition to the brake pedal
sensor, a number of other sensory information
becomes necessary for such a system (e.g. clamp
force, actuator position).
The third type of brake by wire system is based
on electromagnetic retardation. The principles
of electromagnetic retardation are utilized to
generate the braking forces which are contact-
less. As a result, no friction pad is needed in
electromagnetic braking. Two principal types of
actuators are used: (1) Eddy current brakes, and
(2) Regenerative braking via generator.
19
 Drive by Wire Systems

Figure 8a. Schematic of an electro-hydraulic brake by wire actuation system

Figure 8b. Electromechanical brake by wire actuation system

20
Drive by Wire Systems
Figure 8c. Electromagnetic brake by wire actuation system
Figure 8c shows a picture of an eddy current
braking actuator which is composed of a wire
wound stator and solid iron rotor. When the stator
is energized, an electromagnetic field is generated
which gives rise to Lorentz forces according to
Maxwell’s equation when the rotor turns within
the electromagnetic field. These forces oppose the
motion of the rotor, thereby producing retarding
torque. Regenerative braking works in a similar
fashion. One limitation of electromagnetic brak-
ing systems is their dependency on the rotational
speed, that is, the retarding torque is a function
of the rotor speed. As a result, at lower speed, the
retarding torque diminishes significantly and will
not be able to provide sufficient braking torque.
For this reason, electromagnetic brake by wire
systems are rarely used as a standalone brake
system. They are often used in conjunction with
a friction brake system.
Steer B y Wire (SBW) Systems
Steer by wire systems are in the early stage of
development. Major automakers are yet to decide
if and when such system will be incorporated in
a vehicle. Similar to the brake by wire systems,
SBW systems use microcontrollers to actuate
the road wheel in the lateral direction based
on the hand wheel steering angle and vehicle
speed information. Figure 9 illustrates a possible
configuration of a steer by wire system. In this
particular configuration, the hand wheel steering
angle is measured via an appropriate sensor which
is then communicated to two microcontrollers.
These microcontrollers also receive vehicle speed
information from the vehicle communication bus.
Based on the steering wheel angle and vehicle
speed information, the microcontrollers calculated
the road wheel angle command via suitable algo-
rithm. The microcontrollers then tries to maintain
the commanded road wheel angle by actuating
the rack mounted motors in a closed loop fashion
via road wheel angle feedback.
Steer-By-Wire systems are a relatively new
development compared to the traditional me-
chanical, hydraulic or electric steering systems
that are currently used for motor vehicles. In the
SBW system, there is no mechanical coupling
between the steering wheel and the steering
mechanism, i.e. the vehicle’s steering wheel is
detached from the steering mechanism during
normal operation. Even though the mechanical
21

Figure 9. Schematic of an SBW system
linkage between the steering wheel and the road
wheels has been eliminated, a SBW steering sys-
tem is expected not only to implement the same
functions as a conventional mechanically linked
steering system, but also to provide the advanced
steering functions.
There are a number of important steering
functional requirements for a Steer-By-Wire
system:
• Directional control and wheel synchroniza-
tion.
• Adjustable variable steering feel.
• Adjustable steering wheel returns capabil-
ity.
• Variable end of travel stop for steering
wheel.
• Variable steering ratio.
It can be noticed that all of the drive by wire
systems described above comprised of multiple
redundancy in the sensors, actuator, microcon-
trollers, and communications buses. This is ob-
viously by design which ensures a fault tolerant
operation. Of the three drive by wire systems
presented above, both BBW and SBW systems
are considered safety critical systems which
22
Drive by Wire Systems
means that any failure of the system may lead
to catastrophic consequences. As a result, these
systems must be designed to withstand multiple
faults or failures within the system, before reach-
ing conditions of catastrophic proportions. A
fault in sensor, actuator, or microcontroller must
not cause any severe consequences. This fault
tolerance feature requires multiple redundancy
in the hardware architecture which provides fault
tolerant operation.
PERFORMANCE ENHANCEMENTS
AND ACTIVE SAFETY VIA DRIVE
BY WIRE SYSTEMS
Drive by wire systems can significantly enhance
the design and performance of the vehicle active
safety systems such as the collision avoidance
systems, adaptive cruise control, active front
steering, stability control, road condition warn-
ing. Figure 10a-b illustrates trend of incorporation
of various active safety systems in automobiles
and driver error mitigation via DBW systems.
The more advanced systems (e.g. platooning,
highway copilot, autonomous driving) can only
Drive by Wire Systems
Figure 10a. Trend of active safety systems in automobiles
Max Underfloor
Concept
Side Airbag
Side Impact Protection
Airbag
Safety Potential
Deformation
Elements
Compound Environment recognition
Glass Road recognition (LDW)
Seat
Safety Belt Autonomous Cruise Control
Cell
ASRESP
ETC
ABS
Low
960 90 980 990 000
Figure 10b. Driver errors and possible solution via DBW systems
Error
Get into wrong lane
Forget which gear
Only half an eye on the road
Distracted, need to brake hard
Plan route badly
Fail to recollect recent road
Wrong exit from roundabout
Intended lights, switched wipers
Forget light on main beam
Usual route taken by mistake
Misjudge speed of oncoming vehicle
Queing, nearly hit car in front
Driving too fast on dipped lights
Turn left into car's path
Miss motorway exit
Manouver without checking mirror
Fail to see pedestrian crossing
Brake too quickly
Hit something when reversing
Overtake without using mirror
Misjudge gap in car park
Turning left, nearly hit road user
Misjudge interval turning right
Try to pass vehicle turning right
Fail to see pedestrian stepping out
Attempt to drive off in third gear
Try to drive off without starting car
Passive Safety
(reducing personal injury
Autonomous in case of an accident)
Driving
Highway Copilot
Platooning Active Safety
Emergency Brake (avoiding an accident)
Modified Source : Auto-Zeitung Nr.  Jan.9
.....
Automated solution
Navigation system
Automatic gear shift
Fully automated driving system
Anti-lock braking system
Route planning and navigation system
Navigation system
Navigation system
Daylight sensor and automatic lights
Automatic lighting
Navigation system
Collision avoidance system
Collision avoidance system
Vision enhancement system
Collision avoidance system
Navigation system
Collision avoidance system
Collision avoidance system
Anti-lock braking system
Collision avoidance system
Collision avoidance system
Collision avoidance system
Collision avoidance system
Collision avoidance system
Collision avoidance system
Collision avoidance system
Automatic gear shift
Automatic start-up
23

Figure 11. Possible active safety features via the DBW system in an automobile
be effectively achieved via drive by wire system.
Other active safety systems are a natural fit for
drive by wire systems as these systems can eas-
ily be incorporated via the addition of a software
module in a drive by wire equipped vehicle. As
described in (Anwar, 2006), the ABS algorithm
was implemented via only software update to a hy-
brid brake by wire system. In addition to passenger
vehicles, SBW systems have also been investigated
in the context of large construction machineries
such as wheel loaders with superior performance
enhancements (Haggag et al, 2005).
Drive by wire systems can also significantly
enhance the performance of a vehicle by tak-
ing proactive actions in a driving scenario. For
example, vehicle handling can be made more
responsive by adding slight oversteer on high fric-
tion coefficient surface which can easily be done
via a steer by wire system (Hebden et al, 2004;
Yih & Gerdes, 2005; Chang, 2007). Performance
enhancement via SBW system while maintaining
stability has also been investigated by a number
of researchers (Limpibunterng & Fujioka, 2002;
Oh et al, 2004; Segawa et al, 2004; Limpibunterng
& Fujioka, 2004). Similarly vehicle suspension
can be tuned dynamically via active suspension
control system to adapt to road conditions the
vehicle is in. Figure 11 illustrates possible sce-
narios where drive by wire systems are utilized
to improve overall vehicle safety via actively
monitoring vehicle states and taking corrective
actions or warning the driver.
24
Drive by Wire Systems
Both brake-based Yaw Stability Control (YSC)
and Anti-lock Brake System (ABS) have been
demonstrated on BBW systems (Anwar, 2005;
Anwar, 2006). Using sliding mode control (SMC)
theory, the author provided experimental results
for the hybrid BBW system. Here it was empha-
sized that the ABS system was implemented only
through the software and no hardware modifica-
tion was needed.
YSC algorithm via active front wheel steering
has also been developed and tested in a SBW
equipped vehicle (Yih & Gerdes, 2005; Zheng &
Anwar, 2008). The test results from these studies
indicate that the steering based YSC system on a
SBW equipped vehicles has significant safety ad-
vantages in critical situations over a conventional
vehicle where the driver has to manually control
the vehicle against an unexpected yaw motion.
TECHNOLOGICAL CHALLENGES
FOR DRIVE BY WIRE SYSTEMS
AND POSSIBLE SOLUTIONS
The explosion of the usage of the electronic control
systems is nowhere more apparent than in the
modern automobiles. During the last two decades,
advances in electronics have revolutionized many
aspects of automotive engineering, especially in
the areas of engine combustion management and
vehicle safety systems such as anti-lock brakes
Drive by Wire Systems
(ABS) and yaw stability control (YSC). The ben-
efits of applying electronic technology are clear:
improved performance, safety and reliability with
reduced manufacturing and operational costs.
However, only recently has the electronic revolu-
tion begun to find its way into automotive chassis
systems in the form of electronically controlled
variable assist and, within the past few years,
fully electric power assist.
The complexity of the DBW systems arises
from the fact that these systems must incorporate
multiple redundant sensors, actuators, control-
lers, and communications networks to achieve
fault tolerance. The fault tolerant control of
these systems is accomplished via appropriate
fault detection, isolation, and accommodation
(FDIA) algorithms. However, the total number
of redundant components makes the DBW sys-
tems prohibitively expensive. Also, in order to
accurately detect and isolate a component failure
without raising any false alarms, fast and robust
detection algorithms are needed.
A number of issues must be resolved before
drive by wire system equipped vehicles can be
brought to the market. Cost and reliability are
two major challenges facing these systems. While
reliability of these systems can be improved via
increased redundancy and fault tolerant control
algorithms and communication protocols, the
physical redundancy of components makes these
systems prohibitively expensive. In this section,
we highlight the major challenges facing the drive
by wire systems including cost, reliability, packag-
ing, control, and communication protocols.
The numerous benefits of DBW systems out-
weigh the risks in introducing such systems in
the future automobiles. The concept vehicles with
DBW systems have demonstrated such benefits
(Stanton & Marsden, 1997). However, the overall
cost of highly reliable DBW system is still sev-
eral times higher than the conventional systems;
primarily due to the presence of multitude of
redundant components (sensors, microcontrollers,
actuators etc.). Model-based fault detection tech-
niques can be one possible solution, which could
lower the overall cost without compromising
reliability.
Fault Tolerant Control
Fault tolerance is a property of a system that con-
tinues operating properly in the event of failure
of some of its parts. It provides the ability of a
system to provide a service complying with the
specification in spite of faults. If operating qual-
ity decreases at all, the decrease is proportional
to the severity of the failure, as compared to a
naively-designed system in which even a small
failure can cause total breakdown.
In order to bring down the cost, the total number
of redundant components must be reduced without
compromising the fault tolerance. One possible
solution to this problem is to utilize analytical re-
dundancy or model based fault detection, isolation,
and accommodation. Model-based Fault Detection
and Isolation (FDI) explicitly use a mathemati-
cal model of the system. It is motivated by the
conviction that utilizing deeper knowledge of the
system results in more reliable diagnostic deci-
sions. The main idea is “analytical redundancy”
which makes comparison of measurement data
with known mathematical model of the physical
process. It is superior to “hardware redundancy”
generated by installing multiple sensors for the
same measured variable. They offer simplicity,
flexibility in the structure, less hardware, less
weight, and cost.
Model-based Fault Detection and Isolation is
achieved by implementing a more complex failure
detection algorithm that takes careful account of
system dynamics, which may be able to reduce
requirements for costly hardware redundancy.
Analytical redundancy based FDI uses a model
of the dynamic system to generate the redundancy
required for failure detection. In many systems,
all of the states cannot be measured because of
cost, weight and size considerations, therefore,
FDI schemes for such systems must extract the
25

redundant information from dissimilar sensors,
using the differential equations that relate their
outputs. In addition to taking hardware issues
into consideration, the designer should consider
the issue of computational complexity. Most
model-based FDI methods rely on analytical
redundancy. In contrast to physical redundancy,
when measurements from different sensors
are compared, now sensory measurements are
compared to analytically obtained values of the
respective variable and the resulting differences
are called residuals. The deviation of residuals
from the ideal value of zero is the combined result
of noise, modeling errors and faults. A logical
pattern is generated showing which residuals can
be considered normal and which ones indicate
a fault. Such a pattern is called the signature of
the fault. The final step of the procedure is the
analysis of the logical patterns obtained from the
residuals, with the aim of isolating the failures that
cause them. Such analysis may be performed by
comparison to a set of patterns known to belong
to sample failures or by the use of some more
complex logical procedure.
One major difference between pure DBW sys-
tems and other electronically controlled vehicle
systems is that even if these control systems fail,
the basic functionality of the brake and steering
system remains intact. In case of DBW systems,
a component failure can result in loss of function-
Figure 12. Redundancy architecture of a drive by wire system
26
Drive by Wire Systems
ality unless the DBW system has redundancy in
its components by design. A DBW system with
mechanical backup will allow a “limp home”
mode in case of a component or subsystem failure.
However, the mechanical backup is tantamount to
having the basic mechanical system (e.g. steering
or braking) and hence cannot be justified from a
cost standpoint. Following the example of fly by
wire systems, it is only natural to focus on drive by
wire system without mechanical backup. However,
DBW systems are microprocessor-based control
system based on sensor inputs and electronically
controlled actuation systems (via communica-
tion bus) which call for fault tolerance in case of
component failure. Figure 12 shows an example
of the redundancy architecture of a fault tolerant
drive by wire system.
Safety integrity is defined by the probability of
a safety-related system satisfactorily performing
the required safety functions under all the stated
conditions within a stated period of time. Safety
and reliability of a DBW system can generally be
achieved by a combination of the following: fault
avoidance, fault removal, fault tolerance, fault
detection and diagnosis, automatic supervision
and protection.
Fault avoidance and removal has to be accom-
plished mainly during the design and testing phase.
Analytical methods to evaluate the effects of faults
on system reliability and safety exist: reliability
Drive by Wire Systems
analysis, fault tree analysis (FTA), failure modes
and effects analysis (FMEA), hazard analysis
(HA), and risk classification.
Certain component faults / failures can occur
even after designing the hardware based on reli-
ability and safety analysis. These faults must be
tolerated, generally, by adding redundancy in the
components, units, or subsystems. Redundancies
in sensors, actuators, microprocessors, commu-
nication buses, power supplies are considered to
achieve the fault tolerance. Two basic approached
to fault tolerance: static redundancy and dynamic
redundancy.
Static redundancy uses three or more parallel
modules that have the same input signal and the
fault in any one module is detected by two-out-
of-three voting (Figure 13a). A single fault can
such be tolerated by a triple redundant module.
The maximum number of failures that can be
tolerated by an n-redundant system via static
redundancy is (n-1)/2 (n being odd).
Figure 13. Static and dynamics redundancy architectures
Dynamic redundancy requires fewer mod-
ules at the cost of more information processing
(Figure 13b). In dynamic redundancy, at least
two redundant components are required where
one component works in default mode while the
other works as a standby and takes over the main
components if it fails.
The fault in the main component is determined
by consistency checking of its output signal (such
as range of signal and its rate of change) and
comparing with that from the standby component.
For microprocessor, parity checking or watchdog
timers may be used to detect the faulty micropro-
cessor. After a fault is detected, a reconfiguration
module then switches the standby component
to take over and removes the faulty component
from operation.
For static redundancy, at least three redundant
component output signals are required for accu-
rate fault detection via majority voting algorithm.
In case of dynamic redundancy, model-based
27

Figure 14. Fault detection flow diagram
approaches can be utilized to detect fault since
output signals are readily available from various
components. Static redundancy can be found
in various mechanical and electrical systems:
spoke-wheel, multiple brushes in DC motor. Only
single point hazardous failures are considered for
mitigation upon fault detection. It has been shown
that consideration of only single point failures is
sufficient for a DBW system as opposed to a Fly
By Wire (FBW) system.
Fault detection is an integral part of a fault
tolerant control system. Fault detection methods
based on measured signals can be classified as
follows:
• Limit value checking (thresholds) and plau-
sibility checks ranges of single signals.
• Signal-model-based methods for single
periodic or stochastic signals.
• Process-model-based methods for two or
more related signals.
28
Drive by Wire Systems
In order to obtain specific symptoms, it is neces-
sary to have more than one input and one output
signal for parity equations or output observers.
For parameter estimation, it is necessary to have
at least one input signal and one output signal.
Figure 14 illustrates the fault detection process that
utilizes both signal model based fault detection
and process model based fault detection.
Comprehensive overall fault-tolerance design
can obtain fault tolerant components and fault
tolerant control. This, in turn, can be achieved
via fault tolerant lower cost components with
built-in fault tolerance. A fault tolerant sensor
configuration should be at least fail-operational
after one fault. This can be obtained via hard-
ware-redundancy with same type of sensor or
by analytical redundancy with difference sensors
and process models.
Fault tolerant actuators can be designed by us-
ing multiple actuators in parallel, with either static
or dynamic redundancy with cold or hot standby,
Drive by Wire Systems
e.g. static redundancy in hydraulic actuators in a
fly-by-wire system. Fault tolerant actuators can
also be designed with redundancy in actuator
components that have the lowest reliability, e.g.
two servo valves for hydraulic actuators or three
windings on an electric motor are examples of
static redundancy. The cabin pressure flap actuator
in an aircraft uses dynamic redundancy with cold
standby via two independent DC motors acting
on a single planetary gear.
Fault Tolerant Communication
In-car electronics plays an important role in
many automotive applications: Throttle-by-wire,
Steer-by-wire, Brake-by-wire, and many other
systems. The number of electronic components
in a car has therefore significantly increased. In
particular, in-car networks used to interconnect
electronics equipments becomes a key point. A
common practice in the past was the extensive
use of wiring. These wires significantly increased
the complexity and weight of these systems. To-
day, in-car bus networks are largely adopted for
the sake of reducing the vehicle weight and fuel
consumption. The requirements of an automotive
communication network are generated from the
applications it has to support. Major automotive
applications that involve networking and the us-
age of field buses are:
• Drive by wire
• Chassis systems
• Powertrain systems
• Infotainment or multimedia
Multimedia devices require high bandwidth
and plug-n-play capabilities. Two most important
requirements for DBW systems are dependability
and fault-containment. Distributed control is of-
ten used to meet these requirements. Distributed
control algorithms require a coordinated snapshot
of the controlled object. Such coordination can be
achieved by the provision of a global time base.
Figure 15. Common field bus technologies between
microprocessors
Unification of bus systems within vehicles:
Controller Area Network (CAN) will remain the
selection of choice for event-triggered systems
while a new protocol is needed for applications
which require high performance, determinism,
fault-tolerance, and flexibility. Several differ-
ent field buses are used to address these various
communication demands. A big issue that the
automotive system producers deal with is that
too many field bus technologies exist today. To
interconnect these systems there is a need for
high bandwidth together with flexibility and
determinism. It is desirable to move from using
too many technologies to fewer more general.
To reduce the complexity evolving in a modern
automotive system, it would be good to commit
on a set of networking protocols that can be used
in most of the applications typically.
The common field bus technologies intercon-
necting Electronic Control Units or ECUs today
are: Controller Area Network (CAN), Local In-
terconnect Network (LIN), Byteflight, and Media
Oriented Systems Transport (MOST). Figure 15
shows a graphical illustration of the requirements
for various communication protocols.
Analytical Redundancy Based Fault
Tolerant Control of DBW Systems
Physical redundancy based fault tolerant control
and communication protocols can significantly
enhance the reliability. However, this method-
29

ology increases the overall system cost expo-
nentially due to the presence of a large number
of redundant hardware components. One pos-
sible solution to this problem is to replace the
physical redundancy of the fault tolerant drive
by wire systems by analytical redundancy. Ana-
lytical redundancy methodologies can not only
reduce the overall system cost by reducing the
total number of redundant components, but also
further improve overall reliability of the system
through the usage of a diverse array of sensory
information. An up to date in-depth coverage of
analytical redundancy based fault tolerant control
and communication methodologies will be given
in this section including predictive and nonlinear
methodologies.
The concept of analytical redundancy has been
investigated in the context of aerospace applica-
tions, primarily utilizing Eigen-structure theory.
However, most of these articles were aimed at
isolated subsystems in an aircraft or a spacecraft.
Fly-By-Wire (FBW) systems are mostly based on
full hardware redundancies (Bajpai et al, 2001).
As a result, analytical redundancy methodologies
have not been utilized to a great extent in FBW sys-
tems. Nonetheless, the theoretical foundation for
analytical redundancy methodology researched
in aerospace applications can be useful in DBW
system research. Introduction of Drive-By-Wire
technology is more challenging in the automobile
market, mainly because automobile consumers
cannot afford the high cost of redundant systems
the aerospace industry can. Each extra sensor,
actuator, and Electronic Control Unit (ECU) in-
creases the overall cost and weight of the vehicle.
With the profit margin already low, this approach
will not be acceptable to the automobile industries.
By employing analytical redundancy techniques
instead of hardware redundancy, it will be possible
to bring the overall cost of such system down to
a point that will be attractive to the automak-
ers for mass production without sacrificing the
high level of safety and reliability required by
the consumers. Analytical redundancy along
30
Drive by Wire Systems
with appropriate fault detection, isolation, and
accommodation methodologies can be utilized
to make the DBW systems fault tolerant (Gadda
et al, 2007). Through analytical redundancy, the
vehicle steering angle can be estimated from the
states of other vehicle parameters without using
an extra steering position sensor.
Sn ANALnalytical
REDUNDANCY in a STEER BY
WIRir
In this research, the analytical redundancy based
fast fault detection algorithms was developed that
based on physical models, nonlinear estimator
and generalized predictive algorithms. In this
algorithm, outputs from a number of redundant
sensors as well as analytical sensor are checked
against each other for a number of times before
declaring a component to be faulty. The analyti-
cal sensor output is the combination exertion of
the full vehicle model, nonlinear Sliding Mode
Observer, and the long range prediction algorithm.
The yaw rate signal can be measured with inex-
pensive sensors. Therefore with the measured
yaw rate and the measured motor current input,
the road wheel steering angles are estimated
with the Sliding Mode Observer. Thereafter the
steering angles were predicted by a long range
predictor at variable prediction horizons with
the participation of estimated steer angle and the
motor current.
The proposed research concept is utilizes
the long range prediction based fault detection,
based on the analytical model of the SBW system
(Figure 16) which would provide added safety to
the SBW system via fast and robust fault detec-
tion and isolation of a component failure in such
a system.
Since analytical redundancy methods are
model-based, long-range prediction based
Fault Detection, Isolation, and Accommodation
(FDIA) algorithms are appropriate in such an
Drive by Wire Systems

Figure 16. Electronic architecture of an SBW application since modeling errors are inevitable
system in real-world systems. Furthermore, long-range
prediction based FDIA methods (Figure 17)
provide robustness against external disturbances
which are expected in a DBW vehicle having a
multitude of electric and electronics components.
The fundamental concept in this proposition is
that the sensor outputs are compared against the
analytical counterpart (analytical redundancy)
whose outputs are predicted several time step
ahead via generalized predictive algorithm. In the
event of a component failure, the predicted output
will deviate from the sensor outputs several time
steps ahead, thus reducing the detection latency.
Table 1 lists the notations used in this section.

Observer and Predictor Based

Modeling

An observer can be designed by combining the

steering system model and vehicle model (Hasan
& Anwar, 2008):
Table 1. Notations
x = Ax + Bim + E τ f
θ Road wheel angle T
x = β r θ θ 
β Vehicle body side slip angle
 −Cα , f − Cα , r Cα , r b − Cα , f a Cα , f 
r Vehicle yaw rate … −1 + 0 ⁄
… mV mV 2 mV ⁄
a Distance from the front tire to the vehicle’s CoG … Cα , r b − Cα , f a −Cα , f a − Cα , r b 2
2
Cα , f a ⁄
… 0 ⁄
A=… Iz I zV Iz ⁄
b Distance from the rear tire to the vehicle’s CoG
… 0 0 0 1 ⁄
m Vehicle mass … ⁄
…(t p + tm )Cα , f a (t p + tm )Cα , f −(t p + tm )Cα , f −bw ⁄
V Vehicle longitudinal velocity … Jw J wV Jw ⁄
jw 

T T
Iz Vehicle moment of inertia  km   1 
B=… 0 0 0 ⁄ ;E = … 0 0 0 − ⁄
Jw Moment of inertia of the road wheel  Jw   Jw 

bw Viscous damping coefficient for wheel bearing

C = [0 0 1 0] (4.1)
C α,f and C α,r Front and rear tire cornering coefficients
The motor current is the input to the system
km Motor torque constant
and the torque due to Coulomb friction is treated
tp Pneumatic trail
as a disturbance. The above system is fully ob-
tm Mechanical trail
servable.
τf Friction torque at the road wheel
im Motor current
Sliding Mode Observer (SMO)
Ej and Fj Uniquely defined polynomials in the Diophan-
tine equation
The motivations for using Sliding Mode Observer
are, it is model free and robust respect to bounded

31
 Drive by Wire Systems

Figure 17. Road wheel angle estimation and The behavior of the mismatch governed by
fault detection, isolation, and accommodation homogeneous Equation (4.4) is determined by
(FDIA) eigenvalues of matrix (A+LC). For observable
systems, they may be assigned arbitrarily by a
θ des θ proper choice of input matrix, L. It means that
+ any desired rate of convergence of the mismatch
− im θ1 θ2 to zero or estimate x̂(t) to state vector x(t) may be
provided. Then any full-state control algorithms
with vector x̂(t) are applicable.
θ predicted The order of the observer may be reduced due
to the fact that Rank (C) = l and the observed
θ estimated vector may be represented as:

r y = C1x1 + C2 xp
θ FDIA
uncertainty. It can work under much less con-
servative condition. The idea underlying SMO
observer design methods can be illustrated for a
linear time-invariant system (Hasan & Anwar,
2008):
x = Ax + Bu
y = Cx
y ≠ Rl, x ≠ Rn, rank(C) = l (4.2)
x = [x1 xp]T
x1 ≠ R l, xp ≠ R n–l, det(C1) ≠ 0 (4.5)
It is sufficient to design an observer only for
vector xp, then the components of vector x1 are
calculated as:
x1 = C1–1 (y – C2 xp) (4.6)
Write the system Equation (4.2) in space (y, xp)
as:

The pair (C, A) is assumed to be observable and y = A11 y + A12 x p + B1u

n is order of the system. A linear asymptotic ob- x p = A21 y + A22 x p + B2u
server is designed in the same form as the original (4.7)
system (4.2) with an additional input depending
on the mismatch between the real values and the where
estimated values of the output vector:
A A12  B  C C2 
MAM −1 = …11 , MB = …1 ⁄ , M = …1
 A21 A22 ⁄ B 0 I n −1 ⁄
xˆ = Axˆ + Bu + L( y − Cxˆ ) (4.3)  2

where x̂ is an estimate of the system state vector
and L ≠ R n×l is an input matrix. The state vector
of the observer x̂ is available since the auxiliary
dynamic system is implemented in a controller.
The motion equation with respect to mismatch
x  = x –  x̂ is of form:
The coordinate transformation M in nonsin-
gular, det(M) ≠ 0. Therefore, applying the simple
Sliding Mode Observer (SMO) in the state space
system equation,
xˆ = Axˆ + Bu + L sgn( y − Cxˆ ) (4.8)

x = ( A + LC ) x (4.4)

32
Drive by Wire Systems
where, sgn(z) = col(sgn(z1), ..., sgn(zn)
+1 if , z > 0
sgn( z ) = 
−1 if , z < 0
Under a suitable choice of the gain matrix L in
the observer, sliding occurs on the manifold y – C
x̂ = 0, and it becomes equivalent to the reduced
order observer. The discontinuous vector function
v = L sgn(y – ŷ). Now from Equation (4.7),
yˆ = A11 yˆ + A12 xˆ p + B1u + L1 sgn( y − yˆ )
xˆ p = A21 yˆ + A22 xˆ p + B2u + L2 sgn( y − yˆ ) (4.9)
The system for the error y = y – ŷ is of the
form,
y = A11 y + A12 x p + B1u + L1 sgn( y )
x p = A21 y + A22 x p + B2u + L2 sgn( y )
(4.10)
The vector function v ≠  R l is chosen such that
sliding mode is enforced in the manifold y = 0 and
the mismatch between the output vector y and its
estimate ŷ reduced to zero. A vector L2 must be
found such that the mismatch x p = xp – x̂p between
xp and its estimate x̂p decays as the desired rate.
Equivalent value of the discontinuous function:
L1 sgn(y) = A12 x p (4.11)
For simplicity, L1 is considered as 1 and equation
(4.11) becomes:
sgn(y) = A12 x p (4.12)
Now the equation on the sliding manifold appears
from equation (4.10):
x p = ( A22 − L2 A12 ) x p (4.13)
Generalized Prediction Bsed
Predictor
A class of predictive self-tuning controllers,
known as generalized predictive controller (GPC)
(Hasan and Anwar, 2008) have shown robust-
ness against unstable plants, non-minimum-
phase plants, model over parameterization, and
uncertain process dead time. These controllers
have also been observed to provide offset free
behavior for the closed loop system since they
include an integral action. These set of controllers
have been very successful in regulator or track-
ing type observer applications. In the context of
long range prediction, the prediction horizon, j
is a tunable design variable that can be set to any
value according to the desired prediction range.
The predictive nature of the GP (Generalized
Prediction) based predictor algorithm comes from
the use of the Diophantine equation. Through the
use of Diophantine equation, the output of the
plant is predicted j-step ahead of present time.
This prediction output is then used for future
fault detection and identification.
Considering the state space equation (4.2), a
transfer function can be obtained as follows:
U (s)
G (s) = = C ( sI − A) −1 B (4.14)
Y (s)
After discretization of the equation (4.14),
a more general form of the dynamic model of
the dynamic model of the vehicle model can be
written as:
A(z–1) Y(z) = B(z–1) U(z) (4.15)
where A(z-1) and B(z-1) are polynomial of order na
and nb respectively in the backward shift operator
in time, z-1. A(z-1) and B(z-1) have the following
forms:
33
 Drive by Wire Systems

A( z −1 ) = d 0 + d1 z −1 + d 2 z −2 + ... + d na z − na Multiplying equation (4.20) with θ(t+j) and

rearranging that equation, we obtain:
B ( z −1 ) = n0 + n1 z −1 + n2 z −2 + ... + nnb z − nb (4.16)
θ(t + j) = Fj(z–1)θ(t) + Ej(z–1)(n0 + n1z–1
In order to make a prediction of the future + n2z–2 + n3z–3 + n4z–4)∆im(t – j + 1)
output of the road wheel angle, the Diophantine where, ∆ = (1 – z–1) (4.21)
identity is used to derive the j-step ahead predic-
tion of ∆U(t+j). Equation (4.21) predicts the value of the pinion
angle θ in the future ( j - time step ahead).
1 = Ej(z–1) A(z–1) + z–j Fj(z–1) (4.17)
θ(t + j) = F × θ(t) + EB × ∆im(t – j + 1) (4.22)
Where Ej and Fj are uniquely defined polynomials
for a given A(z-1) and the prediction interval j. In The matrices F ≠ R N×5 and EB ≠ R N × (N + 5)are
the present work, the recursive technique has been calculated by using the MATLAB script.
used to obtain Ej and Fj (Hasan & Anwar, 2008).
This makes the procedure computationally very
efficient. It has been shown that with increasing Simulation RESULTS
j only the highest order term in Ej+1(z-1) changes
while the rest of the coefficients remain the same
Initialization of Vehicle Model for
in Ej(z-1). Therefore, we can write:
Simulation
Ej+1(z–1) = Ej(z–1) + ej z–j (4.18)
In the full vehicle model, the parameter initializa-
tion (Anwar & Chen, 2007) is set as follows: Fw
where,
= 2 (N-m); Jw = 0.5 (kg-m2); bw = 20; k m = 0.5; I
= 3136 (kg-m2); Cf = 18000 (N/rad); Cr = 47000
Ej(z–1) = e0 + e1z–1 + e2z–2 + ... + ej–1z–( j–1)
(N/rad); m = 1250 kg; V = 8 (m/s); a = 1.05 (m);
b = 1.71 (m); C0 = (Cf + Cr) (N/rad); C1 = (bCr
In the degree of polynomial A(z-1) is na, then the
– aCf) (N-m/rad); C2 = (a2Cf + b2Cr) (N-m2/rad);
degree of Fj(z-1) becomes na. The coefficients of
tp = 0.0381 (m); tm = 0.04572 (m); C3 = (tp + tm)Cf
the polynomial Fj(z-1) may then be denoted as:
(N-m/rad); K = 80; T = 0.005 (sec).
The vehicle model presented in this thesis is
Fj(z–1) = f j,0 + f j,1z–1 + f j,2z–2 + ... + f j,naz–na (4.19)
evaluated on a validated SIMULINK model (An-
war & Chen, 2007). The cornering coefficients
Steering Angle Prediction have been considered for a light weight passenger
Front Wheel Drive (FWD) car. Therefore the mass
GP based prediction supposed to be executed
and dimensions of the vehicle were perceptibly
for the discrete model. Hence the forth order
standard for the small passenger car. The simula-
vehicle model has been discretized to make itself
tion process has been appraised for a slow moving
compliant with the predictor. From the equations
vehicle and the pneumatic and mechanical trails
(4.15)-(4.16), the Diophantine prediction equation
have been carried out from the standard passen-
( j-step ahead predictor) is given by,
ger car tires specification. The driver’s factor is
chosen as nominal driving effect. But the sample
Ej(z–1)(d0 + d1z–1 + d2z–2 + d3z–3 + d4z–4)
time can be deviated according to the dynamic
∆ + z–j Fj(z–1) = 1 (4.20)
requirement of the system.

34
Drive by Wire Systems

Faults Types and Their Implications stage of the system operation otherwise according
to their nature; they are gradually increased to a
Two major fault types have been introduced into larger extent that could be difficult to control. The
a steer by wire system, namely: amplitude change type incipient faults are shown
in Figures 19a and 20a.
• Permanent fault These two most available types faults are
• Incipient fault introduced into the vehicle model system to
verify the FDIA methodology as well as provide
The persistent fault type is illustrated in Figure the efficient proposition of SMO and GP based
18a. Amplitude change fault can be either positive predictor for the SBW system.
amplitude change or negative amplitude change In order for the SBW system to be robust, the
types. Incipient fault should be handled at the early sensor measurements must be accurate and reli-

Figure 18a. Persistent zero sensor fault introduced in one of the two physical sensors with the fault state
displayed on the right

30
Analytical Sensor
Physical Sensor-1
20
Physical Sensor-2
Steering Angle (Degree)

10

-10

-20

-30
0 2 4 6 8 10 12 14 16 18 20

Time (Sec)

Figure 18b. FDIA output of sensor data after removing the persistent zero introduce fault in one physi-
cal sensor

25
Desired Road Wheel Angle
20 FDIA Output

15
Steering Angle (Degree)

10

-5

-10

-15

-20

-25
0 2 4 6 8 10 12 14 16 18 20

Time (Sec)

35
 Drive by Wire Systems

able. Therefore, any faulty signal must be elimi- of three signals are required for this scheme to
nated to prevent undesirable steering effects. The work. The sensor signals are compared against
Fault Detection, Isolation, and Accommodation each other in real-time to determine the faulty
algorithm (FDIA) (Anwar & Chen, 2007) used signal where majority is assumed to be correct.
in this paper is able to handle single point faults This is based on the assumption that the event of
without interrupting the functionality of the SBW a sensor failure is rare and the event of multiple
system. This algorithm can be easily modified simultaneous sensor failures is extremely rare.
to handle multiple faults if more than three-sen- This algorithm can determine which sensor has
sor signals are compared. The FDIA algorithm failed by comparing its value against other sensors’
implemented in SIMULINK below is based on values. This algorithm can manage hard-failures as
a majority voting scheme in which a minimum well as soft-failures. Hard-failure is characterized

Figure 19a. Negative amplitude change type incipient fault introduced in one of the two physical sensors
with the fault state displayed on the right

30
Analytical Sensor
Physical Sensor-1
20
Physical Sensor-2
Steering Angle (Degree)

10

-10

-20

-30
0 2 4 6 8 10 12 14 16 18 20

Time (Sec)

Figure 19b. FDIA output of sensor data after removing the negative amplitude change type incipient
fault in one physical sensor

25
Desired Road Wheel Angle
20 FDIA Output

15
Steering Angle (Degree)

10

-5

-10

-15

-20

-25
0 2 4 6 8 10 12 14 16 18 20

Time (Sec)

36
Drive by Wire Systems

Figure 20a. Positive amplitude change type incipient fault introduced in one of the two physical sensors
with the fault state displayed on the right

50
Analytical Sensor
40 Physical Sensor-1
Physical Sensor-2
30

Steering Angle (Degree)

20

10

-10

-20

-30

-40

-50
0 2 4 6 8 10 12 14 16 18 20

Time (Sec)

Figure 20b. FDIA output of sensor data after removing the positive amplitude change type incipient
fault in one physical sensor

30
Desired Road Wheel Angle
FDIA Output
20
Steering Angle (Degree)

10

-10

-20

-30
0 2 4 6 8 10 12 14 16 18 20

Time (Sec)

by an abrupt or sudden sensor failure and soft- angle to detect and control them. And some simu-
failure is characterized by biases or drifts in the lations were observed to verify the advantages
signal over time. When a sensor fails, its signal is of higher prediction horizon into the dynamic
no longer used in the road wheel angle calculation. systems. The SBW controller, the yaw angle ob-
In such a situation the driver would be alerted of server, the road wheel angle estimator, the FDIA
the sensor failure, but would also still be able to algorithms, and the Generalized Predictive (GP)
maintain safe control of the vehicle. based predictor are combined with a simplified
A number of simulation runs were performed vehicle model with an SBW actuation system. The
in order to evaluate the developed methodology combined model was given a sinusoidal steering
of estimating and then predicting the road wheel input. Fault was then injected to one of the three
road wheel angle sensors whether one of them

37
 Drive by Wire Systems

Figure 21a. Effect of prediction horizon for the amplitude change type incipient fault (lower magnitude)
into the physical sensor

4.5

Fault detection time by FDIA (Sec)

4.48

4.46

4.44

4.42

4.4

4.38
0 1 2 3 4 5 6 7 8 9 10

Prediction Horizon, j

Figure 21b. Effect of prediction horizon for the amplitude change type incipient fault (higher magnitude)
into the physical sensor.
23.25
Fault detection time by FDIA (Sec)

23.2

23.15

23.1
0 1 2 3 4 5 6 7 8 9 10

Prediction Horizon, j

was analytical sensor. Fault flags and the output sensor though m-file. These two figures show
of the road wheel angle from the FDIA block are the nature of incipient fault as gradual increment
then recorded. with time. According to the proposed estimated
Persistent zero introduce fault was injected and predicted fault tolerant control methodology,
by making the sensor out a constant value. It these faulty signals were eliminated and the FDIA
was considered a faulty signal by the FDIA block output signals (Figures 19b and 20b) verify the
and was eliminated from the FDIA block output. efficiency of the GP based prediction.
Figure 19b shows the FDIA block output with the Now to show the effect of GP based prediction
removal of the fault from the system. with the different prediction horizons, it is con-
Figures 19a and 20a show the sensor signals venient to observer the effects with the incipient
of the amplitude change type incipient faults. fault. Because by definition the incipient fault is
These faults are introduced into the 2nd physical gradually increases with time, the consequence

38
Drive by Wire Systems
of prediction horizon is more prominent. Figures
21a-b show the effect of prediction horizon for
the various magnitudes of amplitude change type
incipient fault. It illustrates that the fault detection
time for incipient faults decreases significantly
with increase in the prediction horizon, thereby
improving the efficiency of fault detection.
This research work has demonstrated that it
is possible to increase the level of robustness of a
fault tolerant SBW system via successful imple-
mentation of SMO and GP based fault tolerant
control. In the present work, a nonlinear Sliding
Mode Observer (SMO) was designed and imple-
mented to estimate the road wheel angle with
available sensor output of yaw angle and motor
input current. Through the proposed predictive
analytical redundancy based fault detection and
isolation algorithm, an extra level of redundancy
was possible without any extra hardware. The
proposed algorithms rendered the SBW system
with a robust fault tolerant system as evidenced
by the simulation results. It was also observed
that the reliability of the proposed methodology
increases with the increase of prediction horizon
as it reduces the detection time for the faulty
sensors.
FUTURE TRENDS
Various drive by wire systems in the same vehicle
may share a diversity of sensors and actuators
in order to bring down the overall system cost.
This poses a unique challenge/opportunity for
integration of all the sensory information via data
fusion methodologies that will take the analytical
redundancy features of these systems to a new
level. In addition, integration of fault diagnosis
and control bring about new challenges since
the control objectives and diagnosis objectives
may have conflicting features. For example, in
an SBW system, the control objectives could be
noise/disturbance rejection and command track-
ing whereas the diagnosis objective could be
faulty signal tracking and noise reduction (Zheng
et al, 2006). Integrated control of various drive
by wire systems (e.g. TBW, BBW, SBW) is also
being investigated for future application on DBW
vehicles (Abe & Mokhiamar, 2007; Sameshima
et al, 2004; Lu & Hedrick, 2005; Setlur et al,
2006). Robust sliding mode control of a throttle
body for drive by wire operation of automotive
engines has been critical in the success of TBW
systems (Rossi et al, 2000).
Future communication bus for DBW systems
are also being developed and implemented (Sha-
heen et al, 2003). We will focus on the future trend
of this important DBW technology which is very
critical to the success of these systems.
Modern automotive system needs both time-
triggered (e.g. FlexRay) and event-triggered (e.g.
CAN) traffic. We analyze the relevant feature of
these communication buses below.
Controller Area Network (CAN):
• Very good in handling event-triggered traffic
• Can be found on many places in a car, rang-
ing from engine control and ABS-systems
to body and comfort systems
• Has also been extended with a time-trig-
gered session layer, as TT-CAN, to support
time-triggered traffic
• May provide a smooth transition using CAN
in the first generation “wet” DBW where a
hydraulic backup is used
• However, since it relies on the CAN lower
layers, it is lacking fault-tolerant mecha-
nisms and bandwidth capabilities
Time Triggered Protocol (TTP):
• Highly fault-tolerant network
• Intended for safety-critical systems such as
DBW and avionics
• Implemented fault tolerant mechanisms,
such as atomic broadcast using membership
service, distributed clock synchronization
and bus guardians
• Ensures that there can be no single point of
failure
39

Figure 22. Application of communication bus protocols in a drive by wire vehicle
• Very deterministic at the cost of being less
flexible in terms of message transmissions
• Weakness: Inflexible message transmis-
sion
FlexRay:
• Allows for both time-triggered and event-
triggered message transmissions
• Developed with safety-critical applications
in mind, just like TTP
• Hence, using FlexRay it is possible to de-
velop both DBW systems and powertrain
systems, reducing the need for several field
bus technologies
• Has the biggest potential for becoming the
next generation automotive network for
safety-critical fault-tolerant applications,
mainly because it is heavily backed up by
industrial partners and pushed by most major
industrial automakers
40
Drive by Wire Systems
Although FlexRay is developed and mainly
intended for safety-critical communications such
as DBW, it is possible to use it in several other
automotive applications. The possible solution
is the implementation of higher-layer extensions
on top of FlexRay, which will make it suitable
for a broader class of automotive applications.
The higher-layer extensions could be especially
devised for better supporting dynamic traffic
scenarios. Figure 22 illustrates an in-vehicle
network architecture with FlexRay. As shown,
other non-time triggered communication buses
can be used for non-safety critical applications
(e.g. door locks, climate control).
Considering the most likely application do-
mains, mixed configurations will be dominant.
Minislotting offers FlexRay with dynamic seg-
ment capability. With FlexRay, the dynamic
segment is scheduled in a cyclic manner using a
minislotting technique similar to Byteflight. As
the dynamic segment is started, the minislotting
Drive by Wire Systems
mechanism is used to schedule the messages. In
order for the minislotting mechanism to work,
all messages must have unique identifiers (IDs),
like in CAN. Moreover, all nodes in the system
keep a slot-counter. As a dynamic segment is
started, all slot counters are reset to 0 indicating
the start of the first dynamic slot. Whenever a
node has a message with an ID matching the slot
counter, the node will send its message. Once
the message has been sent, then all the nodes in
the system will detect a new dynamic slot and
increase their slot counters by one. If there is no
message transmitted within a short time D after
the initiation of a dynamic slot, a new dynamic slot
is detected and all the slot-counters are increased
again, etc. In this way messages are scheduled
in increasing order based on their IDs, serving
the lower value IDs first. Hence, a low value ID
gives a high priority to a message and thus a
higher chance of being transmitted. A dynamic
slot can be of varying size depending on the size
of the message transmitted in the dynamic slot.
However, the slot-counters might not reach their
maximum value due to the fact that the dynamic
segment does not fit all possible messages. As
the dynamic segment is of fixed size, depending
on the number of dynamic slots used and their
individual lengths, the slot-counter will only reach
Figure 23. (a) Composability based on ECU nodes; (b) Composability of data from different channels
(a)
a certain value. Therefore some messages might
have to wait one or more communication cycles
in order to be scheduled. It should be noted here
that the slot-counters always begin with value 0
for all dynamic segments.
In summary, FlexRay offers high speed time
triggered communication. One of the constraints
of current automotive electronic communications
systems is the limited bandwidth. For a start, the
FlexRay Consortium selected a data transfer rate
of 10 MBit/s on each of the two communication
channels as a viable trade-off between the need for
high performance and the economic constraints.
As the current FlexRay specification is open for
higher bit rates, faster physical layers are possible
in the future.
FlexRay communication bus can also be
designed for fault-tolerance. Scalable system
fault-tolerance can be achieved via the support of
either single or dual channels. Since the needs of
automotive applications differ quite substantially
in terms of fault-tolerance, FlexRay can do the
splits between distributed fault-tolerant and non
fault-tolerant systems:
• Single-channel system (doubling the data
throughput)
(b)
41

Figure 24. Possible configuration of FlexRay bus architecture
• Dual-channel system (full redundancy)
• Dual-channel system with mixed connectiv-
ity (some nodes connect to both channels,
some connect only to one)
For safety-critical systems, communication
reliability cannot be based on probabilistic mea-
sures. A CAN system is not possible to calculate
the communication delay except for some high pri-
ority message. FlexRay guarantees the transport
for safety-critical message with known latency.
FlexRay offers deterministic system design.
FlexRay combines a time-triggered along with
an event-triggered system. Based on an extended
TDMA (time-division multiple access) media ac-
cess strategy, a so-called communication cycle is
set up, consisting of a mandatory static segment,
an optional dynamic segment and two protocol
segments called symbol window (optional) and
network idle time (mandatory). Application cycles
can be formed by one or more communication
cycles, which are executed from start-up of the
42
Drive by Wire Systems
network until its shutdown. A time-triggered com-
munication system depends upon a common time
base, the so-called global time. This time base has
to be shared by all communication controllers in
the network. In order to synchronize the clock,
certain messages are marked as synchronization
messages in the static segment. By means of a
special algorithm, the local clock-time of a com-
ponent, for example, a communication controller,
is corrected in such a way that all local clocks run
synchronously to a global clock.
A bus guardian ensures that the data traffic
moves in a timely fashion. Access to the com-
munication bus has to be carefully managed,
especially with regard to safety relevant applica-
tions. An independent component that protects a
channel from interference is needed. It must limit
the times that any communication controller can
transmit to those pre-assigned intervals, in which
it is allowed to do so. This component is the Bus
Guardian. Basically, FlexRay systems allow to use
bus guardians with independent clocks and are
Drive by Wire Systems
configured in a way that an error in the controller
cannot influence the guardian and vice versa.
Composability
As mentioned above, currently used systems are
event-triggered, which means that whenever a
station has data to send it tries to access the bus.
Adding a new message is convenient to realize
by assigning an identifier to it. The pitfall comes
with the testing of such a system since any new
message or change of the application modifies the
behavior of the communication system. FlexRay
puts a high demand on composability of the new
communication system. Both the static and dy-
namic segments strictly adhere to the exclusive
ownership of slots, a key foundation for the com-
posability. As each ECU owns slots exclusively, it
is possible to develop ECUs autonomously and to
later integrate them without side effects. Figure
23 (a) and (b) illustrate the composability feature
of FlexRay.
The final FlexRay system architecture may
take the form shown in Figure 24.
A number of vendors already started marketing
FlexRay systems. Some available hardware are:
• National Instrument and Vector FlexRay
CARDBUS card
• DeComsys FlexRay Node
• dSPACE MicroAutoBox
• Fujitsu Evaluation Kit
• TI DPS with FlexRay communication in-
terface
• Freescale integrated and stand-alone
FlexRay controllers
FlexRay Communication: Features
and Challenges that Lie Ahead
It offers a fault tolerant communication system
for advanced automotive control applications.
Fault-tolerant clock synchronization is achieved
via a global time base. It offers Collision-free bus
access and guarantees message latency. It offers
Message oriented addressing via identifiers. Scal-
able system fault-tolerance is achieved via the
support of either single or dual channels. It has
been chosen as the standard by most OEM’s.
However, the following challenges still lie
ahead with FlexRay development:
• New protocol with new hardware & im-
mature software tools
• Little experience both internally and exter-
nally with the development of a FlexRay
communication system
• Nevertheless, internal development with
TTP system has provided a good basic
understanding of a real-time safety-critical
communication system
• Tedious definition of information transfer
using the FlexRay protocol
• Entering into a real-time world where the
design and definition of our system is very
different from past system designs
• Defining tasks to be executed within our
system
• Defining worst-case execution times for
these tasks
• Defining offsets for the execution of tasks and
deadlines when tasks should be complete
• Must work closely with other suppliers to
integrate into the overall FlexRay commu-
nication system of the vehicle
FlexRay System Design: Two Level
Design Approach
1. Cluster level design: At this design level,
the physical cluster layout, the subsystems
and the interaction between the subsystems
are defined. Start with system requirement.
Define the subsystems which are mapped
to the nodes. Specify the signals provided
by the subsystem, which are mapped to the
position of the FlexRay frame. Generate the
cluster specification.
43

2. Component level design: At this level, the
precise subsystem interaction as specified
in the cluster design is defined. Start with
subsystem requirement specified during
Cluster Level Design. Define the tasks that
implement the behavior of the subsystem.
The task specification must be consistent
with the signal specification from the cluster
design. Calculate the worst-case execution
time.
Some Design Considerations
• Numbers of subsystems (nodes): It needs at
least 4 nodes to make communication system
true fault tolerant. Increasing the number of
nodes will increase the cost. Nodes should
be added only when it is necessary.
• Numbers of tasks within the subsystem:
Number more available tasks gives the
designer more scheduling flexibility. More
tasks result in the schedule complexity and
increase switching overhead. Trade off has
Figure 25. FlexRay based distributed steer by wire control system
44
Drive by Wire Systems
to be made between number of tasks and
complexity/switching overhead.
• Communication cycle: The communication
cycle needs to meet the real time control
sampling time requirement. The cycle is
also restricted by bus speed, the number of
signals to be exchanged within the cycle, and
number of tasks to be executed within the
cycle. Considerable experiments need to be
conducted to validate final communication
cycle.
• Slot size: The larger the slot size, the fewer
low level buffer reads and writes required.
With larger slot sizes, the receiver might get
many signals that it does not need, wasting
buffer space. The proper slot size is applica-
tion dependent, and is the judgment of the
system engineer.
The culmination of a robust fault tolerant
DBW system can be a drive by wire system that
is designed based on FlexRay communication
bus. One such design for steer by wire system is
illustrated in Figure 25.
Drive by Wire Systems
CONCLUDING REMARKS
In what presented above, it is clear that drive
by wire systems have clear benefits over their
conventional counterpart in the area of safety
and performance. DBW systems can, in fact,
enhance the safety and performance of a vehicle
equipped without any additional hardware, since
these enhancements can be achieved through
software upgrades. While drive by wire systems
have made significant progress in fault tolerance
and reliability, these systems, particularly the
safety critical BBW and SBW systems, are still
some steps away from full commercialization.
One of the main reasons for this is the high cost
of these systems due to multiple redundancies in
sensors, actuators, microcontroller, and commu-
nications buses. As demonstrated in this chapter,
analytical redundancy / model based fault tolerant
control can reduce the total number of redundant
components to certain extent. However, further
reduction in the overall cost of the DBW system
may still be warranted so that the automakers
and their suppliers can competitively price these
systems when compared with the conventional
systems. The additional cost reduction can be
achieved via innovative actuator design, manu-
facturing process optimization for microproces-
sors and communications buses, and inexpensive
sensor development with diversity in the sensors.
Historically the cost of microprocessors and com-
munication electronics have seen significant drop
in cost year over year. Therefore, the introduction
of BBW and SBW systems in passenger vehicles
is almost certain. However, the timing of such
commercialization will largely depend on further
research and development of these systems in
order to address the aforementioned challenges
and the willingness of the automakers and their
suppliers to take on such a task to mass produce
DBW systems.
REFERENCES
Abe, M., & Mokhiamar, O. (2007). An integration
of vehicle motion controls for full drive-by-wire
vehicle. Proc. IMechE Part K: J. Multi-body
Dynamics, 221, 117-127.
Anon (2003, December). Cutting down on the
chips. Automotive Engineer, (pp. 48-51).
Anwar, S. (2005). Generalized Predictive Control
of Yaw Dynamics of a Hybrid Brake-By-Wire
Equipped Vehicle. International Journal of Me-
chatronics, 15(November), 1089-1108.
Anwar, S. (2006). Anti-Lock Braking Control
of a Hybrid Brake-By-Wire System. Journal of
Automobile Engineering – IMechE Proceedings
Part D., 220(8), 1101-1117.
Anwar, S., & Chen, L. (2007). An Analytical
Redundancy Based Fault Detection and Isolation
Algorithm for a Road-Wheel Control Subsystem
in a Steer-By-Wire System. IEEE Transactions on
Vehicular Technology, 56(5), 2859-2869.
Bajpai, G., Chang, B.C., & Lau, A. (2001, Sep-
tember). Reconfiguration of Flight Control Sys-
tems for Actuator Failures. IEEE AESS Systems
Magazine, (pp. 29-33).
Chang, S. C. (2007). Adoption of state feedback
to control dynamics of a vehicle with a steer-by-
wire system. Proc. IMechE Part D: J. Automobile
Engineering, 221, 1-12.
Daniels, J. (2005). Practicalities. Automotive
Engineer, 30(10), 33.
Davis, B. (2001). Making the right connections.
Professional Engineering, February 28, (pp.
36-39).
Fowler, D. (2003). Down to the wire. The Engi-
neer, (pp. 29-30).
Gadda, C. D., Laws, S. M., & Gerdes, J. C. (2007).
Generating Diagnostic Residuals for Steer-by-
45
 Drive by Wire Systems

Wire Vehicles. IEEE Trans. on Control Sys. Tech.,
15(3), 529-540.
Haggag, S. Alstrom, D., Cetinkunt, S., & Egelja,
A. (2005). Modeling, Control, and Validation of
an Electro-Hydraulic Steer-by-Wire System for
Articulated Vehicle Applications. ASME/IEEE
Trans. on Mechatronics, 10(6), 688-692.
Hasan, M. S., & Anwar, S. (2008). Sliding Mode
Observer Based Predictive Fault Diagnosis of a
Steer-By-Wire System. Proceedings of the 17th
Int’l Federation of Automatic Control (IFAC)
World Congress, Seoul, S. Korea.
Hebden, R. G., Edwards, C., & Spurgeon, S. K.
(2004). Automotive steering control in a split-mu
maneuver using an observer based sliding mode
controller. Vehicle System Dynamics, 41(3), 181-
202.
Higgins, A., & Koucky, S. (2002). Mercedes
pumps ‘Fly By Wire” brakes into new roadster.
Machine Design, (p. 26).
Hoseinnezhad, R. (2006). Position Sensing in
Brake-By-Wire Calipers Using Resolvers. IEEE
Transactions on Vehicular Technology, 55(3),
924-932.
Isermann, R., Schwarz, R., & Stolzl, S. (2002,
October). Fault tolerant drive-by-wire systems.
IEEE Control Systems Magazine, (pp. 64-81).
Kendall, J. (2005). Current Technology. Com-
mercial Motor, 201(5117), 62-64.
Lee, A. (2003). Driven by wire. The Engineer,
5(8), 28-29.
Limpibunterng, T., & Fujioka, T. (2002). A new
design approach for steer-by-wire system by dual
port system. Vehicle System Dynamics, Suppl.
37, 197-208.
Limpibunterng, T., & Fujioka, T. (2004). Bilateral
driver model for steer-by-wire controller design.
Vehicle System Dynamics, Suppl. 41, 381-390.
Lu, X-Y., & Hedrick, J. K. (2005). Heavy-duty ve-
hicle modelling and longitudinal control. Vehicle
System Dynamics, 43(9), 653–669.
Margolin, B. (1997). Analog, mixed-signal ICs
steering drive-by-wire toward reality. Computer
Design, 36(12), 31-34.
Oh, S-W., Chae, H-C., Yun, S-C., & Han, C-S.
(2004). The design of a controller for the steer-by-
wire system. JSME International Journal, Series
C, 47(3), 896-907.
Ross, P. (2003, February). Top Ten Techno-Cool
Cars. IEEE Spectrum, (pp. 30-35).
Rossi, C., Tilli, A., & Tonielli, A. (2000). Robust
Control of a Throttle Body for Drive by Wire Op-
eration of Automotive Engines. IEEE Transactions
on Control System Technology, 8(6), 993-1002.
Sameshima, H., Ogasa, M., & Yamamoto, T.
(2004). On-Board Characteristics of Lithium Ion
Batteries for improving energy regenerative ef-
ficiency. Quarterly report of RTRI, 45(2), 45-52.
Segawa, M., Kimura, S., Kada, T., & Nakano,
S. (2004). A study of the relationship between
vehicle behavior and the steering wheel torque on
steer by wire vehicles. Vehicle System Dynamics,
Suppl. 41, 202-211.
Setlur, P. J., Wagner, R., Dawson, D. M., & Bra-
ganza, D. (2006). A Trajectory Tracking Steer-
by-Wire Control System for Ground Vehicles.
IEEE Transactions on Vehicular Technology,
55(1), 76-85.
Shaheen, S., Heffernan, D., & Leen, G. (2003).
A comparison of emerging time-triggered proto-
cols for automotive X-by-wire control networks.
Proc. Instn Mech. Engrs Part D: J. Automobile
Engineering, 217, 13-22.
Stanton, N., & Marsden, P. (1997). Drive-by-wire
systems: some reflections on the trend to automate
the driver role. Proceeding of Inst. Mech. Engrs.
Part D, 211, 267-276.

46
Drive by Wire Systems
Wagstaff, I. (1999, April). Braking new ground.
Automotive Engineer, (pp. 61-62).
Yih P., & Gerdes, J.C., Modification of Vehicle
Handling Characteristics via Steer-by-Wire. IEEE
Trans. on Control Sys. Tech., 13(6), 965-976.
Zheng, B., & Anwar, S. (2008). Fault Toler-
ant Control of the Road Wheel Subsystem in a
Steer-By-Wire System. International Journal
of Vehicular Technology, Article ID 859571,
doi:10.1155/2008/859571, 1-8.
Zheng, B., Huang, M., Daugherty, B., & Anwar,
S. (2006). Integrated Steer-By-Wire control and
diagnostic system. Proceedings of the American
Control Conference, Minneapolis, Minnesota.
47
48

Chapter III
Electromagnetic Compatibility
Issues in Automotive
Communications
Todd H. Hubing
Clemson University International Center for Automotive Research, USA

ABSTtract

Engineers and engineering managers involved in the design of automotive electronic systems need to
have a basic familiarity with electronic noise and the electromagnetic compatibility (EMC) issues that
influence the design and the performance of automotive systems. When EMC issues are addressed early
in a product’s design cycle, the resulting designs often meet all EMC requirements without significant
cost or performance problems. EMC problems detected after a product has been built and tested, on the
other hand, can be very difficult and costly to fix. This chapter reviews automotive EMC requirements
and discusses the design of automotive electronics for EMC. The objective of the chapter is to provide
non-EMC engineers and engineering managers with basic information that will help them recognize
the importance of designing for electromagnetic compatibility, rather than addressing electronic noise
problems as they arise.

Introduction and fuel economy. They also navigate, entertain,

There are more electronic systems in the average
car than there are in the average house. Dozens
of computers, connected to sensors and actuators
through kilometers of wiring, control everything
from the engine to the brakes. Electronic systems
determine the vehicle’s performance, handling,
communicate and help to keep the driver and
passengers safe and comfortable.
Electromagnetic interference is a significant
concern to the automotive system designer.
Packing all of these electronic systems in close
proximity and getting them to function reliably
in all possible situations is becoming increas-

Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.
Electromagnetic Compatibility Issues in Automotive Communications
ingly difficult. Microcontroller circuits generally
operate at radio frequencies. They emit radio
frequency noise and are susceptible to radio
frequency fields. In an automotive environment,
electronic systems must be able to function in
the presence of intentional radio transmitters
that are mounted on the vehicle, located along
the vehicle’s route, or carried into the vehicle by
the driver or passengers.
This chapter provides an overview of the elec-
tromagnetic compatibility requirements that au-
tomobiles must meet. Significant electronic noise
sources and coupling paths are described, and
aspects of the major automotive communication
buses that affect electromagnetic compatibility are
reviewed. The chapter concludes with a discussion
of design features that affect the electromagnetic
compatibility of automotive systems.
AutomotiC REQUuir
In order to help ensure that a vehicle will not
have problems with electromagnetic interference,
extensive electromagnetic compatibility (EMC)
testing of each model is performed as the vehicle
and its components are being developed. There are
a number of EMC test standards for automobiles.
The most widely referenced standards for unin-
tentional electromagnetic emissions from vehicles
are known as CISPR 12 (IEC, 2008) and CISPR
25 (IEC, 2002). CISPR, which is an acronym for
Comité International Spécial des Perturbations
Radioélectriques, is an organization that develops
and supports a variety of international standards
governing EMC. CISPR 12 defines test procedures
and emission limits for radiated emissions from
vehicles. CISPR 25 defines conducted and radiated
emissions tests for vehicular components.
CISPR 12 and CISPR 25 only address elec-
tromagnetic emissions. There are no CISPR stan-
dards for evaluating the susceptibility of vehicles
or components to electromagnetic disturbances.
Electromagnetic susceptibility (or immunity)
standards are developed and maintained by an-
other international standards organization, the
International Electro-Technical Committee (IEC).
Various IEC standards describe procedures for
evaluating the susceptibility of a vehicle or com-
ponent to various disturbances such as strong RF
(radio frequency) fields, power bus transients, and
electrostatic discharge.
CISPR 12
According to the introduction in the standard,
“CISPR 12 has been developed to serve the road
vehicle and related industries with test methods
and limits that provide satisfactory protection
for radio reception.” It is designed to protect
broadcast receivers operating at frequencies
between 30 MHz and 1 GHz from unintentional
electromagnetic emissions from a vehicle located
10 meters or more away. The standard applies to
automobiles, trucks and boats, but not to aircraft
or trains.
The test procedure calls for the vehicle to be
parked in a flat area free of buildings, trees or
other objects that might reflect electromagnetic
fields, as illustrated in Figure 1. Electromagnetic
emissions from the vehicle are measured with an
antenna located 10 meters away. Measurements
are made with the vehicle’s engine idling at 1500
RPM, and also with the engine off but the rest
of the vehicle powered on. Electric propulsion
vehicles are measured on a dynamometer. For
hybrid vehicles, a separate test is performed for
each form of propulsion.
Making electromagnetic measurements in an
open field presents a few technical challenges
including weather and ambient noise from local
transmitters such as radio and television stations.
For this reason, the CISPR 12 standard allows
these measurements to be made in an absorber
lined shielded enclosure (ALSE) provided these
measurements can be shown to correlate to open
field measurements. Most large automotive manu-
facturers operate or lease an EMC test facility
with an ALSE.
49
 Electromagnetic Compatibility Issues in Automotive Communications
Figure 1. CISPR 12 Radiated emissions test configuration
10 METERS
CISPR 25
The CISPR 25 standard is designed to evaluate
the conducted and radiated emissions from indi-
vidual components and systems within a vehicle.
It covers frequencies from 150 kHz to 1 GHz and
includes an in-vehicle test procedure as well as
bench-top tests.
For the in-vehicle test, the test receiver is
connected to an antenna mounted on the vehicle.
This antenna is an existing radio antenna when
measurements are made at frequencies within
the radio receiver bandwidth. Out-of-band test-
ing is done with a monopole antenna mounted
in the same location. Each electronic system be-
ing evaluated is powered on and operated while
monitoring the voltage induced at the antenna
terminals. If the voltage exceeds the limit speci-
fied in the standard, the system under evaluation
is non-compliant.
Figure 2. CISPR 25 radiated emissions ALSE test configuration
ALSE
ANTENNA
AN EUT
1 METER
50
ANTENNA
EMITEST RECEIVER
Bench-top tests are done on a 0.9-meter
tall table with a metal top located in a shielded
room as illustrated in Figure 2. The bench top is
bonded to the wall or floor of the shielded room
and the equipment under test (EUT) is placed on
the table. Both the EUT and its wiring harness
are insulated from the bench top by a 5-cm thick
sheet of dielectric material. The EUT is powered
through an artificial network (AN) that provides
a stable well-defined power bus impedance at
measurement frequencies. Another box attached
to the wiring harness emulates all the sensors,
actuators and communication devices needed to
operate the EUT.
CISPR 25 describes four bench-top tests, two
that measure conducted emissions from the EUT
and two that measure radiated emissions. The
EMC test plan for a particular product may call
for all or a subset of these tests to be performed. In
one of the conducted emissions tests, the voltage
EMITEST RECEIVER
Electromagnetic Compatibility Issues in Automotive Communications
on each wire in the wiring harness is measured
relative to the ground plane. In the other conducted
test, the common-mode (i.e. net) current flowing
on the wiring harness is measured.
The two radiated emissions tests do not actu-
ally measure radiated emissions directly. In the
first test, called “Radiated Emissions – ALSE
Method,” the EUT is set up on the bench top as
described above and measurements are made with
an antenna located 1 meter away in the shielded
room. Measurements made between 150 kHz and
30 MHz are done with a short monopole antenna
mounted in a narrow ground plane that extends
out from the bench top. Between 30 MHz and
200 MHz, a biconical antenna is used. Above
200 MHz, a log-periodic dipole array antenna
is employed. At a 1-meter test distance, none of
these antennas is in the radiated far-field of the
EUT and its harness; therefore this procedure does
not measure the actual radiated field strength.
However, the voltage measured at the antenna
terminals is an indication of the field strength
at the antenna location, which (in many cases)
correlates to the likelihood that a particular EUT
will be the source of a radiated emissions problem
in the vehicle.
The second radiated emission test defined in
CISPR 25 employs a device known as a TEM
cell. TEM is an acronym for “Transverse Electro-
Magnetic” and a TEM cell is a large waveguide
designed to support only TEM mode propagation
at frequencies below a cut-off frequency deter-
mined by the dimensions of the cell. As indicated
in Figure 3, the EUT is placed inside the TEM
Figure 3. CISPR 25 radiated emissions TEM cell test configuration
EUT
cell and configured in basically the same way
that it would be configured on the bench top.
The test receiver measures the voltage induced
at the TEM cell termination and this voltage is
compared to the limit specified in the standard.
CISPR 25 specifies two possible configurations
for the wiring harness, one that emphasizes emis-
sions from the harness and one that minimizes
the contribution from the harness.
Like the ALSE tests, TEM cell measurements
do not directly measure radiated emissions. In-
stead they measure the electric and magnetic field
coupling to the TEM cell. These measurements
generally do not correlate with the results of true
radiated emissions tests, but they can provide
useful information about the nature of the field
coupling from the EUT, e.g. (Tippet & Chang,
1976), (Crawford & Workman, 1980), (Sreeniva-
siah et. al, 1981), and (Deng et. al., 2007).
Susceptibility (Immunity) Tests
As indicated earlier, there are a number of pub-
lished test procedures for evaluating the suscepti-
bility of electronic components to electromagnetic
disturbances. The International Standards Organi-
zation (ISO) maintains a number of standards for
evaluating the performance of vehicles or vehicle
components when subjected to electromagnetic
interference. For example, (ISO 11452-2, 2005)
and (ISO 11452-3, 2001) describe ALSE and TEM
cell RF immunity tests with test configurations
similar to the CISPR 25 bench-top and TEM cell
tests. A transmitter drives the antenna (Figure 4)
EMI TEST RECEIVER
51
 Electromagnetic Compatibility Issues in Automotive Communications
Figure 4. ISO 11452-2 radiated immunity test set-up
ALSE
AN EUT
or TEM cell to produce a strong EM field in the
vicinity of the EUT. The strength of the applied
field is increased until the limit is reached or a
failure is observed. The limit depends on how
critical the EUT is and what type of failure oc-
curs. Minor “annoyance” failures may be allowed
at relatively low field strengths, but failures that
might affect the safety of the vehicle occupants
must not occur even when the applied field strength
is relatively high.
(ISO 11452-4, 2005) describes a bulk current
injection (BCI) test, where common-mode current
is injected on the wiring harness, as illustrated in
Figure 5. This type of immunity testing is very
Figure 5. ISO 11452-4 bulk current injection test set-up
SIGNAL GENERATOR
AMPLIFIER
AN
52
ANTENNA
SIGNAL GENERATOR
AMPLIFIER
popular in the automotive industry, because most
automotive components are connected to a wir-
ing harness and most EM susceptibility problems
result from noise brought into the component
through the harness.
The waveforms used for the component-level
susceptibility tests described above are continuous
wave (CW) signals that are swept in frequency
with different forms of modulation. (ISO 7637-2,
2004) and (ISO 7637-3, 2007) describe transient
immunity tests, where the noise coupled to the har-
ness is short pulses. (ISO 10605, 2008) describes
an electrostatic discharge (ESD) test, where a
simulator is used to create very short transients that
EUT
Electromagnetic Compatibility Issues in Automotive Communications

Table 1. ISO EM susceptibility test standards

Document Number Title
ISO 11451-2 Road vehicles — Vehicle test methods for electrical disturbances from narrowband radiated electromagnetic
energy — Part 2: Off-vehicle radiation sources
ISO 11451-3 Road vehicles — Electrical disturbances by narrowband radiated electromagnetic energy — Vehicle test methods
— Part 3: On-board transmitter simulation
ISO 11451-4 Road vehicles — Vehicle test methods for electrical disturbances from narrowband radiated electromagnetic
energy — Part 4: Bulk current injection (BCI)
Road vehicles — Component test methods for electrical disturbances from narrowband radiated electromagnetic
ISO 11452-2
energy — Part 2: Absorber-lined shielded enclosure
Road vehicles — Component test methods for electrical disturbances from narrowband radiated electromagnetic
ISO 11452-3
energy — Part 3: Transverse electromagnetic mode (TEM) cell
Road vehicles — Component test methods for electrical disturbances from narrowband radiated electromagnetic
ISO 11452-4
energy — Part 4: Bulk current injection (BCI)
Road vehicles — Component test methods for electrical disturbances from narrowband radiated electromagnetic
ISO 11452-5
energy — Part 5: Stripline
Road vehicles — Component test methods for electrical disturbances from narrowband radiated electromagnetic
ISO 11452-7
energy — Part 7: Direct radio frequency (RF) power injection
Road vehicles — Component test methods for electrical disturbances from narrowband radiated electromagnetic
ISO 11452-8
energy — Part 8: Immunity to magnetic fields
Road vehicles — Electrical disturbances from conduction and coupling — Part 2: Electrical transient conduction
ISO 7637-2
along supply lines only
Road vehicles — Electrical disturbance by conduction and coupling — Part 3: Vehicles with nominal 12 V or 24 V
ISO 7637-3 supply voltage — Electrical transient transmission by capacitive and inductive coupling via lines other than supply
lines
ISO 10605 Road vehicles — Test methods for electrical disturbances from electrostatic discharge

are conductively coupled to all exposed metallic OEM-Specific Requirements

contact points on the EUT. Table 1 lists various
automotive electromagnetic susceptibility test
standards published by ISO.
The Society of Automotive Engineers (SAE)
also develops and maintains automotive EMC
standards. The American National Standards
Institute (ANSI), which represents the United
States on international standards committees,
has delegated their automotive standards writing
activity to SAE. Many components of the CISPR
and ISO automotive EMC standards were initially
published as SAE standards. For this reason, there
is a great deal of overlap between the SAE stan-
dards and other international standards. However,
there are distinct differences that prevent similar
standards from being used interchangeably. SAE
Vehicular EMC Standards are listed in Table 2.
The CISPR, ISO and SAE EMC standards are
based on test practices developed by automotive
companies to meet specific internal requirements.
They are the result of many compromises and they
have evolved slowly over the years. Most automo-
tive companies adapt or supplement these proce-
dures to meet their specific needs. For example, it
is not uncommon for an automotive manufacturer
to specify different test environments, test limits
or excitation waveforms from those specified in
the standards. Because of this, test laboratories
that make EMC measurements for one automotive
company may not be able to make the same type
of measurements for another company.
Many factors affect the electromagnetic cou-
pling to and from an electronic system. Small
variations in cable position, load impedance or

53
 Electromagnetic Compatibility Issues in Automotive Communications

Table 2. SAE EM susceptibility test standards

Document Number Title
SAE J551-11 Vehicle Electromagnetic Immunity—Off-Vehicle Source
SAE J551-12 Vehicle Electromagnetic Immunity—On-Board Transmitter Simulation
SAE J551-13 Vehicle Electromagnetic Immunity—Bulk Current Injection
SAE J551-15 Vehicle Electromagnetic Immunity—Electrostatic Discharge (ESD)
SAE J551-16 Vehicle Electromagnetic Immunity—Reverberation Chamber Immunity
SAE J551-17 Vehicle Electromagnetic Immunity—Power Line Disturbances
SAE J1113/1 Electromagnetic Compatibility Measurement Procedures and Limits for Components of Vehicles, Boats (Up to 15
M), and Machines (Except Aircraft) (50 Hz to 18 GHz)
SAE J1113/2 Electromagnetic Compatibility Measurement Procedures and Limits for Vehicle Components (Except Aircraft)—
Conducted Immunity, 15 Hz to 250 kHz—All Leads
SAE J1113/3 Conducted Immunity, 250 kHz to 400 MHz, Direct Injection of Radio Frequency (RF) Power
SAE J1113/4 Immunity to Radiated Electromagnetic Fields-Bulk Current Injection (BCI) Method
SAE J1113/11 Immunity to Conducted Transients on Power Leads
SAE J1113/12 Electrical Interference by Conduction and Coupling—Capacitive and Inductive Coupling via Lines Other than
Supply Lines
SAE J1113/13 Electromagnetic Compatibility Measurement Procedure for Vehicle Components—Part 13: Immunity to Electro-
static Discharge
SAE J1113/21 Electromagnetic Compatibility Measurement Procedure for Vehicle Components—Part 21: Immunity to Electro-
magnetic Fields, 30 MHz to 18 GHz, Absorber-Lined Chamber
SAE J1113/24 Immunity to Radiated Electromagnetic Fields; 10 kHz to 200 MHz—Crawford TEM Cell and 10 kHz to 5
GHz—Wideband TEM Cell
SAE J1113/26 Electromagnetic Compatibility Measurement Procedure for Vehicle Components—Immunity to AC Power Line
Electric Fields
SAE J1113/27 Electromagnetic Compatibility Measurements Procedure for Vehicle Components—Part 27: Immunity to Radiated
Electromagnetic Fields—Mode Stir Reverberation Method
SAE J1113/28 Electromagnetic Compatibility Measurements Procedure for Vehicle Components—Part 28—Immunity to Radi-
ated Electromagnetic Fields—Reverberation Method (Mode Tuning)
SAE J1113/42 Electromagnetic Compatibility—Component Test Procedure—Part 42—Conducted Transient Emissions

test environment for example can result in large Sourcof NOISE

changes in the coupled energy at any given fre-
quency. For this reason, it is not uncommon for
a system tested in one lab to perform much dif-
ferently than the same system tested in another
lab. Also, since the laboratory test environment
is different from the environment in which the
system will ultimately be installed, a system
which does well in a component-level test may
still cause a failure in the vehicle-level test.
Nevertheless, component level testing helps to
identify EMC problems early in the design stage
when it is still possible to make changes that
do not significantly impact the product cost or
development schedule.
No other consumer product packs as many inde-
pendently designed sources of electronic noise
in such close proximity as the automobile. The
intentional transmitters in a vehicle (e.g. cell
phones, citizen’s band radios and wireless com-
munications equipment) are responsible for their
share of electromagnetic interference; however
the electrical noise generated by the engine and
the hundreds of communications and control cir-
cuits in a vehicle can be an even greater source of
electromagnetic compatibility problems.

54
Electromagnetic Compatibility Issues in Automotive Communications
Microprocessors
There are dozens of microcontrollers, digital sig-
nal processors and other kinds of microprocessors
in a typical passenger car. Each processor’s speed
is determined by a digital clock that typically
switches at a frequency between a few megahertz
(106 times per second) and a few gigahertz (109
times per second). Energy at the clock frequency
and harmonics of the clock frequency is coupled
to virtually every electrical conductor near or
connected to the processor. The power gener-
ated at any given frequency associated with a
clock is generally on the order of milliwatts (10 -3
watts) and most of this power is delivered to its
intended destination and ultimately dissipated
as heat. However, a few nanowatts (10 -9 watts) of
radiated power are typically all that is necessary
to interfere with intentional radio receivers in
or near a vehicle. Therefore great care must be
taken to control or contain the energy generated
by the clocks and signals associated with every
microprocessor in the vehicle.
PWM Signals
A common way to transfer continuously variable
levels of power to automotive actuators (e.g. the
brightness of an LED, the position of a valve, or the
speed of a motor) is using pulse width modulated
(PWM) signals. Instead of varying the amplitude
of a DC signal, PWM signals transmit pulses
that switch between 0 and a constant amplitude
at a fixed frequency. The width of the pulses de-
termines the amount of power transferred to the
Figure 6. PWM representation of an analog signal
V V
t t
actuator, as illustrated in Figure 6. The primary
advantage of this technique is that relatively little
power is dissipated in the switching device (which
is always full-on or full-off) and therefore it is
possible to build relatively efficient controllers.
However, PWM signals are a major source of
electromagnetic interference in automobiles. Con-
trolling an actuator with a DC amplitude setting
generates very little high-frequency energy. PWM
signals, on the other hand, generate significant
amounts of energy at the switching frequency and
at harmonics of the switching frequency. The fast
transition times required for efficient switching
of PWM signals can result in significant amounts
of energy being created at frequencies that are
100 (or more) times higher than the switching
frequency. High-frequency energy is much more
easily coupled from one circuit to another and
much more likely to interfere with radio receivers
and wireless communication.
Power Bs Noise
Even when the digital or PWM signals are care-
fully routed and isolated from other circuits, the
active devices that produce these signals must
draw current from the power supply. Switching
devices tend to draw power from the supply in
short bursts. These bursts of current and the
corresponding dips in supply voltage can be
propagated in an automobile’s power distribution
bus, causing the bus to become a major conduit
for electromagnetic interference. CISPR 25
specifies limits on the amount of noise that any
given component is allowed to put on the power
55
 Electromagnetic Compatibility Issues in Automotive Communications
bus. Nevertheless, anyone who has ever used an
oscilloscope to measure the voltage on the power
bus wiring in a modern automobile knows that
spikes, dips, bursts and periodic high-frequency
noise are common occurrences.
Engine Noise
The electronic ignition circuitry in internal com-
bustion engines generates sharp pulses that can
be detected everywhere within and around the
vehicle and can be especially strong in the engine
compartment. Similarly, vehicles with electric
drives employ motor controllers that generate
very fast, high-current PWM signals. This type
of noise can be difficult to deal with because the
total amount of energy and the frequency content
of these signals depend on many factors including
the engine speed and load.
External Noise Sources
The list of external noise sources that a vehicle
might encounter is virtually endless, because cars
and trucks are used for many different purposes
and are driven in a wide variety of environments.
External sources known to have caused automo-
tive EMC problems include:
• Radio and TV towers
• Overhead power transmission lines
• Electrostatic discharge generated by the
occupants
• Electrostatic discharge generated by the
vehicle-road interface
• Amateur radio equipment installed in the
vehicle
• CB or fleet communications radios installed
in the vehicle
• Police radar
• Nearby lightning strikes
• Cell phones in or near the car
56
The EMC standards described in the first sec-
tion of this chapter are designed to emulate as many
types of external sources as possible. However,
the electromagnetic environment is constantly
changing and it is difficult to anticipate every
situation. Also, automotive manufacturers must
balance the cost of designing for every possible
situation against the impact of specific types of
EMC-related failures in their market.
EM COUPLING Path
Before discussing the relative advantages and dis-
advantages of one automotive design over another,
it is important to have a basic understanding of
how electromagnetic energy is coupled to or from
electronic systems. Electromagnetic coupling can
be divided into 4 categories:
• Conducted coupling
• Electric-field coupling
• Magnetic-field coupling
• Radiated field coupling
The type of coupling most likely to occur
depends on the frequency of the source, the
impedances of the source and victim, and the
proximity of the source and victim. Once a circuit
is understood well enough to know which types
of EM coupling are likely to be a problem, it is
much easier to make design decisions that will
minimize the chances of having an electromag-
netic interference problem.
Conducted Coupling
Conducted coupling (also called common-imped-
ance coupling) can only occur when there are at
least two conductors connecting the source to
the victim. Examples of conducted coupling in
an automobile include:
Electromagnetic Compatibility Issues in Automotive Communications
• Lights dimming or systems resetting when
the starter is engaged
• Power bus noise due to a motor controller
being heard as static in the radio
• Damage to an ECU occurring as a result of
arc welding on the vehicle
Figure 7 shows an example of how conducted
coupling occurs. In this case, two systems in the
automobile share the same current return conduc-
tor (the body of the vehicle). When the starter is
engaged, it draws 20 amps of current from the
battery causing the battery voltage at the terminals
to drop to 6 volts. Other systems powered by the
same battery must be able to function with an
input of only 6 volts for a second or two while the
engine is cranking. Any problems resulting from
this momentary drop in power supply voltage are
EMC problems.
In the starter motor example, the two conduc-
tors are the power and ground conductors. Con-
ducted coupling occurs because both the source
and the victim circuits share a conducted circuit
Figure 7. Conducted coupling example
OTHER
SYSTEMS
BATTERY
B ATTE R Y
0 .3 O H M S
12 V +
-
R > > 0 .3 O H M S
O TH E R
S Y S TE M S
path with a finite impedance (in this case, the
battery resistance). Conducted coupling can also
occur between two systems that both use body
surface metal as their “ground” conductor. Cur-
rents from one system flowing in the body surface
metal induce voltages across the metal that can
affect the operation of other circuits that use the
body surface metal as a “zero-volt” reference.
Conducted coupling tends to dominate at low
frequencies and is the only type of electromag-
netic coupling that can occur at DC. Conducted
interference problems are usually solved by
reducing the current in the “culprit” circuit or
by separating (isolating) the current paths in the
culprit and victim circuits.
Electric-Field Coupling
Electric-field coupling (also called capacitive
coupling) is most likely to occur between high-
impedance circuits. An example of electric-field
coupling in an automobile is shown in Figure 8.
A few electric field lines from a high-voltage igni-
OTHER
SYSTEMS
STARTER
MOTOR
IG N ITIO N
S W ITC H
S TA R TE R
M O TO R
0 .3 O H M S
57
 Electromagnetic Compatibility Issues in Automotive Communications
Figure 8. Electric-field coupling example
CM
V (t)
IG N ITIO N E LE C TR IC -FIE LD
C IR C U ITR Y C O U P LIN G
tion circuit terminate on the AM radio antenna.
Schematically, this can be represented by a mutual
capacitance as illustrated in the lower part of
Figure 8. The current induced in the AM radio
receiver is proportional to C times dV/dt, where
C is the mutual capacitance and dV/dt is the time
rate of change of the excitation voltage.
Electric-field coupling problems are gener-
ally solved by reducing the mutual capacitance
between the culprit and the victim. This can be
accomplished by increasing the distance between
the two circuits or by shielding. It is important to
note that electric-field shielding is different from
magnetic-field shielding or enclosure shielding.
The goal of an electric field shield is to terminate
or redirect electric field lines before they reach
the victim circuit. Figure 9 shows an example of
how a shield might be used to eliminate the AM
radio interference illustrated in Figure 8. A metal
shield (in this case a metal hood) is placed in a
position where it can intercept the electric field
58
ANTENNA
RADIO
IGNITION
V N O IS E
R A D IO
R E C E IV E R
lines emanating from the culprit circuit. The shield
must be grounded to the culprit circuit’s refer-
ence conductor in order to effectively terminate
the field lines.
We could have protected the AM radio from
this culprit and others by surrounding the antenna
with a metal shield and grounding it to the AM
radio’s reference conductor. However, this would
have affected the antenna’s ability to perform its
intended function. Whether it is better to shield the
culprit or victim depends on many factors includ-
ing cost and convenience, but the shield should
always be grounded to the reference conductor
that most effectively diverts the field lines.
Magnetic-Field Coupling
Magnetic-field coupling (also called inductive
coupling) is most likely to occur when the culprit
is a low-impedance (high current) circuit. Figure
Electromagnetic Compatibility Issues in Automotive Communications

Figure 9. Electric-field shielding example

ANTENNA

RADIO

IGNITION

C CM
HO O D

V (t) V N O IS E
HOOD

IG N ITIO N E LE C TR IC -FIE LD R A D IO
C IR C U ITR Y C O U P LIN G R E C E IV E R

Figure 10. Magnetic-field coupling example 10 shows an example of magnetic-field coupling

between a motor driver and an audio circuit on
a printed circuit board. Lines of magnetic flux
passing through the circuit loop induce a voltage
in the loop that appears across the input of the
audio amplifier. A schematic representation of this
coupling is shown in the lower part of Figure 10.
In this case, the voltage induced across the ampli-
fier input is proportional to M dI/dt, where M is
the mutual inductance between the culprit circuit
and the victim circuit and dI/dt is the time rate of
change of the current in the culprit circuit.
The mutual inductance can be reduced by
increasing the distance between the culprit and
I(t) V N O IS E victim or by diverting the magnetic flux lines with
a shield. Unlike electric field lines, magnetic flux
lines can’t be terminated on a conductor (or any
M O TO R LO O P A R E A
CURRENT O F C IR C U IT other material). Since they cannot be terminated,
M A G N E TIC -FIE LD
they must be diverted. One of the best ways to
C O U P LIN G divert low-frequency magnetic flux lines is to use
steel (or a similar material with high magnetic
permeability). Lines of magnetic flux prefer to

59
 Electromagnetic Compatibility Issues in Automotive Communications
Figure 11. Low-frequency magnetic-field shielding
with magnetic material
Figure 12. High-frequency diversion of magnetic
field lines due to eddy currents
travel through steel rather than air when given
the opportunity. As indicated in Figure 11, shields
made of steel can effectively divert magnetic flux
lines from a victim circuit when properly oriented.
Note that simply putting a steel plate above or
below the circuit would be relatively ineffective,
because the magnetic flux lines would simply
pass through the plate to the other side.
At higher frequencies (e.g. 100s of kHz to
MHz), magnetic flux lines are reluctant to flow
through large conductors due to “eddy currents”
60
that are induced in these conductors. The eddy
currents produce their own magnetic flux that op-
poses the flux creating the eddy currents. The net
effect is that higher frequency lines of magnetic
flux will flow around plates made of materials that
are good conductors (e.g. aluminum or copper).
Consider, for example, a magnetic interference
problem between a PWM dome light circuit
switching at 200 kHz and a nearby circuit board
as illustrated in Figure 12. At 200 kHz, eddy cur-
rents induced in the copper plane of the circuit
board are sufficient to divert the magnetic flux
lines and prevent an interference problem.
Radiated Field Coupling
Many people confuse radiated field coupling with
electric or magnetic field coupling. In fact, the
automotive EMC standards divide all tests into two
categories, conducted and radiated, without any
attempt to describe the specific type of coupling
being measured. However, in order to properly
design and troubleshoot for EMC, it is essential
to distinguish electric and magnetic field coupling
from truly radiated coupling.
Radiated field coupling can only occur when
the victim circuit is not in the near field of the
culprit circuit. This generally means the victim
circuit must be approximately a wavelength
away (or farther) from the culprit. At 100 MHz,
a wavelength in air is 3 meters. This means that
radiated field coupling between circuits in an
automobile is not likely to occur at frequencies
below 100 MHz. Low frequency coupling will be
predominantly electric or magnetic field coupling
and all circuits and shields should be designed
to minimize the type of coupling most likely to
occur based on the impedance and environment
of the systems involved.
Radiated field coupling requires an “antenna”
of some kind on both the culprit and victim cir-
cuits. Relatively efficient antennas may take the
form of wires or other metallic objects. They can
be made of copper, aluminum, steel or virtually
Electromagnetic Compatibility Issues in Automotive Communications

any highly conductive material. They can be
bare metal or covered in plastic, but they all have
one thing in common; efficient antennas are not
small relative to the wavelengths they radiate. In
fact, an efficient antenna must drive one metallic
object with a voltage relative to another metallic
object and NEITHER object can be very small
relative to a wavelength. This means that good
antennas at frequencies below a few GHz will be
large enough to easily spot.
Radiated emissions from automotive compo-
nents are usually the result of common-mode cur-
rents induced on the wiring harnesses. Common-
mode currents can be unintentionally induced on
a wiring harness in two ways; differential signals
on a harness can be transformed into common-
mode currents by imbalances (asymmetries) in the
harness structure or the signal source; or, electric
or magnetic fields generated by the source can
produce a voltage between the harness and the
component that drives common-mode current
on the cable. At frequencies above 30 MHz, a
few microamps of common-mode current can be
sufficient to cause radiated emissions that exceed
some radiated emissions limits.
The best way to minimize the chances of pro-
ducing excessive radiated emissions is to make
sure all of the large pieces of metal in your design
are held to the same potential. This generally
means all wires, shields, heatsinks and enclo-
sures should be bonded together, and all power
and signal wires should have a high-frequency,
low-impedance connection to these metal parts.
The more high-frequency circuitry that is located
between two metal objects, the more well-bonded
or filtered those objects have to be. It is also a good
idea to limit the frequency content of your circuits.
Higher frequencies have shorter wavelengths and
are therefore efficiently radiated by more of metal
objects in the vehicle.
Shielding to prevent radiated emissions or
susceptibility can be difficult. It is generally
necessary to completely surround the culprit or
victim circuits with an enclosure that has no large
apertures or seams. All wire penetrations through
the enclosure must be carefully filtered. This is
a common practice in military vehicles, but not
in automobiles. Partial shields or shields with
significant apertures, seams or unfiltered wire
penetrations can actually increase the radiated
emissions from a circuit board by providing the
large metal objects needed to create an efficient
antenna.
Table 3 lists a few design guidelines that can
help to reduce the chances of having a radiated
emissions problem. However, successfully design-

Table 3. EMC design guidelines

Design Guideline #1: Mini-
mize the loop areas associated
with high-frequency power
and signal currents.
Design Guideline #2: Do not
split, gap or cut the signal
return path
Design Guideline #3: Don’t
Locate high-speed circuitry
between connectors on a
printed circuit board
Design Guideline #4: Control
signal transition times
Digital circuit designers like to think of signals in terms of voltage. Signal integrity and EMC engineers
must think of signals in terms of current. Signal currents always flow in loops. Keep these loops as
small as possible.
The safest rule-of-thumb is to provide one solid plane for returning all signal currents. In situations
where it is expected that a particular signal is susceptible or is capable of interfering with the other cir-
cuits use a dedicated trace or wire to return that current to its source. Do not try to implement a scheme
that “worked” in someone else’s “similar” design.
High-speed circuitry located between connectors on a printed circuit board can easily develop potential
differences of a few millivolts or greater between the connectors. These voltages can drive currents
onto attached cables causing a product to exceed radiated emissions requirements.
A well-formed digital signal will have a significant amount of power in the lower harmonic frequen-
cies, but not so much power in the upper harmonics. Power in the upper harmonic frequencies is best
controlled by controlling the transition times in digital signals.

61
 Electromagnetic Compatibility Issues in Automotive Communications
ing and integrating the myriad electronic systems
in modern automobiles cannot be accomplished
by simply following a few design guidelines. It
is important to hold regular design reviews with
engineers who specialize in automotive EMC.
EMC ISSUES Wwith AUTutomoti
NETtwork
The dozens of microprocessors in the typical
automobile must be able to exchange electronic
information with various sensors, actuators and
other microprocessors. This information can
be conveyed by an analog signal level, a PWM
signal, or digitally using one of the many data
communications network protocols developed for
automotive applications. There are many factors
to consider when choosing an appropriate network
for a given communications task including the
amount of data, the importance of the data, and
the cost of hardware necessary to implement the
network. Electromagnetic compatibility is also
a significant factor in this decision. This section
discusses design aspects of each of the major
automotive networks that have an impact on
electromagnetic compatibility.
Controller Area Network (CAN)
The CAN bus, developed in the mid-1980s, can be
found in virtually every automobile manufactured
today. It is also used in non-automotive applica-
tions such as factory automation and aerospace
systems. The need to send and receive digital sig-
nals reliably in an electrically noisy environment
was a significant factor driving the design of the
CAN bus. CAN bus signal drivers and receivers
employ a balanced, differential interface. The
signals are represented as voltage differences
between two signal carrying wires rather than as
voltages relative to a reference ground. Both signal
carrying conductors have the same impedance
relative to ground and the bus is terminated with a
62
matched resistance at both ends. As a result, CAN
signals are relatively unaffected by noise currents
flowing in the ground structure. They are also less
likely to produce significant radiated emissions,
since the currents flowing in each signal wire are
equal in amplitude and opposite in sign.
CAN communications employ Cyclic Re-
dundancy Checks (CRCs) and all messages are
acknowledged by the recipient. In the event that
transient electrical noise does couple to the circuit
and cause a data error, these features reduce the
likelihood of a communication failure.
Figure 13 shows a basic CAN bus layout. EMC
problems due to conducted coupling are rare in
properly implemented CAN bus systems because
the bus is isolated from the ground structure. How-
ever individual nodes can be disabled by conducted
interference, and improperly designed nodes can
introduce high-frequency noise on the bus that is
ultimately radiated by the bus or by other nodes
on the bus. CAN bus implementations should be
tested to ensure that each side of the bus wiring
has the same high impedance to ground when all
terminations and nodes are connected.
Local Interconnect Network (LIN)
The LIN bus was developed in the late 1990s as
a lower-cost, lower-weight alternative to CAN. It
employs a single-ended (as opposed to differential)
signaling scheme, so it only requires one signal
wire, as illustrated in Figure 14. Signal voltages
are referenced to the ground structure (usually the
vehicle frame) and signal currents return to the
driver on this structure. LIN signals can also be
transmitted over existing wires, such as a vehicle’s
power distribution wiring.
Since LIN bus data rates are less than 20
Kbps, radiated emissions are not a significant
problem. There is simply not enough power in
LIN bus signals at frequencies high enough to
radiate efficiently. However, since the LIN bus
uses the vehicle ground as the return conductor
for its signal current, it is more sensitive to noise
Electromagnetic Compatibility Issues in Automotive Communications
Figure 13. Basic CAN bus layout
NODE 1
NODE 2
Figure 14. LIN bus layout
NODE 1
currents flowing on the body of the vehicle. To
partially compensate for this vulnerability, LIN
buses typically employ a 0-12 volt signaling am-
plitude. Stronger signal amplitudes mean higher
levels of noise are required to cause an interference
problem. Also, the bus termination impedance is
high (nominally 1 kΩ), which reduces the levels of
noise current that flow in the signal conductor.
LIN bus communication takes place one mes-
sage at a time between a master and slave node.
Although it is not required by the LIN specifica-
tion, transceivers typically have built-in ESD and
overvoltage protection. Many also have risetime
control and ground-shift protection.
NODE 3
NODE 4
NODE 2
FlexRay
FlexRay was developed in the late 1990s to meet
the demand for greater bandwidth and guaranteed
response times. Like CAN, it employs a differ-
ential signaling scheme that can be implemented
on twisted wire pairs with matched terminations.
Unlike CAN, it employs a 3-state bus where signal
levels can assume 3 different voltage (non-ground)
levels. In its most robust configuration, all com-
munication is point-to-point with connections to
the rest of the network made through bidirectional
repeaters called “active stars”. Passive star and
63
 Electromagnetic Compatibility Issues in Automotive Communications
linear passive bus configurations are also possible.
The maximum data rate is 10 Mbps.
The FlexRay standard includes a document
called the “EMC Measurement Specification”
(FlexRay Consortium, 2005). This document does
not specify test limits, but it does define EMC
test procedures, test configurations, excitation
waveforms, and failure criteria for evaluating
the electromagnetic susceptibility of FlexRay
networks. The FlexRay standard also specifies
ESD protection levels for network processors.
MOST
Media Oriented Systems Transport (MOST) is a
high-speed networking scheme designed to meet
the needs of audio and video entertainment sys-
tems in automobiles. Data is conveyed over optical
fiber at rates as high as 23 Mbps. Optical fiber
is immune to electromagnetic fields, but MOST
networks are still susceptible to electromagnetic
interference. The most vulnerable part of a fiber
optic system is often the optical signal receiver.
The high-gain, wide-band amplifiers required to
convert flashes of light back to electrical signals
can be relatively sensitive to conducted or elec-
tric-field coupled noise. These amplifiers must be
carefully laid out, shielded and filtered to prevent
system noise from obscuring the signal. The rest
of the circuitry must also be carefully designed to
avoid the electromagnetic interference and signal
integrity problems that are commonly associated
with circuits operating at these frequencies.
Bluetooth and other Wireless
Communications
Bluetooth is a wireless network commonly de-
ployed in automobiles. It is a relatively low-power,
short range wireless network that operates in the
2.40 – 2.48 GHz Industrial-Scientific-Medical
(ISM) band. This is the same unregulated fre-
quency band occupied by WiFi networks, micro-
wave ovens, and a variety of other ISM devices. To
64
minimize interference, Bluetooth networks divide
the band into 79 1-MHz channels and Bluetooth
communication hops between channels. In this
way narrowband interference that blocks only a
few channels does not have a significant overall
impact on communications. Of course, broadband
or even out-of-band noise can disrupt Bluetooth
networks if the power levels are sufficient.
Automobiles must also co-exist with other
wireless transmitters and receivers such as cell
phones, WiMax devices, satellite communica-
tions devices and dedicated short range com-
munications (DSRC) equipment. Unfortunately,
current international EMC test standards do not
adequately address potential interference prob-
lems with these devices. The technology has
advanced faster than the standards organizations
can react. Some manufacturers have developed
their own tests and others test the vehicle by
simply installing representative equipment and
driving around.
Fortunately, the transmitted power levels from
most of these devices are relatively low. Also,
the electromagnetic immunity is enhanced by
the fact that the newer wireless communication
technologies tend to be either spread spectrum or
very narrow-band. Spread spectrum communica-
tion networks (like Bluetooth) shift in both time
and frequency making them relatively immune
to narrow-band interference, while narrow-band
communication is relatively insensitive to inter-
ference occurring at frequencies other than the
operating frequency.
Nevertheless, as the number of interference
sources in an automobile rises, the signal-to-noise
ratio (and the range) of spread spectrum devices
degrades and the probability of interfering with
a narrow band transmission increases. New test
procedures and design practices will be required
to insure that future automobiles are electromag-
netically compatible with an increasingly wide
variety of wireless devices.
Electromagnetic Compatibility Issues in Automotive Communications
EC DESIign CONSIidration
When automotive components and systems are
carefully laid out and properly grounded and
shielded, the automobile and its components can
usually meet all EMC requirements with very little
cost invested in EMC-specific design features.
EMC problems with poorly designed components
and systems, on the other hand, can be very
difficult to fix without significantly impacting
costs or development schedules. Therefore, it is
critically important for automotive manufactur-
ers and Tier 1 electronics suppliers to design for
EMC early in the product development cycle and
to have component and system design reviews on
a regular basis.
Wiring Harnesses
One aspect of automotive system design that de-
serves special attention is the design and layout of
the wiring harnesses. Coupling from one circuit to
another in a wiring harness is often referred to as
crosstalk. It is important to pay careful attention
to the currents that will flow out and back on each
of the wires in a wiring harness. High-speed sig-
nals (with data rates greater than 1 Mbps) should
always be routed on twisted wire pairs, shielded
Figure 15. An example of a circuit board with DC-isolated circuit and chassis grounds
C h a ssis co n n e ctio n
to ch a ssis g ro u n d
C a p a cito rs co n n e ctin g
ch a ssis g ro u n d to th e
d ig ita l re tu rn p la n e
twisted wire pairs or coaxial lines. Twisted wire
pairs should be used for differential signals, and
coaxial lines (with the shield grounded at both
ends) should be used for single-ended signals.
Low-frequency analog signals can be routed on
a single wire using the body as a current return
path provided these signals are able to tolerate the
noise currents typically present on body surface
metal. Low-frequency analog signals that are
sensitive to conducted noise should use a wire for
the return current that is isolated from ground at
one or both ends.
Grounding and Shielding of
Components
Automotive components often have two separate
reference grounds; one labeled “chassis” ground
that connects directly to the component enclosure
and automobile body surface metal, and another
labeled “digital” ground that serves as the cur-
rent return path for all of the digital circuits on
the circuit board. Since both of these ground
structures are rather large metallic objects (i.e.
good antenna parts), it is important that they are
held to the same potential at high frequencies.
Even a few microvolts of noise between these
two grounds at high frequencies can create a
W irin g H a rn e ss
C h a ssis co n n e ctio n
to ch a ssis g ro u n d
D ig ita l R e tu rn P la n e
C h a ssis G ro u n d P la n e
65
 Electromagnetic Compatibility Issues in Automotive Communications
significant radiated emissions problem. Bonding
them together directly in as many places as pos-
sible is the best way to accomplish this, however,
sometimes it is necessary to isolate the chassis and
digital grounds at low frequencies in order to keep
noise currents that flow on the vehicle body from
flowing across the circuit board ground. In other
words, we want to isolate these grounds to prevent
conducted (low-frequency) interference, but we
want to bond them together at high frequencies
to prevent a radiated emissions problem. This
can usually be accomplished by routing different
grounds on different layers of a multi-layer board
and connecting them with many low-inductance
capacitors. Figure 15 illustrates the grounding
scheme in a typical multi-layer circuit board in
an automotive application where chassis ground
and digital ground must be DC-isolated.
When there is a chassis ground plane on a
printed circuit board, it is very important to ensure
that this plane has a low inductance connection to
both the chassis and to the digital ground plane,
especially in the vicinity of external cable con-
nections. Even a small, high-frequency voltage
between the chassis and a wiring harness can result
in excessive levels of radiated emissions.
As described earlier, any shields that are ap-
plied to the product should be appropriate for the
field coupling they are intended to suppress. Low-
frequency electric-field shields should intercept
electric flux lines and route the displacement
current harmlessly back to its source. Low-fre-
quency magnetic-field shields should divert lines
of magnetic flux away from the circuit being pro-
tected. High-frequency shielding enclosures must
be carefully designed to eliminate leakage through
seams and apertures, and all wire penetrations
through the shield should be filtered.
Antennas
High-end cars on the market today typically have
about a dozen intentional antennas in them, a num-
ber that continues to rise each year. These antennas
66
are used for remote keyless entry; AM, FM and
satellite radio reception; tire pressure monitoring;
GPS Navigation; cell phone communication and
wireless data communication. Fortunately, most
of these systems employ antennas and receivers
that are relatively insensitive to out-of-band in-
terference. However, the performance of each of
these systems can be severely degraded by in-band
interference, and the high-gain amplifiers in the
front end of these systems can be overwhelmed
by particularly strong transient or out-of-band
continuous wave signals.
Shielding is not usually a practical, cost-effec-
tive solution for protecting antennas and receivers.
Thus it is important to be familiar with the relative
sensitivities of a particular communication system
and to locate potential sources of interference far
enough from the antenna to minimize the potential
for an interference problem.
EMC Design Reviews
Many EMC problems that are first observed in
vehicle prototypes are due to relatively obvious
design flaws that could have been detected and
easily corrected in the early stages of the vehicle
design. Correcting problems after the first proto-
types are built, however, can be relatively difficult
and expensive. For this reason, it is very important
to regularly review the design of the automobile
and all electronic components on a regular basis
throughout the product development process.
Experienced EMC engineers can identify
potential problems and recommend fixes that
often add nothing to the product cost. No-cost
EMC fixes include measures such as rerouting
traces on a printed circuit board, changes in
grounding, changes in cable routing, or changes
in connector pin assignments. An EMC engineer
may also recommend adding fixtures for filters
or shields that are initially unpopulated and only
used in the event that a problem arises later in
the development cycle. This type of preparation
costs nothing initially and can save substantial
Electromagnetic Compatibility Issues in Automotive Communications

amounts of time and money if EMC problems
arise during prototype testing.
ConcluSIion
For the first 60 to 70 years of its existence, the
automobile was primarily a mechanical machine
with few electric components. Automobiles today
have dozens of microprocessors and kilometers
of electric wire occupying a volume of a few cu-
bic meters. As the number of electronic systems
in automobiles rises, the potential for electro-
magnetic interference increases. Designing for
electromagnetic compatibility is not difficult or
expensive, but it is important to devote proper at-
tention to EMC at every stage of an automobile’s
development. Problems discovered late in the
design cycle can seriously impact development
schedules and product cost.
This chapter has provided a basic overview of
EMC issues affecting automotive communications
for non-EMC engineers and engineering managers
who work with automotive networks. Additional
information on this topic can be found in (Paul,
2006), (Kaiser, 2005) or the Clemson Vehicular
Electronics Laboratory website (http://www.cvel.
clemson.edu).
REFERENCES
Crawford, M., & Workman, J. (1980). Predicting
free-space radiated emissions from electronic
equipment using TEM cell and open-field site
measurements. Proc. IEEE Int. Symp. on Electro-
magn. Compat., (pp. 80-85). Baltimore, MD.
Deng, S., Hubing, T., & Beetner, D. (2007).
Characterizing the electric field coupling from IC
heatsink structures to external cables using TEM
cell measurements. IEEE Trans. on Electromagn.
Compat., 49(4) 785-791.
FlexRay Consortium. (2005). FlexRay Communi-
cations System, Physical Layer, EMC Measure-
ment Specification Version 2.1.
International Electrotechnical Commission – IEC.
(2002). CISPR 25—Radio disturbance charac-
teristics for the protection of receivers used on
board vehicles, boats, and on devices—Limits
and methods of measurement.
IEC. (2008). CISPR 12—Vehicles, boats and in-
ternal combustion engines—Radio disturbance
characteristics—Limits and methods of measure-
ment for the protection of off-board receivers.
International Organization for Standardization—
ISO 11452-2. (2005). Road vehicles — Vehicle
test methods for electrical disturbances from
narrowband radiated electromagnetic energy
— Part 2: Off-vehicle radiation sources.
ISO 10605. (2008). Road vehicles — Test methods
for electrical disturbances from electrostatic
discharge.
ISO 11452-3. (2001). Road vehicles — Component
test methods for electrical disturbances from
narrowband radiated electromagnetic energy
— Part 3: Transverse electromagnetic mode
(TEM) cell.
ISO 11452-4. (200). Road vehicles — Component
test methods for electrical disturbances from
narrowband radiated electromagnetic energy
— Part 4: Bulk current injection (BCI).
ISO 7637-2. (2004). Road vehicles — Electri-
cal disturbances from conduction and coupling
— Part 2: Electrical transient conduction along
supply lines only.
ISO 7637-3. (2007). Road vehicles — Electrical
disturbance by conduction and coupling — Part
3: Vehicles with nominal 12 V or 24 V supply
voltage — Electrical transient transmission by
capacitive and inductive coupling via lines other
than supply lines.

67
 Electromagnetic Compatibility Issues in Automotive Communications

Kaiser, Kenneth L. (2005). Electromagnetic
Compatibility Handbook. CRC Press.
Paul, C. R. (2006). Introduction to Electromag-
netic Compatibility, 2nd. ed. Wiley-Interscience.
Sreenivasiah, I., Chang, D. C., & Ma, M. T.
(1981). Emission characteristics of electrically
small radiating sources from tests inside a TEM
cell. IEEE Trans. on Electromagn. Compat., (23),
113-121.
Tippet, J. C. & Chang, D. C. (1976). Radiation
characteristics of electrically small devices in a
TEM transmission cell, IEEE Trans. on Electro-
magn. Compat., 18(4), 134-140.

68
 69

Chapter IV
Automotive Network
Architecture for ECUs
Communications
Fabienne Nouvel
Laboratory IETR-UMR, INSA, France

Wilfried Gouret
Laboratory IETR-UMR, INSA, France

Patrice Mazério
Laboratory IETR-UMR, INSA, France

Ghais El Zein
Laboratory IETR-UMR, INSA, France

ABSTtract

This chapter deals with automotive networks and the emerging requirements involved by the X-by-wire
and X-tainment applications. The introduction of ECUs (Electronic Control Units) has been driven by
new market (like navigation, multimedia, and safety). Furthermore, the automotive industry has to face a
great challenge in its transition from mechanical engineering towards mechatronical products. Combining
the concepts of networks and mechatronic modules makes it possible to reduce both the cabling and the
number of connectors. To connect the ECUs, a variety of network technologies are already widespread.
A review of the most widely-used automotive networks and emerging ones is given first. To fulfill the
increasing demand of intra-car communications, a new technique based on power line communication
(PLC) is proposed and reviewed in the second section. On the other hand, there are several infotain-
ment applications (like mobile phones, laptop computers) pushing for the adoption of intra-car wireless
communications. Some of the most common wireless technologies that have potential to be used in the
automotive domain are considered and different experimentations are presented. Finally, the challenges
of these wired or wireless alternative solutions to automotive networks are highlighted.

Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Automotive Network Architecture for ECUs Communications
Introduction
Fieldbuses are a family of automotive communi-
cation networks that have evolved as a response
to the demand to reduce cabling costs in factory
automation systems. Moving from a situation in
which every controller has its own cables connect-
ing the sensors to the controller to ECU sharing
a bus, costs could be cut and flexibility could be
increased. Pushing for this evolution of technol-
ogy was the fact that the number of cables in the
system increased as the number of sensors and
actuators grew. Furthermore, the ECUs share their
CPU power and knowledge with other control-
lers. Several fieldbus technologies, usually very
specialised, were developed by different manu-
facturers to meet the demands of their application.
The standardisation process stabilized not until
the mid 90’s with CAN standardized in1993 by the
International Standardisation Organisation (ISO)
(Johansson, Törngren, & Nielsen, 2005).
Indeed, today’s vehicle networks are not just
collections of discrete, point-to-point signal cables.
They distribute the electronic systems according
to their domain sharing networks. Replacing
rigid mechanical components with dynamically
configurable and even reconfigurable electronic
elements triggers an almost organic, system wide
level of integration. Furthermore, the amount of
cables is drastically reduced by using these field-
buses. As a result, the cost of advanced systems
should plummet. Furthermore, highly reliable
and fault-tolerant electronic control systems, X-
by-wire systems do not depend on conventional
mechanical or hydraulic mechanisms. They make
vehicles lighter, cheaper, safer, and finally more
fuel-efficient.
In (Len & Hefferman, 2001), the authors
demonstrate the advantages of X-by-wire and
embedded networks. For example, Leen and Hef-
ferman write “X-by-wire steering systems under
70
development will replace the steering column
shaft with angle sensors and feedback motors.
This removal will improve driver safety in col-
lisions and allow new styling freedom. It will
also simplify production of left- and right-hand
models. It is natural to add advanced functions
to such electronic systems. For example, consider
systems that reduce steering-wheel feedback to
the driver. In mechanical steering systems, the
driver actually feels the vehicle losing control in
unstable conditions and can react appropriately.
These self diagnosing and configurable systems
adapt easily to different vehicle platforms and
make the diagnosis easier” (Len & Hefferman,
2001, p.90).
Figure 1 represents the future topology of
embedded networks inside luxury cars. We can
notice the different multiplexed networks which
may need gateways to communicate between
different application domains.
Considering these large number of distributed
ECUs, it makes the automotive system complex in
many ways. Several different networks are used
to address the different levels of communications
requirements, including safety-critical, and fault-
tolerant. High bandwidth becomes necessary to
interconnect these systems with predictability.
This last issue is a great challenge for car manu-
facturers. To reduce this complexity, without
reducing safety, it is desirable to commit to a
limited set of networks. However, it is not in the
near future that the number of networks can be
reduced to only one or two as such technology
would provide the properties supporting the most
demanding automotive systems. It would probably
make it too expensive. Hence, it is more likely that
a few network technologies will be used while
providing timeliness and fault tolerance.
The next section gives an overview of networks
used today in the automotives, identifying their
strengths and possible drawbacks.
Automotive Network Architecture for ECUs Communications
Figure 1. Network vehicle architecture (source: Automotive Systems Research Group, 2008)
EVOolution of automoti
nETtworking t
The purpose of this section is to give a description
of the most representative networks for differ-
ent main domain of application. We will focus
on the three most common vehicle networks,
namely LIN, CAN and Flexray. These embedded
networks have both increased the functionality
and decreased the amount of wires. However, the
usage of different wires for the different networks
still has the disadvantage of heavy, complex and
expensive.
The Local Interconnect Network: LIN
Conceived in 1998, the LIN consortium comprises
many car manufacturers like Audi, Volvo, BMW
(“LIN Consortium”, 2003). LIN is an inexpensive
slow and serial bus used for distributed body
control electronic systems in vehicle. It enables
effective communication for sensors and actua-
tors where bandwidth, speed and versality are not
required (i.e inside mechatronic based subsystems
generally made of an ECU and its set of sensors
and actuators). LIN is commonly used as a sub
bus for CAN and Flexray.
A LIN network is based on one master node
and a couple of LIN slaves (up to 16 over 40 meters
line length). The communication in an active LIN
network is always initiated by a master task: the
master sends out a message header that comprises
the synchronisation break and message identifier
(frame header). Exactly one slave is activated upon
reception and filtering of the identifier (up to 64),
which starts the transmission of the response, up
to eight data bytes. The frame response may be
sent by one LIN slave or LIN master. Each LIN
frame is composed of a frame header and a frame
response. Figure 2 depicts the LIN frame format
transmitted during its frame slot.
The master node decides when and which
frame shall be transmitted according to the sched-
ule table containing the transmission order. At the
71
 Automotive Network Architecture for ECUs Communications
Figure 2. the LIN frame (source: Navet, Song, Simonot-Lion & Wilvert, 2005).
moment a frame is scheduled for transmission,
the master sends the header inviting a slave node
to send its data in response. Any node interested
can read a data frame transmitted on the bus. The
reliability of LIN is high but it does not have to
meet the same levels as CAN. The LIN can be
implemented using just a single wire, while CAN
needs two. The physical layer (PHY) supports a
data rate equal to 20 kb/s (due to electromagnetic
limitations) but other transmission supports en-
abling higher data rates are possible.
Another representative of low-cost network is
TTP/A (Kopetz & all, 2002). It implements mas-
ter-slaves exchanges where the slave answers the
master’s request with a data frame having a fixed
length. However, to the best of our knowledge,
TTP/A is not currently in use in production cars.
On the other hand, LIN is widely used in middle
range cars but it can not support high data rate as
required by devices like portable DVD players or
multimedia applications.
The Control Area Network: CAN
The Basic CAN
The Controller Area Network (CAN) (“CAN”,2001;
“Road Vehicles”, 2004) is a widely communica-
tion fieldbus used in automotive and other real
time applications. It is a serial communications
protocol which efficiently supports distributed
realtime control with a middle level of security.
In automotive electronics, engine control units,
sensors, anti-skid-systems, etc. are connected us-
72
ing CAN with bit rates up to 1 Mbit/s. However,
in today’s car, CAN is used as an SAE (Society
of Automotive Engineers) class C (classification
defined in (“J2056/2 Survey”, 1994)) network for
real time control in the powertrain and chassis
domains (at 250 or 500 kb/s). It is also implemented
as an SAE class B network for the electronics in
the body domain (at 125 kb/s).
In CAN, data is transmitted in frames contain-
ing between 0 and 8 bytes of data and a number
of control bits. A CAN frame is labelled by an
identifier whose numerical value determines the
frame priority. Depending on the CAN format
(standard CAN 2.0A or extended CAN 2.0B) the
size of the identifier is 11 bits (CAN 2.0A) or 29-
bits (CAN 2.0B). Between CAN frames sent on
the bus, there is also a 3 bit inter-frame space. The
standard CAN frame format (and the inter-frame
space) is depicted in Figure 3.
CAN is a collision-avoidance broadcast bus
(CSMA/CA for carrier sense multiple access with
collision avoidance), which uses deterministic
collision resolution to control access to the bus.
It implements a fixed-priority based arbitration
mechanism that can provide real time guarantees
and that is amenable to timing analysis.
As distributed real time systems become
more and more complex, the computing power
is steadily growing, and the number of ECUs
attached to CAN buses is growing. Thus CAN’s
maximum speed of 1 Mbps can lead to perfor-
mance bottlenecks. Hence, methods for increasing
the achievable utilisation are needed, e.g., novel
analysis methods that allow increased utilisation
Automotive Network Architecture for ECUs Communications
Figure 3. The CAN frame (source: Nolte, 2006)
while guaranteeing timing requirements to be ful-
filled, and novel approaches to schedule CAN.
Derivative Products and CAN
Optimization
In (Nolte, 2003; Nolte, 2006), the author ad-
dresses two issues to increase real time. First,
he investigates a novel way of scheduling the
CAN through a server-based scheduling. This
increases the flexibility of CAN, by providing
efficient handling of sporadic and aperiodic
message streams. He also studies the bit stuffing
which could degrade the real time and data rate.
The bit-stuffing mechanism of CAN causes the
message frame length to vary, depending on the
original bit-pattern of the CAN frame
Currently, time-triggered communication
systems are being considered for use in X-by-
Wire applications. In this category, multi-access
protocols based on time division multiple access
(TDMA) are particularly well suited. Time-Trig-
gered communication on CAN is specified by
TT-CAN (“Road Vehicles”, 2004), a standardised
session layer extension to original CAN. In TT-
CAN, the exchange of messages is controlled by
the temporal progression of time, and all nodes are
following a pre-defined static schedule. One node,
the master node, is periodically (or on the occur-
rence of a specific event) transmitting a specific
reference message, which acts as a reference in
time. All nodes in the system are synchronising
with this message, which gives a reference point
in the temporal domain for the static schedule of
the message transactions. The static schedule is
based on TDMA where message exchanges may
only occur during specific time slots or in time
windows. Hence, the master’s view of time is
referred to as the network’s global time. TT-CAN
appends a set of new features to the original CAN,
and being standardised, several semiconductor
vendors have started manufacturing TTCAN
compliant devices. Probably, carmakers might
consider the use of TTCAN for some systems
during a transition period as FlexRay technology
seems to be in a very strong position for becoming
the standard in the industry.
The Flexray Protocol
FlexRay (“Flexray Consortium”, 2004) is a
communication system developed by a consor-
tium founded in 2000, including both car and
semiconductors manufacturers. The consortium
members realized that despite the numerous
automotive communications protocols out on
the market, none would fulfil future automotive
control requirements.
FlexRay is a fault-tolerant protocol designed
for high-data-rate, advanced-control applications,
such as X-by-wire systems. The protocol specifica-
tion promises time-triggered communications, a
synchronized global time base, and real time data
transmission with bounded message latency.
An overview of the FlexRay frame format is
given in Figure 4. The frame consists of three
segments. These are the header segment, the
payload segment containing up to 254 bytes of
73
 Automotive Network Architecture for ECUs Communications

Figure 4. Flexray frame (source Flexray Consortium, 2004)

data, and the trailer or CRC segment. A node
shall transmit the frame on the network such that
the header segment appears first, followed by the
payload segment, and then followed by the trailer
segment, which is transmitted last. The header
of 5 fields includes the identifier, the cycle count
and the length of the data payload.
Communication is done in a communication
cycle consisting of a static part and a dynamic
part, where each of the parts may be empty. The
sending slots are represented through the identifier
(ID) numbers that are the same on both channels.
The sending slots are used deterministically (pre-
defined TDMA strategy) in the static part. In the
dynamic part there can be differences in the phase
on the two channels. Nodes that are connected to
both channels send their frames in the static part
simultaneously on both channels. It is possible to
send different data in the same sending slot on
different channels (Figure 5).
An interesting feature of Flexray is that it can
provide scalable dependability i.e., the “ability
to operate in configurations that provide various
degrees of fault tolerance.” Indeed, the protocol
allows for mixing links with single and dual trans-
mission supports on the same network, or with
different fault-tolerance capability with regards to
clock synchronization, etc. Proposed applications
include chassis control, X-by-wire implementa-
tions, and body and powertrain systems. Currently,

Figure 5. Communication cycle ( from (“Flexray Consortium”, 2004))

74
Automotive Network Architecture for ECUs Communications
FlexRay can handle communications at 10 Mbps.
In the automotive context, where critical and
non-critical functions will increasingly coexist
and interoperate, this flexibility can prove to be
efficient in terms of cost and reuse of existing
components.
Flexray is considered by manufacturers as
the backbone network for the other networks
like CAN or LIN. It is supposed to be foremost
used in safety-critical X-by-wire applications.
However, this network still uses specific wires,
which do not achieve compatibility with other
networks. So, gateways are necessary to transfer
information from ECUs connected on a domain
network to ECUs connected on other networks.
Those gateways could introduce latency, errors,
bottlenecks, and so on.
Other fault-tolerant networks have been de-
veloped, namely TTP and TT-CAN, but it seems
they are not the best choice for automotive manu-
facturers due to limited flexibility, high costs and
conflicting interests (Nolte, 2006).
Some Other Automotive Networks
Two other major networks have been developed
for emerging multimedia applications, namely
MOST (“MOST”, 2006; Nolte, 2006), and IDB
1394 (Navet, Song, Simonot-Lion & Wilvert, 2005;
“IDB Forum”, 2008). MOST, initiated in 1998, is
expected to be available with speeds of 150 Mbps.
It is using plastic optical fibers as communication
medium (and coax cables) and supports up to 64
devices (nodes). It is sharply more expensive than
other networks. The reader can have more details
at (“MOST Cooperation”, 2008).
On the other hand, IDB Customer Convenience
Port 1394 (IDB-1394), an automotive version of
IEEE 1394, supports data rate of 100 Mbps and
supports up to 63 nodes over a specific six-wire
cable. The IEEE 1394 interface is a serial bus in-
terface standard, for high-speed communications
and isochronous real time data transfer, frequently
used in digital audio and digital video. The inter-
face is also known by the brand names of FireWire.
In the automotives, IDB-1394 is a digital network
that runs at speeds of up to 400Mbps carrying
high quality video, multi-channel sound and high
speed data for vehicle applications. Designed for
entertainment and communication applications,
IDB-1394 enables devices to operate digitally car-
rying multiple programs of up to high definition
video along with multi-channel audio between
devices on a vehicle. In (“IDB1394”, 2008), an
IDB-1394 demonstrator has been presented and
focuses on the advantages of such solution.
These two systems answer to the actual au-
tomotive trends namely multi-media networks,
data storages, multiple users, navigation and
entertainment applications. However, these two
high speed solutions need specific wires and allow
only peer to peer communications. They answer
specifically to multi-media applications require-
ments. It doesn’t seem to be a so much flexible
seamless solution. Some common technologies are
discussed in (Navet, 2008) and (Nolte, 2006).
Conclusion
Table 1 sums the different embedded protocols
used in today vehicles.
Considering these specific domain embedded
networks, we can observe that each solution uses
its specific wires and communication system. We
can see the wide diversity of the solutions and the
necessity to find a limited set of networks which
answers to these multiple applications. An at-
tractive solution to reduce the wires is the power
line communication (PLC) using the power lines
(12/42V) to transmit both the power and the mes-
sages without functional barriers domain. It can
answer the vehicles requirements namely cost,
decrease of the amount of wires, flexibility and
bandwidth. First of all, we will describe the power
line channels and the available PLC techniques
in the next section.
75
 Automotive Network Architecture for ECUs Communications
Table 1. Embedded automotive networks
Networks LIN CAN
Data rate 20 Kbps 1Mbps
Medium Access Master/slave CSMA/CA
TDMA
Number of nodes 16 / 40 meters 32 / 40 meters
Security/fault low /no medium/some
tolerant
Applications Sub-networks Body domain,
power train
PowER lin communication
(PLC) ntwork: an altrnati
olution
Considering the current embedded vehicle net-
works, these fieldbuses require different wires
of several kilometres. It leads to increasingly
complex network architectures. Thus gateways
are necessary to switch from one network to
another one. This increases the propagation time
and does not guarantee real time. One solution to
reduce the amount of wires would be to use the
PLC technology that is currently being developed
for indoor networks to transmit information or
at least some of it, over the 12V power distribu-
tion system found in automobiles (Rubin, 2002).
This solution would reduce the amount of wires
as well as simplify the network architecture. The
possible applications of automotive PLC are very
wide, extending from low-speed data buses for
activating actuators to high-speed multimedia
applications. To our knowledge, there are no cars
actually that implement PLC network instead
of dedicated protocol buses. Different studies
have been carried out to demonstrate the PLC
feasibility, like the multi-user and low data rate
system in (Nouvel, El Zein & Citerne, 1994). It
is possible to interface PLC with CAN bus, as
described in (Beikirch & Voss, 2000). Other
low data PLC solutions have been studied in
76
Flexray MOST IDB-1394
10 Mbps 25 Mbps -> 150 100 Mbps
Mbps
Master/salve Peer to peer Peer to peer
TDMA
22/passive star 64 /fiber 63/fiber
high/yes some /no some /no
X-by-wires Mutli-media Mutli-media
(Schiffer, 2000) and (Huck, Schirmer & Dostert,
2005). Recently, different applications have been
tested like in (“Valéo”, 2006) which is a video
transmission between a camera at the back of the
vehicle and a screen placed on the dashboard up
to 30 frames/s.
Although high data rate and flexibility obtained
for indoor domestic PLC are proven, it is not pos-
sible to apply them directly to cars because the
geometrical characteristics and wires topologies
are totally different. Moreover, the PLC channels
are affected by the variable activation schedules
of electrical functions, such as brakes, indica-
tors or ABS, which produce sharp modifications
in the circuit’s load impedances over brief time
intervals, as presented in (Lienard, Carrion,
Degardin & Degauque, 2008). To optimize high
bit-rate communication, the PLC channel transfer
function must be carefully studied because it is
frequency selective.
The most promising candidate techniques for
PLC transmission are spread spectrum combined
with code division multiple access (CDMA)
and orthogonal frequency division multiplex-
ing (OFDM). The first one has been studied in
(Nouvel, El Zein & Citerne, 1994), in which the
PLC network offers about 50 kbits/s per user,
combined with multiplexing. On the other hand,
OFDM has been adopted for different wireless
wide area or personal area networks such as
Automotive Network Architecture for ECUs Communications

in IEEE 802.11a/g and IEEE 802.16a (Hara & is labelled “direct” if the shortest connection
Prasad, 1997; Heiskala & Terry, 2002). The main between the modems does not pass through the
advantages of this technique are well known and battery, a path is labelled “indirect” in all other
included robustness, high spectrum efficiency in cases. This distinction is based on the fact that
frequency-selective, flexible spectrum adaptation in the frequency range, the battery acts as a very
and time-variant fading channels and capability low impedance which partially short circuits the
of portable and mobile reception. Furthermore, transmission line. The schematic representation
OFDM allows high data rate and can be com- of the network used in the simulation and mea-
bined with CDMA. The choice of the modulation surements is given in Figure 6.
is driven by the PLC channel and the different Two paths, D1D3 and D2D3 are considered,
parameters are optimised according to the chan- corresponding to a direct and an indirect path,
nel behaviour: the bandwidth, the modulation respectively. In this test bed, the distances D1D3
(number of sub-carriers), the coding, the guard and D2D3 are about 8 meters and 10 meters
interval, and so on. respectively. Furthermore the number of wires
Below the embedded wiring harnesses are inside cable bundles varies from 1 to 30. The
presented in order to point out the variations on scattering matrix was taken on the 12V power
the PLC channels. Then, the OFDM HomePlug line while moving through urban and suburban
standard used in indoor PLC is presented and is traffic. Examples of transfer functions for direct
studied for the embedded vehicle PLC. and indirect connection in the frequency range 500
kHz-30 MHz are shown in Figure 7. The results
Transfer Function and Propagation
Channel
Figure 6. Testbed for power line measurement
Above all, it is very important to investigate both
channel capacity and noise environments on the
power line channel. As such a channel is not ideal,
it is characterised by noise disturbance, loss, in-
terference of impulsive type, and also attenuation.
Otherwise, the goal of the measurements on the
vehicle is to investigate the influence of the state
of the elements connected on the power line.
To determine the capacity of the channel
Figure 7. Power line channel measurements
through the cables for a PLC transmission, many
transfer functions between two nodes have been
measured in the study carried out by (Lienard,
Carrion, Degardin & Degauque, 2008; Degardin,
Laly, Liénard & Degauque, 2006). The authors
describe the aims of the tests and the measure-
ments set up. The most important results are
given here.
In order to analyse the battery effects, two
types of channels must be considered, depending
on the network architecture between a transmit-
ting point TX and a receiving point RX. A path

77
 Automotive Network Architecture for ECUs Communications
show the insertion losses over the [0-30] MHz
bandwidth are about -15 dB and -36 dB (influ-
ence of the battery) for direct and indirect path
respectively. The transmitted signal will be more
attenuated in indirect path than direct path.
To carry out the statistical study, measurements
were taken while the electronic equipments are
alternatively switched ON and OFF. Figure 8
represents the variations of the transfer function
between these two nodes in the vehicle accord-
ing to their state (ON/OFF) and/or other elements
present on the same cable. ‘Hi’ refers to different
configurations of electric equipments (ON/OFF).
One can see that the power line is very frequency
selective and the attenuation depends not only on
the distance between nodes but also on the states
of each node. The selectivity of the channel is
coupled with attenuation due to the distance be-
tween the nodes. Deep fading in many frequency
bands, especially in the PLC band [2 - 30 MHz]
used in residential area appears.
Liénard, Carrion, Degardin & Degauque
(2008) have then focus on the coherence band-
width and delay spread. The statistical analysis
shows that the coherence bandwidth is greater
than 800 kHz for direct path and 400 kHz for
Figure 8. a. Testbed setup, b. Channel variation according to loads
a.
78
indirect path that have to be compared with the
OFDM/HomePlug (“Home Plug V1.0”, 2008 )
subcarrier spacing of 195 kHz. The delay spread
remains under 380 ns for all the geometrical and
electrical configurations. This value is ten times
smaller than the cyclic prefix duration recom-
mended by the HomePlug standard.
In (Degardin, Laly, Liénard & Degauque,
2006), the authors have also considered the im-
pulsive noise. The analysis of the background
noise measured on the vehicle DC line during
cruising, shows that this noise can be modelled
by a white Gaussian noise with a Power Spectral
Density (PSD) of -130 dBm/Hz. Additional narrow
and wide band noise generated by the electronic
devices will also disturb the transmission. Figure
9 displays impulse noise measurements. A clas-
sification of the pulses into two classes has been
proposed: the first type is a single transient or
pulse, while the second one is a burst character-
ized by a succession of elementary pulses.
To overcome this noisy PLC environment,
the used of OFDM modulation technique can
be investigated. Using OFDM has in fact many
benefits as it can cope with deep fading, provides
good interchannel interference performance and is
b.
Automotive Network Architecture for ECUs Communications
Figure 9. Typical noise records in a vehicle
good at minimizing the effect of in-band narrow
interference. Furthermore, one other requirement
for PLC is the automotive EMC (electromag-
netic compatibility) constraints Ray_CISPR25
(“CISPR25”, 2002) which allow transmission in
the [0.15-30] MHz bandwidth under 50 dBµV
(-35dBm/Hz).
HomePLug V 1.0 Sandard
As OFDM (Hara & Prasad, 1997) is used in dif-
ferent communication systems we present in this
section, a brief overview of this modulation widely
used in both wired and wireless PHY layers is
given. OFDM is a frequency-division multiplexing
(FDM) scheme using the fast Fourier transform
(FFT) to generate the signal. A large number of
closely-spaced orthogonal sub-carriers are used
to carry symbols. The incoming stream is divided
into several parallel data streams or channels equal
to the number of sub-carriers. Each sub-carrier
is modulated with a conventional modulation
scheme (such as QPSK (quadrature phase-shift
keying)) at a low symbol rate, maintaining total
data rates similar to conventional single-carrier
modulation schemes in the same bandwidth.The
primary advantage of OFDM over single-carrier
schemes is its ability to cope with severe channel
conditions, for example attenuation of high fre-
quencies in a long wire, narrowband interference
and frequency-selective fading due to multipath,
without complex equalization filters. Channel
equalization is simplified because OFDM may
be viewed as using many orthogonal slowly-
modulated narrowband signals rather than one
rapidly-modulated wideband signal. To eliminate
intersymbol interference (ISI), guard interval is
inserted between symbols making it possible to
handle time-spreading. OFDM has been adopted
for different wireless wide area or personal area
networks (for example in IEEE 802.11a/g) or over
copper wires (for example xDSL). It is used in
new applications such as high definition video
and audio broadcasting.
In 2000, a coalition of manufacturers has
established a new protocol HomePlug 1.0 that
enables the establishment of an Ethernet-IP class
network over power line channels (“Home Plug
V1.0”, 2008). The HomePlug process is based on
an OFDM technique whose major advantage for
79
 Automotive Network Architecture for ECUs Communications

Table 2. Different transmission modes: coding and bit rate

Transmission Mode Modulation Coding Maximum bit rate

1 DBPSK ½ code rate Reed Solomon 4.53 Mb/s

interleaver (t=8)
2 DQPSK 23/29 to 238/254 9.06 Mb/s
and convolutional encoder
3 DQPSK (constraint length = 7) 13.59 Mb/s

Figure 10. HomePlug V1.0 PHY frame format (source: Home Plug V1.0, 2008)

Figure 11. HomePlug spectrum in the [1-25] MHz bandwidth and transmitted signal

128 sub car r ier s

synchr o
84 car r ier s for data

80
Automotive Network Architecture for ECUs Communications
the embedded PLC application is to cope with the
frequency selectivity of the power line channel
caused by the multiple reflections of the loads
connected to the power grid and by the coupling
to the other cables placed in the same bundle. The
modulation is based on 128 sub-carriers equally
spaced from 4.3 MHz to 25MHz, in conjunction
coding applied before differential encoding. The
differential encoding allows simplifying the Rx
part by suppressing the channel equalization. Only
84 carriers are used in the transmission bandwidth,
with DBPSK (differential binary PSK), DQPSK
(differential QPSK) or ROBO modulation ( a
DBPSK with heavy error correction) on each
subcarrier. HomePlug uses CSMA/CA protocol
to access to the network. Figure 10 represents the
PHY frame format as Figure 11 represents the
frequency response and the transmitted signal over
a 220 V AC network. All the coding parameters
are summarized in Table 2.
Recently, the HomePlug AV has been intro-
duced and will be the second major standard
released by the HomePlug Powerline Alliance
(Gavette, 2006; Afkhamie, Katar, Yonge, & New-
man, 2005). While HomePlug V1.0 was designed
mainly to distribute broadband internet access at
home, the HomePlug AV’s objective is to distribute
multi-media content within the house as well as
data. The PHY layer still operates in the frequency
range of [2 – 28] MHz and provides a 200 Mbps
PHY channel rate (150 Mbps net information rate).
Long OFDM symbols with 917 usable carriers
(tones) are used in conjunction with a flexible
guard interval. Modulation densities from BPSK
(binary PSK, 1 bit of information per carrier per
symbol) to 1024 QPSK (10 bits of information per
carrier per symbol) are independently applied to
each carrier based on the channel characteristics
between the transmitter and the receiver. Ex-
perimental systems of HomePlugAV have been
field tested in houses, suggesting that on average
HomePlug AV system achieves 10 times the data
rate of a 1.0 system.
Nowadays, the HomePlug Alliance and the
IEEE (“IEEE 1901”, 2008) are strongly commit-
ted to delivering a single mature solution that
will be endorsed by the IEEE 1901 work group
as the baseline standard. This alliance will pro-
pose a standard for high speed (over 100 Mbps
at the physical layer) communication devices via
alternating current electric power lines, so called
Broadband over Power Line (BPL) devices. It
will use transmission frequencies below 100
MHz. This solution could be investigated in an
automotive environment.
Some papers have compared HomePlug V1.0
and HomePlug AV with the indoor wireless system
(Lee, Latchman, Newman, Katar & Yong, 2002;
Lin, Latchman, Katar &. Lee, 2003). The results
show that PLC solution can outperform wireless
networks in indoor.
HomePLug V1.0 for Vehicle:
Experimental Results
In (Gouret, Nouvel, El Zein, 2006; Gouret, Nou-
vel, & El Zein, 2007) the authors experiment two
HomePlug V1.0 PLC modems inside the cars. The
first, designed by LEA uses the standard Home-
plug v1.0 and the Intellon chip (“LEA”, 2006), and
the second from Spidcom uses an owner standard
(“Spidcom”, 2008). In the Spidcom system, the
OFDM modulation is based on 896-carriers from
0 to 30 MHz divided into 7 equal sub-bands. For
the data transmission, BPSK up to 256-QPSK
modulation is used in conjunction with error cor-
rection (Viterbi & Reed Solomon) and a bit-loading
scheme. The MAC layer provides a mechanism
based on TDMA (CSMA/CA also available) with
a master/slave prioritisation.
For each modem, coupling and filters have
been optimized to fulfil the class 4 CISPR 25
with a typical limit of -35 dBm/Hz in the [0-30]
MHz bandwidth. The testbed, illustrated in Fig-
ure 12, is constituted of a pair of identical PLC
modems connected to PCs through Ethernet
81
 Automotive Network Architecture for ECUs Communications

Figure 12. PLC Testbed for PLC

LAN and placed in different point of the vehicle
(near trailer hitch, cockpit, engine and cigarette
lighter). Each one is connected to the electrical
12V supply network with the coupler and the
filter. A spectrum analyser and an oscilloscope
are connected to the DC line through a filter that
behaves as a high-pass filter with a 500 kHz cut-off
frequency. The first path corresponds as a direct
way between the engine and the trunk. For the
second, it corresponds also as a direct way, but
between ECUs of the cockpit and the cigarette
lighter. The throughput and transmitted power
were measured. Only a summary of findings is
presented here.
The first results (Figure 13a and Figure 13b)
show that the levels of broadcast power are differ-

Figure 13. Spectrum Measurement [0-30] MHz: a. HomePlug V1.0 b. Spidcom

a. b.

82
Automotive Network Architecture for ECUs Communications
ent for the two modems while keeping the power
under EMC/CISPR25 requirements. Concerning
the Spidcom one, its broadcast power is of the order
of 5 dB lower than LEA. To be able to compare
both modems, the bands 6 and 7 were switched off
to obtain the same used bandwidth. Both modems
transmit TCP or UDP data flow.
Iperf tool (“Iperf”, 2008) is used for measuring
maximum TCP and UDP bandwidth performance.
Iperf allows the tuning of various parameters and
UDP characteristics. Iperf reports bandwidth,
delay jitter, and datagram loss. During the data
rate measurement, the Spidcom modem transmit-
ted up to 9 Mbps against 5.67 Mbps for the LEA.
Spidcom modem integrates a pre-equalization
function which improves the data rate by tak-
ing into account the SNR (signal-to-noise ratio)
for modulation on each sub-carrier. When this
function is involved, we obtain a data rate of 12
Mbps. Such results can be obtained by using up
to 256-QPSK modulation on sub-carrier with high
signal-to-noise ratio, not available on Homeplug
chip which uses the same modulation for all the
sub carriers. The best data rate obtained with
the whole spectrum and optimization reaches
20 Mbps.
Discussion
The results are promising and demonstrate the
possibility of implementing a PLC inside vehicle
based on a modified OFDM/HomePlug V1.0 while
fulfilling vehicle requirements namely low cost,
high data rate similar to Flexray and flexibility. As
Spdicom modems can achieve 20Mbps using the
30 MHz bandwidth, it seems it can compete with
Flexray. However, only no critical functions have
been considered. Redundancy is only available
through multiple transmissions as error correction
and detection is applied in the receiver. Similar
PLC tests have been carried out in (Degardin,
Liénard, Degauque & Laly, 2007). However, to
improve data and bit error rate, additional work
is necessary leading to the following questions: 1)
What is about the throughput when more than two
PLC nodes are transmitting? 2) What is about the
real time performance of such implementation?
3) How does the PLC solution facilitate integra-
tion and diagnose? 4) How can it be interfaced
with other wired protocols? To answer the ques-
tions above, the research has to be conducted by
surveying the research community. We will now
focus on wireless solutions and their viability for
automotives.
WirELESS Ccommunication in
automo
During the last several years, the interest in wire-
less networking has grown significantly due to
the availability of many wireless products, such
as cell phones, wireless enabled mice, keyboards,
modems, and many other products. The reason
for the rapid growth of wireless technology is that
it provides the users with additional convenience
over the wired technology. In automotive systems,
both within and between the vehicles, there are
several applications pushing for the adoption of
wireless communications. Looking at in-vehicle
communications, more and more portable devices,
e.g., mobile phones, laptop computers and DVD
player can exploit the possibility of interconnec-
tion with the vehicle. Wireless communication
could be an attractive solution to reduce the
number of cables and disturbances in cars. To the
best of our knowledge, there are not a lot of wire-
less experiments in intra-car communications.
Below, we review potential wireless solutions
and discuss them. Among them, we specifically
focus on three techniques working in the 2.4
GHz band and allowing data rate from 250 kbps
up to 50 Mbps, namely Bluetooth, IEEE 802.11
b/g and ZigBee.
83
 Automotive Network Architecture for ECUs Communications
Bluetooth Wireless Technology
Bluetooth (“Bluetooth”, 2008) operates in the
2.4GHz ISM (Industrial Scientific Medical band)
band. It is a low-power, low-cost, short-range
wireless communication system developed by
a consortium of many manufacturers (Ericsson,
IBM, Intel, Nokia and Toshiba, etc). Dividing the
ISM band into 79 channels, the Bluetooth system
uses the Frequency Hopping Spread Spectrum
(FHSS) technology to operate on difference
channels with the frequencies changing rate of
1600 times per second when connected and 3200
times per second when in inquiry and page mode.
Bluetooth topology is made up of the piconets that
contain a single master node and up to seven slave
nodes. Communication among the slave nodes
has to be performed from the master node. Low
power is one of Bluetooth’s main advantages, from
100mW to 1mW, with a coverage range from 100
meters down to 1 meter. It provides speeds of up
to 3 Mbps (Bluetooth V2.0).
General Motors Corporation introduced a
Bluetooth network in its 2003 Saab 9-3 model car.
If the Bluetooth technology is used to establish
the wireless connection among the nodes in the
CAN or LIN bus, the volume of wires may de-
crease. Bluetooth wireless connection can supply
enough transmission rate for the control network.
Indeed, the asynchronous data transmission rate
in the Bluetooth network is up to 723.2Kbps
asymmetric (433.9Kbsps symmetric). While in
CAN, the data transmission rate for the real time
control sub network is typically > 250Kbps, and
about 125Kbps for the information sharing sub
network.
As demonstrated in (Chen & Chen, 2005),
though an in-vehicle wireless network can provide
the users with many convenient features, it can
also make the vehicle system vulnerable to many
types of security attacks and also has to face some
problems. As interference and setup delay time is
very dangerous for the vehicle’s safety, right now
84
the Bluetooth wireless technology is not proper to
be used in X-by-wire systems. However, it can be
used in the nonsafety-critical control systems. It
is estimated that diagnosis process may become
the first practical Bluetooth system level appli-
cation in the vehicles (Leen, Hefferman, 2001).
Other studies have been carried out in order to
build a secure Bluetooth network (Mahmud &
Shanker, 2006).
IEEE 802.11 B/G Wireless Technology:
Wi-Fi
IEEE 802.11 (Wi-Fi is in current usage) is a set of
standards for wireless local area network (WLAN)
computer communication, developed by the IEEE
LAN/MAN Standards Committee (IEEE 802) in
the 2.4 GHz, 5 GHz and public spectrum bands.
It includes IEEE 802.11b/g standards in the 2.4
GHz band and IEEE 802.11a in the 5 GHz band
(“802.11”; Gast, 2002).
IEEE 802.11b is a based on the direct-sequence
spread spectrum modulation technique. Techni-
cally, IEEE 802.11b applies Complementary
code keying (CCK) as its modulation technique.
As it uses the CSMA/CA media access method,
the maximum IEEE 802.11b throughput that an
application can achieve is about 5.9 Mbit/s using
TCP (7.1 Mbit/s using UDP). Then, the IEEE
802.11g was the third modulation standard for
Wireless LAN. The modulation scheme used in
802.11g is OFDM. Like IEEE 802.11b it works in
the 2.4 GHz band but operates at a maximum raw
data rate of 54 Mbit/s (19 Mbit/s net throughputs).
802.11g hardware is fully backwards compatible
with 802.11b hardware. For the two standards, two
network topologies are available namely ad-hoc
or infrastructure. A wireless ad hoc network is a
decentralized wireless network where each node
is willing to forward data for other nodes, and so
the determination of which nodes forward data
is made dynamically based on the network con-
nectivity. On the other hand, in the infrastructure
Automotive Network Architecture for ECUs Communications
network, a wireless access point connects wire-
less communication devices together to form a
wireless network.
Data rate measurements have been carried
out between two ECU/modems according to the
engine and equipments states using the infrastruc-
ture configuration. The test bed is built up of two
802.11g modems connected to laptop for IP trans-
fers. They are placed as for PLC cases. The authors
observe that the data rate decreases less than in the
PLC configuration according to the engine states
and equipments. We can achieve about 12 Mbps.
The worst configuration is when the transmitter
is placed in the trunk and the received antenna
is placed near the dash-board. Many reflections
appear and the signal is altered.
As for Bluetooth, IEEE 802.11 b/g can be pro-
posed as a potential wireless solution. However to
our knowledge there are not a lot of experimen-
tations of 802.11 b/g in vehicles. Besides, IEEE
802.11 b/g devices suffer interference from other
products operating in the 2.4 GHz. When 802.11b/g
is used in a point-to-multipoint configuration,
the overall bandwidth is dynamically demand
shared across all the users on a channel. Thus,
real time is no more guaranteed. In (Lin, Haniph,
Latcman & Neuman, 2003), the authors have
conducted field test in houses in order to compare
the performances of IEEE 802.11 in infrastructure
mode and HomePlug V1.0. Field tests show that
HomePlug had a better coverage than IEEE 802.11.
However, IEEE 802.11 was nearly always very
high (above 4 Mbits/s in 802.11b mode in 70% of
connections) but decreases more rapidly (less than
3 Mbits/s) than HomePlug (above 4Mbits/s in 58%
of connections) when distances are larger than 20
meters. In fact, the Homeplug performance is not
correlated with the line sight distances measured in
the experiment as the signals have to pass through
power lines cables. In (Lee, Latchman, Newman,
Katar & Yonge, 2002) the authors have carried
out similar experiments. This study compared
two technologies for implementing a small of-
fice network without extra cable installation. The
overall results show that the power line network
outperforms the wireless network.
It is not possible to directly apply the results
obtained for indoor power lines to automobiles
because the environment and topologies are to-
tally different. Thus, additional measurements
must be carried out to investigate these wireless
network solutions.
ZigBe Wireless Technology
An additional intra-car wireless sensor network
is based on ZigBee, which offers lower data rate
but can be used for many embedded applications.
It can be a good alternative to LIN network.
ZigBee is an industry alliance that promotes
a set of rules which builds on top of the IEEE
802.15.4 standards (“ZigBee”; Tsai, Saraydar,
Talty, Ames, MacDonald, & Tonguz, 2007).
ZigBee is a low-cost, low-power, wireless mesh
networking standard. It operates in three ISM
radio bands: 868 MHz in Europe, 915 MHz in
countries such as USA and Australia, and 2.4 GHz
(16 channels) in most jurisdictions worldwide. The
modulation uses direct-sequence spread spectrum
coding, which is managed by the digital stream into
the modulator. BPSK and QPSK could be used in
the 2.4 GHz band. The data rate is 250 Kbps per
channel in the 2.4 GHz band (40 Kbps in the 915
MHz band and 20 Kbps in the 868 MHz band).
Transmission range is between 10 and 75 meters,
although it is heavily dependent on the particular
environment. The maximum output power of the
radios is generally 0 dBm (1 mW). CSMA/CA is
the usually channel access mode, except three
cases (beacons, acknowledgment, messages for
beacon oriented network devices).
As ZigBee provides network speeds of up
to 250 Kbps, it is expected to be largely used
as a sensor network for monitoring and control
purposes (air conditioning, heating, ventilation,
lighting control, etc.).
85
 Automotive Network Architecture for ECUs Communications
In the article (Tsai, Saraydar, Talty, Ames,
MacDonald, & Tonguz, 2007), the channel be-
haviour under various scenarios is observed for
ZigBee nodes placed throughout a midsize sedan.
Results indicate that the change in link quality
with respect to locations of nodes in the car is
significant. Engine noise can increase the Packet
Reception Rate throughput received sensitivity
threshold by 2 to 4 dB. In general, the experiments
and measured results indicate that ZigBee is a vi-
able and promising technology for implementing
an intra-car wireless sensor network.
However, the communication between sensor
nodes and a base station in the car will depend on
several factors, such as the power loss, coherence
bandwidth, and coherence time of the underly-
ing communication channels between the sensor
nodes and the base station.
Other high data rate technologies are avail-
able such as IEEE 802.15.3a (UWB) which is a
competitor to the other IEEE 802.11 solutions. It
can provide speeds of up to hundred of Mbps/s.
It can be likely used in multimedia applications.
However, UWB components and boards are not
actually still available.
Wireless Technologies Comparison
and Design for Automotives
Todays, several applications push for the adoption
of intra-car wireless applications. Considering
voice and multimedia applications up to sensor/
controls applications, the technologies are already
available. The main differences between the wire-
less technologies we have considered come from
the different target applications they are designed
for. Bluetooth can be suitable for voice and audio
applications thus eliminating short distance ca-
bling. It can address small file transfer between
devices in a network range of a few tens of meters
and with network speeds of a few (1-3) Mbps.
IEEE 802.11 b/g can address audio, multi-media
applications as file transfer with much higher
throughput. Plug and play devices like personal
86
digital assistants, DVD players can communicate
with each other while they could also connect
with the other nodes already installed in the car
to provide a multifunctional device interface.
Documents from one PC could be displayed on
LCD screens installed in different places of the
vehicle (dashboard, back of seat). This solution
can compete with MOST and IDB1394 whose
main disadvantage is their dedication to a specific
application. Another potential use is inter-vehicle
communications as it consumes high power.
ZigBee, on the other hand, addresses sensors and
control. ZigBee applications are consisting of lots
of devices typically requiring small data packets.
It can provide speeds of 250Kbps which is not so
important for non-critical and safety applications.
Nodes can be scattered around in a slightly larger
area compared with Bluetooth.
Bluetooth can claim to replace some CAN’s
applications as ZigBee replaces some LIN func-
tions. But actually, no wireless network can an-
swer X-by-wire requirements allowing potential
Flexray replacement.
If we consider power usage and real time
constraints, both ZigBee and Bluetooth require
much less power than IEEE 802.11 b/g whereas
low power is a strong cars manufacturer’s require-
ment. From a real time point of view, multi-media
applications do not require strictly real time. On
the other-hand, safety-systems have real time
requirements, e.g., collision detection/avoidance
systems or active suspension systems that respond
to road conditions. Moreover, diagnostics tools can
make real time data available during operation of
the vehicle. Also, wireless networks can operate
as a redundant link between nodes linked with
wired different networks for real time applications.
Furthermore, in order to study these alternative
solutions, wireless channel measurements must
be carried out, considering different experimental
scenarios (engine ON/OFF, driving, passengers
present or not, driver always present, equipment
ON/OFF, …).
Automotive Network Architecture for ECUs Communications
Conclu
The partitioning of automotive applications into
several functional domains with their own require-
ments can be useful to manage complexity but it
leads to a great number of several independent
networks whereas more and more cross domain
data exchanges are necessary. Then, gateways
must be implemented and can impact perfor-
mances in term of real time. Such example of
cross-domain can be cited as the engine controller
(powertrain domain to body domain) or upcoming
X-by-wire functions.
An initial foreseeable solution with PLC has
been presented to improve interoperability be-
tween applications with different requirements.
PLC network answers both to cost, flexibility and
throughput requirements. Future work should be
devoted to optimize power line architectures in
order to minimize the number of cables, splicing,
etc. This implies rethinking the DC bundles as
rethinking the implementation of networks as
independent domain. Additionally, the technical
solution must be optimized taking into account
more channel measurements.
Wireless solutions can be adopted inside au-
tomotive applications as more and more devices
are both used in home and in cars while operating
as non-critical application or as redundant links.
Studies should be devoted to test these wireless
systems in intra-car communications while op-
timizing them to fulfil flexibility, fault tolerance,
security, real time and EMC requirements.
In 2003, AUTOSAR (AUTomotive Open Sys-
tem Architecture) (Fennel & Bunzel, 2006) has
been jointly developed by automobile manufac-
turers, suppliers and tool developers. It is defined
as an open and standardized automotive software
architecture. AUTOSAR aims as minimizing the
barriers between functional domains while facili-
tating the re-use of soft and hardware components
between different vehicle platforms. To achieve
this, AUTOSAR defines a methodology that sup-
ports a distributed, function-driven development
process and standardizes the software-architec-
ture for each ECU in such a system. AUTOSAR
also specifies compatible software-interfaces at
application-level. Applying AUTOSAR will im-
prove the interoperability between applications
and sub-networks. The automotive industry has
recognized that a technological breakthrough is
required and the development and standardiza-
tion of AUTOSAR must be handled by many
companies. Today, LIN, CAN and MOST are
AUTOSAR compliant.
In its concept, AUTOSAR introduces a
Virtual Functional bus (VFB) which separates
applications from infrastructure (i.e ECU and
networks). An application consists of intercon-
nected “AUTOSAR Software Components”.
The VFB provides standardized communication
mechanisms and services for the components.
It acts independently from the chosen mapping
of these components to the infrastructure of the
interconnected ECUs. Taking into account AUTO-
SAR concept, if the proposed networks (PLC or
wireless) are introduced in vehicles, they must be
adapted and must demonstrate their conformance
to the AUTOSAR standard.
REFERENCES
802.11 (1999). Technical Specification. Retrieved
from ref http://standards.ieee.org/getieee802/
download.
Afkhamie, K. H., Katar, S., Yonge, L., & Newman,
R. A. (2005). Overview of the upcoming Home-
Plug AV Standard. Power Line Communications
and Its Applications, International Symposium,
6-8 April, (pp. 400- 404).
Automotive Systems Research Group. (2008).
Retrieved from http://www.ttcan.com/ASRG
Beikirch, H., & Voss, M. (2000). CAN-trans-
ceiver for field-bus power line communication.
ISPLC Symposium, Limerick, Ireland, April,
(pp. 257–264).
87
 Automotive Network Architecture for ECUs Communications

Bluetooth (2008). Retrieved from http://www. Gouret, W., Nouvel, F., & EL Zein, G. (2006).
bluetooth.org/ Additional Network Using Automotive Powerline
Communication. IEEE International Power line
CAN. (2001). Retrieved from official web site
Communications and Its Applications Conference,
http://www.can-cia.org/
ISPLC 2006, 26-29 March, (pp 1087-1092).
Chen, Y., & Chen, L. (2005). Using Bluetooth
Gouret, W., Nouvel, F., & EL Zein, G. (2007). High
wireless technology in vehicles. IEEE Interna-
Data Rate Network Using Automotive Power Line.
tional Conference Vehicular Electronics and
IEEE International Power line Communications
Safety, 14-16 October, (pp. 344- 347).
and Its Applications Conference, ISPLC 2007,
CISPR25. (2002). Limits and methods of measure- 26-28 March. (pp. 1-4).
ment of radio disturbance characteristics for the
Hara, S., & Prasad, R. (1997). Overview of multi-
protection of receivers used on board vehicles.
carrier CDMA. IEEE Communications Magazine,
Technical report, CISPR-25, 2002.
December, (pp. 126-133).
Degardin, V., Laly, P., Liénard, M., & Degauque,
Heiskala, J., & Terry, J. (2002). OFDM wireless
P. (2006). Impulsive Noise on In-Vehicle Power
LANS: A Theorethical and Practical guide. Sams
Lines: Characterization and Impact on Commu-
Publishing.
nication Performance. IEEE International Power
line Communications and Its Applications Confer- Home Plug V1.0 (2008). HomePlug Powerline
ence, ISPLC 2006, 26-29 March, (pp. 222-226). Alliance. Retrieved from official web site www.
homeplug.org.
Degardin, V., Laly, P., Liénard, M., & Degauque,
P. (2007). Performances of the HomePlug PHY Huck, T., Schirmer, J., & Dostert, K. (2005). Tuto-
layer in the context of in-vehicle powerline com- rial about the implementation of a vehicular high
munications. IEEE International Power line Com- speed communication system. IEEE ISPLC.
munications and Its Applications Conference,
IDB Forum. (2008). Retrieved from web site
ISPLC 2007, 26-28 March, (pp. 93-97).
http://www.idbforum.org.
Fennel, H., & Bunzel, S. (2006). Achievements
IDB1394, (2008). 1394 Automotive Presentation,
and exploitation of the AUTOSAR development
Max Bassler, Shenzen, China, 10 April 2008. Re-
partnership, Copyright © 2006 CTEA. Retrieved
trieved from web site http://www.idbforum.org
from official web site http:www.autosar.org
IEEE 1901 (2008). Retrieved from official web
FlexRay Consortium. (2004). FlexRay Commu-
http://grouper.ieee.org/groups/1901/Scope_Pur-
nication System, Protocol Specification, Version
pose.html
2.0. Retrieved from: http://www.flexray.com
Iperf, (2008). Iperf tool and documentation,
Gast, M. S. (2002). 802.11 Wireless Networks, the
Retrieved from official web http://sourceforge.
definitive guide. O’Reilly Publishing.
net/projects/iperf.
Gavette, S., & other HomePlug members Home-
J2056/2 Survey. (1994). J2056/2 survey of known
Plug Powerline Alliance (2006). HomePlug
protocols. In SAE Handbook. Warrendale, PA:
AV Technology Overview. Technical report.
Soc. Automotive Eng. (SAE), vol. 2.
Sharp Laboratories of America. Retrieved
from download.microsoft.com/download. Johansson, K., Törngren, M., & Nielsen, L. (2005),
Handbook of Networked and Embedded Control

88
Automotive Network Architecture for ECUs Communications

Systems, D. Hristu-Varsakelis and W. S. Levine, MOST (2006). MOST Core Compliance Test
Eds. Boston, MA: Birkhäuser, 2005. Specification V2.00. Technical Report. Retrieved,
from http://www.mostcooperation.com/publica-
Kopetz, H., et al. (2002). Specification of the TTP/A
tions.
Protocol. (Technical Report). Vienna, Austria,
University. Technology. MOST Cooperation (2008). MOST Specification
Revision. Retrieved from http://www.mostnet.
LEA. (2008). LEA Organisation. Retrieved from
de.
official web site http://www.leacom.fr.
Navet (2008). The Automotive Embedded Systems
Lee, M. K., Latchman, H. A., Newman, R. E.,
Handbook. Industrial Information Technology
Katar, S., & Yonge, L. (2002). Field performance
series, CRC Press / Taylor and Francis, ISBN
comparison of IEEE 802.11b and HomePlug
978-0849380266, to appear November 2008.
1.0 . IEEE Conference Conference Local Com-
puter Networks LCN 2002. 27th Annual (pp. Navet, N., Song, Y-Q., Simonot-Lion, F., &
598- 599). Wilvert, C. (2005, June). Trends in Automotive
Communication Systems. Proceedings of the
Leen, G., & Hefferman, D. (2001). Vehicles with-
IEEE, special issue on Industrial Communica-
out wires. Computing & Control Engineering
tions Systems., 93(6), 1204-1223.
Journal, 12(5), 205-21.
Nolte, T. (2003). Reducing Pessimism and Increas-
Lienard, M., Carrion, M., Degardin, V., & De-
ing Flexibility in the Controller Area Network.
gauque, P. (2008). Modeling and Analysis on
Mälardalen University Licentiate Thesis, May
In-vehicle power line communication channels.
2003 Department of Computer Science and
IEEE Transaction Vehicular TEchnologie, 57(2),
Engineering Mälardalen University Västeras,
670-679.
Sweden, Copyright Thomas Nolte, Printed by
Lin, Y., Latchman, H., Katar, S., & Lee, M. K. Arkitektkopia, Västeras, Sweden Distribution .
(2003, May). Theoretical and Field Performance
Nolte, T. (2006). Share-Driven Scheduling of
Comparison between 802.11 Wireless and Power-
Embedded Networks. Mälardalen University,
line HomePlug 1.0 Protocols. IEEE Communica-
Dissertation, May 2006. Department of Computer
tions Magazine with Focus Theme on Powerline
Science and Engineering Mälardalen University
Local Area Networks, (pp. 54-63).
Västeras, Sweden, Printed by Arkitektkopia,
Lin, Y. J., Haniph, A., Latchman, H., Neuman, R. Västeras, Sweden Distribution.
(2003, April). A comparative Performance study
Nouvel, F., El Zein, G., & Citerne, J. (1994, June).
of Wireless and Power Line Network. IEEE Com-
Code division multiple access for an automo-
munications Magazine, (pp. 54-63).
tive area network over power-lines. Proceed-
LIN Consortium. (2003). LIN Specification Pack- ings of IEEE Vehicular Technology Conf., (pp.
age. Revision 2.0.Retrieved from http://www. 525–529).
lin~subbus.org/.
Road Vehicles. (2004). Controller Area Network
Mahmud, S., & Shanker, S. (2006, May). In-Ve- (CAN)—Part 4: Time-Triggered Communication,
hicle Secure Wireless Personal Area Network ISO 11 898-4.
(SWPAN). IEEE Transactions On Vehicular
Rubin, A. (2002). Implementing automotive
Technology, 55(3), 1051-1061.
protocols for communications over noisy battery

89
 Automotive Network Architecture for ECUs Communications
power lines. Proceedings of the IEEE Conference
on Conv. Elect Electron. Eng. Israel, December,
(p. 306).
Schiffer, A. (2000). Statistical channel and noise
modelling of vehicular dc-lines for data com-
munication. Proceedings of IEEE Vehicular
Technology Conf.,Tokyo.
Spdicom (2008). Spdicom Inc., Retrieved from
official web Official web site http://www.spid-
com.com.
Tsai, H., Saraydar, C., Talty, T., Ames, M., Mac-
donald, A., & Tonguz, O. (2007). ZigBee-based
Intra-car Wireless Sensor Network, Communi-
cations. IEEE International Conference ICC’07,
24-28 June, (pp. 3965-3971).
Valéo (2006). Valéo Inc., Electrical and Elec-
tronic Distribution Systems: Focus on Power
Line Communication. Retrieved from official
web http://www.valeo.com/ .
ZigBee (2000). ZigBee Specification. Version 1.0,
December. Retrieved from http://www.zigbee.
org;
KEY TERMS
Bluetooth: Wireless master-slave network in
the 2.4 GHz band
90
Control Area Network (CAN): Protocol
network
CSMA/CA: Carrier sense multiple access
with collision avoidance
ECU: Electronic Control Units
Flexray: protocol network
HomePlug: Power line communication net-
work
IDB 1394: Intelligent transportation systems
data bus using IEEE1394 technology—protocol
network
Local Interconnect Network(LIN): Protocol
network
Media Oriented Systems Transport
(MOST): Protocol network
OFDM: Orthogonal frequency division mul-
tiplexing
PLC: Power line communication
TDMA: Time division multiple access
IEEE 802.11 b/g: Wireless local area network
in the 2.4 GHz band (Wi-Fi)
ZigBee: Wireless sensor protocol network in
the 2.4 GHz band
 91

Chapter V
Enabling Secure Wireless
Real-Time Vehicle Monitoring
and Control
Lek Heng Ngoh
Institute for Infocomm Research, A*STAR, Singapore

ABSRACT

The use of in-car networked electronic controller units (ECUs) for monitoring and control of various
vehicle subsystems has become a common practice among the automotive manufacturers. In this chapter,
the author surveys one of the most popular in-car networking technologies, the Controller Area Network
(CAN), as well as newer and emerging in-car network technologies called FlexRay and Media-Oriented
System Transport (MOST). Currently, these networks are deployed for in-car applications such as en-
gine diagnostics, and infotainment systems. In this chapter, however, the author extends the use of these
embedded vehicular networks by proposing to remotely monitor and control the vehicles through them,
in order to realize safety and driver assistance related applications. To accomplish this task, additional
technologies such as real-time wireless communications and data security are required, and each of
them is introduced and described in this chapter.

Introduction and send information from and to the on-road ve-

In this chapter we present the increasingly im-
portant technological advances in a networked
vehicle and the related technologies in achieving
secure wireless real-time monitoring and control
of these vehicles. As more electronics found their
way into the vehicle, the ability to remotely gather
hicles opens up many application opportunities in
the effort to make future vehicles safer (both the
drivers and road users), more economical, more
fun (e.g. multimedia entertainment) and last but
not least, to simplify the often mammoth tasks of
traffic management and road maintenance. How-
ever, for this vision to become a reality technology

Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Enabling Secure Wireless Real-Time Vehicle Monitoring and Control
development must take place on several fronts.
This chapter attempts to present an overview and
analysis of these key technical components.
The chapter is organized as follows. We begin
by elaborating the controller area network (CAN)
standard which is the most widely deployed in-
vehicle network system. Our description includes
an overview of the CAN technology (Section III),
its deployment in a vehicle (Sections IV and V).
The above is followed by a description of the vari-
ous CAN-base car telematic applications (Section
VI). A brief introduction is given in Section VII
on other networking standards that are currently
deployed in vehicles. In Sections VIII and IX,
we elaborate two key components in achieving
remote vehicle monitoring and control, namely
the wireless communication component and data
security component. In Section X, we present an
overall architecture of the secure wireless real-
time vehicle monitoring and control environment.
Finally, in Sections XI and XII, we discuss some
possible future trends in this field and conclude
this chapter.
II. Background
The Controller Area Network (CAN) (CiA, 2008;
Audi CAN-bus, 1998) is a serial bus communica-
tions protocol developed by Bosch in around1983.
CAN is a standard suitable for efficient and reli-
able communication between sensors, actuators,
Figure 1. CAN Replaces traditional mesh wiring in a vehicle
device 1
ECU
device 2
Without CAN Bus
92
controller, and other nodes in real-time applica-
tions. Since then, CAN has evolved to become
the de facto low-cost standard in a large variety
of networked embedded control systems. Many
of the early deployments of CAN are found in
automotive vehicles. To-date, however, CAN has
been used in a variety of application domains such
as transportation (e.g. trains), manufacturing (e.g.
process automation), construction (e.g. structural
monitoring), agriculture (e.g. agriculture equip-
ment), healthcare (e.g. medical devices), enter-
tainment (e.g. arcade game machine), and even
in the control of scientific apparatus (e.g. proton
accelerators). In this chapter, however, we will
only focus on applications of CAN within the
automotive vehicles (see Figure 1).
In 2007, it is estimated that sale of vehicular
CAN nodes has reached more than 600 million
each year (Johansson, K.H., Törngren M., &
Nielsen L., 2005). Looking ahead, the number
of CAN nodes deployed in automotive vehicles
is expected to increase substantially as by 2008,
all new vehicles sold in the US will be required
to be CAN-compliant, in particular, all new ve-
hicles are to equip with CAN-based diagnostic
system. Moving forward, the number of automo-
tive manufacturers adopting CAN for in-vehicle
embedded data communications and control is
expected to increase as more Japanese models
have started to adopt CAN-based technologies
which are relatively low-cost and a well-estab-
lished standard.
device 1
ECU
device 2
With CAN Bus
Enabling Secure Wireless Real-Time Vehicle Monitoring and Control
In the rest of this chapter, we will discuss
the various system components involved to ac-
complish the goal of monitor and control vehicle
remotely and securely. We first elaborate the basis
and the general technology landscape of CAN
deployments in automotive industry; followed
by an overview of the deployment of in-car CAN
networks. Next we discuss the various existing and
future applications that make use of the CAN data.
The two key technologies components needed to
achieve secure remote vehicle monitoring, namely
data security and wireless communications are
discussed toward the end of the chapter.
III. OVEF CAN
TECHNOLOGIES
An overview of CAN technologies is presented
in this section.
CAN is a twisted-pair serial lines with real-
time communications protocol particularly suited
for connecting sensors (for monitoring purposes),
and actuators (for control purposes) which are
commonly found in today’s vehicles – a high-end
passenger car today can have up to hundreds of
sensors and actuators connected by 80 electron-
ics control units (ECUs) (Audi CAN-bus, 1998)
which form the CAN nodes (see Figure 1). Details
of the CAN bus operations are described in many
literatures elsewhere. An overview is provided
here and interested readers are referred to these
literatures for further information (CiA, 2008).
All CAN nodes are connected to the CAN bus
in parallel, and communicate with each other via a
data frame which forms part of the CAN standard
(see Figure 2 and Figure 3). A data frame begins
with a single start-of-frame (SOF) bit, followed
by eleven bits of identifier (ID) (there is also an
extended data frame with twenty-nine bit identi-
fier), and the remote transmission request (RTR)
bit. The ID and RTR bits form the arbitration field
(see Figure 3). The control field consists of six
bits, and indicates how many bytes of data there
are in the data field – which can be zero to eight
bytes long. The data field is followed by the cyclic
redundancy checksum (CRC), which allows the
receiver to check if the data has been corrupted
during transmission. Next the two-bit acknowl-
edgement (ACK) field is for any of the receiver to
indicate to the transmitter that it has received a
valid frame. Finally, the end of message is signally
through a seven-bit end-of-frame (EOF). We shall
describe the transmission process next.
As all CAN nodes are allow to transmit data
onto the CAN bus any time, therefore it belongs
to the class of protocols known as carrier sense
multiple access/collision avoidance (CSMA/CA).
In order to provide the necessary arbitration among
the conflicting CAN nodes, it uses a non-destruc-
tive arbitration process. This process requires
the transmitting nodes to listen to the network
in order to avoid collisions. This is how it works:
whenever the CAN bus is free, any ECU can start
to transmit a message, any possible conflicts due
to more than one ECU transmitting simultane-
ously are resolved by bit-wise arbitration using
the identifier of each message. During arbitra-
tion, each ECU transmits its message identifier
and compares it with what level it “hears” from
the network. If it wins in the bit-wise arbitration,
the transmission process for this ECU continues,
otherwise it quits transmitting. Using this process,
CAN bus guarantees that there will be only one
transmitter left transmitting at the end of the
arbitration phase, with the rest of the ECU revert
to receiving (or listening) mode.
For cost control reasons, CAN bus comes in
different speed ratings depending on the functions
that it is deployed to control. Currently there are
three classes of CAN bus (Class A to Class C).
Class A speed rating is the slowest, typically
around 10 kbps (thousand bits per second). It is
typically deployed to control components such
as power mirror, power windows, door locks and
lights. Class B CAN, by comparison, may operate
for 10 kbps to 125 kbps. It is fast enough to carry
more complex information and time-sensitive
93
 Enabling Secure Wireless Real-Time Vehicle Monitoring and Control
data such as electronic instruments, electronic
transmission controls, climate controls. Finally,
Class C CAN bus is the fastest with bit rate up to
1 Mbps. Many of the vehicles that are currently
using a Class C data bus are operating at speeds
of around 500kbps. At this rate, it is fast enough
for power-train control modules, air bag modules,
and fast acting anti-lock break and stability control
system. The development of CAN continues, and
it can be expected that faster versions of CAN bus
will make their way into a shop near you soon.
The CAN protocol defines the lowest two
layers (i.e. physical and data-link layer) of the
Open System Interconnection model i.e. OSI’s
Figure 2. Parallel ECU connections to CAN bus
Sensors/Actuators
Electronics
Control Unit (ECU)
CAN Bus
Figure 3. Details of CAN message data frame
Arbitration Field
(12 bits/30 bits)
RTR (1 bit)
Start field
(1 bit)
Control field
(6 bits)
Message ID
(11 bits/29 bits)
94
7-layer model, in order to meet certain applica-
tion requirements, higher layer protocols have
also been developed to be added to the current
two-layer specification. Many of these higher
layer protocols were developed for applications
outside of automotive industry (e.g. distributed
industrial automation) (Kvaser, 2008; AALCAR,
2008). However, these protocols have now become
relevant as we explore transmitting of CAN bus
data remotely (see Section VIII).
An example of a higher layer protocol is the
CANopen (Kvaser, 2008) which is a standardized
layer defined for distributed industrial automation.
It specifies communication profiles and device
Sensors/Actuators
Electronics
Control Unit (ECU)
CAN Bus
End field
CRC (16 bits) (7 bits)
Data field ACK field
(max 64 bits) (2 bits)
Enabling Secure Wireless Real-Time Vehicle Monitoring and Control

profiles, which enable an application-indepen- Figure 4. An example CAN deployment in car

dent use of CAN. In this case device profiles
are defined for most common devices such as Headlamp
Wipers
analogue-to-digital converters. Furthermore,
CANopen also distinguishes real-time data verses
Power train
less time-sensitive data. Other standardized higher CAN ECU
layer protocols include DeviceNet (Kvaser, 2008)
CAN Bus
(adding Internet protocols TCP/IP on top of CAN).
Another higher layer protocol is CAN Kingdom
(Kvaser, 2008). CAN Kingdom was developed door Audio
mainly for maritime industry. Its main feature is speakers
that it allows changing of CAN network behavior
at any time, including whilst in run-time. CAN
Kingdom allows for system troubleshooters to
turn off individual nodes. It will also allow for
dynamic and run-time change of node identifier
Table 1. Examples of CAN ECU for powertrain
and conditions for sending messages.
and chassis subsystems
Finally, new developments have also taken
place for basic CAN protocols. The latest ex- Action Data Speed
tension of CAN includes the capabilities such as Engine control Monitor/Control 500kps
Time-Trigger communication on CAN (TTCAN,
Gearbox control Monitor/Contro 500kps
(TTCAN, 2008)) verses the traditional event-trig-
Gear selector lever Monitor/Contro 500kps
gered CAN. TTCAN allows the nodes to have
Break control unit Monitor/Contro 500kps
guaranteed access to the communication bus with
a fixed periodic interval.

IV. EXAMPLE DEPLOYMENT OF Table 2. Examples of CAN ECU for convenience

CAN NETWORK IN VEHICLES subsystems

Action Data Speed

In the automotive industry, there has been a
remarkable evolution over the last few years in Left turn signal Monitor 125 kbps

which embedded control systems have grown from Right turn signal Monitor 125 kbps

standalone control systems to highly integrated High Beam Monitor 125 kbps

and networked control systems. Many car manu- Wiper Hi Monitor 125 kbps

factures today design their control architectures Wiper Lo Monitor 125 kbps
based on CAN, including BMW, Chevrolet, Fiat, Wiper int. approx 17 seconds
Monitor 125 kbps
interval
Ford, Mazda, Mercedes, Renault, SAAB, Toyota,
Wiper int. approx 15 seconds
Volkswagen, and Volvo. interval
Monitor 125 kbps
Typical CAN-based distributed control archi- Head light “on” Lo Monitor 125 kbps
tecture would consist of multiple CAN networks of
Head light “on” Hi Monitor 125 kbps
ECUs namely power-train and chassis subsystem,
Reverse light “on” Monitor 125 kbps
infotainment subsystem, and body electronics
Brake light “on” Monitor 125 kbps
subsystem (see Figure 4). Tables 1-2 show the
details of these three CAN networks.

95
 Enabling Secure Wireless Real-Time Vehicle Monitoring and Control
V. MANUFACTUERE -SPECIFIC CAN
AND LoCATION DATA
Even though CAN bus is a well-established
standard for intra-car data communications, car
manufactures have taken the liberty to implement
and deploy the network freely. The reality is that
there is no standard of allocating the data bits
with respect to the various CAN sub-systems.
Each car manufacturer has therefore assigned and
defined their own CAN bit pattern for respective
CAN node. This vendor-specific information
must therefore be obtained in order for any soft-
ware to decode and make use of the CAN data.
Furthermore, most vehicles today are equipped
with Global Positioning System (GPS) receivers
where the location information of the vehicles
is constantly obtained. As we will elaborate in
the next section, location information for each
vehicle is crucial to many of the vehicle telematic
applications, as well as future multi-hop car-to-car
communications, and should therefore be handled
and communicated along with the CAN bus data
described in this section.
VI. VEHICLE APPLICATIONS USING
CAN BUS DATA
Today the adoption of CAN bus is not only
motivated by saving and simplifying cabling as
was the case in the beginning. A number of ap-
plications such as diagnostic and engine control
functions have been developed. Beyond these,
some highly integrated and networked control
systems that involve distributed coordination of
many sub-systems are also been developed. In
many exiting and future vehicles, information of a
car’s multiple ECUs will be shared and exchanged
over the CAN bus. Examples of use of this data
include vehicle diagnostics, dynamics control
and even drive-by-wire systems (e.g. brake-by-
wire, steer-by-wire etc.). Whilst up to now data
is only available and used within the vehicle,
96
the transmission and use of this data externally
from one or more vehicles are yet to be exploited
for a wide range of applications, some of which
are even beyond our current imagination. As
described in previous section, CAN bus has been
used for not just vehicle’s power-train and chassis,
it has also been deployed at both the comfort and
convenience (e.g. seat, parking aid, door, interior
climate control), as well as infotainment (radio,
phone, navigation) subsystems. All these mean
that with the aplenty CAN basic data, a host of
remote applications can be developed to cater for
all aspects of need of the drivers and their vehicles
(Jaman G., 2007). In the rest of this section, some
of these futuristic applications will be described
at conceptual level to give readers a sense of the
emerging applications at the horizon.
1. Maintenance and diagnosis: Based on the
above description of CAN operations thus
far, it should not be difficult to conclude that
in order to read the data on CAN bus, one
has to do so through a device mimicking
the operations of a “receiver-only” ECU
node. Therefore, CAN-compliant SCAN
tools (AALCAR, 2008) are currently
available commercially to read the vehicle
CAN data serially. All CAN ECU modules
can send and receive “I’m OK” signals to
inform each other their status. On the other
hand, whenever a problem occurs in one of
the modules, it will not generate the “ok”
signal and this in turn will be detected
by the SCAN tool and inform the user. A
faulty ECU module in this case may not
start a vehicle, and hence the follow-up ac-
tion is for the vehicle to be checked and the
faulty ECU replaced at a workshop nearby.
Similarly, making the CAN data remotely
will enable such condition to be detected in
advance and preventative maintenance can
be carried out regularly.
2. Safety: CAN networks have played a
significant role in providing extra safety
Enabling Secure Wireless Real-Time Vehicle Monitoring and Control
features to the driver. Case in point is the
electronic stability control (ESC) that is
fast becoming a standard feature in even
mid-range passenger vehicles. Essentially
the ESC systems can detect unstable driving
situation by comparing the driver’s intended
course with the vehicle’s actual moment
using CAN ECU data such as wheel speed,
angle, yaw rate and lateral acceleration, and
take necessary control action to stabilize the
car and keep it on the road. Control actions
of ESC include the use of a combination of
anti-lock brake (ABS) and electronic force
distribution etc. With more sophisticated
sensory technologies (e.g. radar and video),
future ESC systems will be able to detect
signal such as driver’s physical conditions
(e.g. drowsiness) and take appropriate cor-
rective action such as to take-over control
of the vehicle and park it somewhere safe!
All the monitoring and actuation are com-
municated to the respective ECU via CAN
networks.
3. Traf. c telemetric: The use of on-the-road
vehicles as speed sensors and to rely the
information to a centralized server, where
information is gathered and learned about the
traffic condition, about a stretch of a road,
have proven to be feasible (Johansson, K.H.,
Törngren M., & Nielsen L., 2005). With CAN
bus information made available, collective
information regarding the road conditions
and even weather condition (via information
on the use of headlights and wipers) can be
obtained. A host of vehicular applications
and services can then be developed to make
use of this information.
4. Driver assistance: Some of the up-and-com-
ing driver assistive functions for automotive
are the adaptive cruise control (ACC) and
parking assistance. ACC’s main function is
to adapt vehicle speed with respect to the
information such as distance to the front
car, obstacles and even weather (e.g. fog)
conditions. Similar to the sensor modalities
used for providing safety features such as
ESC, ACC and parking assistance systems
will make use of the information collected
through CAN networks, as well as using the
CAN network to communicate commands
to both the driver and relevant electronic
systems
Next, automatic parking assistance systems
would probably rely on ultrasonic sensors mounted
on the sides of the vehicle to measure available
parking space length and depth. CAN bus would
be used by the system to communicate commands
to the subsystems such as steering control and
brakes.
Similar to the above cases, with remote moni-
toring and control, such features can be provided
cheaper and earlier as many of these systems
would require large computing resources (e.g.
CPU power) which may not be readily available
in a car.
5. Infotainment-on-the-go: With the prolifer-
ation of personal entertainment devices and
the availability of intelligent multipurpose
monitor screens and audio devices in passen-
ger cars, it is only a matter of time before the
drivers are able to download infotainment
content such as popular songs and surf the
net on-the-go. CAN bus is already expected
to play a significant role in the control and
configuration of these in-car infotainment
devices to facilitate these applications.
Clearly business aspects such as market in-
troduction and key success factors need to be
considered together as the related technologies
are being overcome. It is worth noting that the
relevant business issues here are no different
from any Internet-based applications and ser-
vices which we are familiar. However, one key
important difference is that the availability (i.e.
adoption rate) of car communication will greatly
97
 Enabling Secure Wireless Real-Time Vehicle Monitoring and Control
influence the kinds of applications that are likely
to be introduced first. The two key factors are, the
different data quality of service (QOS) that each
mode of communication can deliver (e.g. data
timeliness and reliability), and the availability of
cars equipped with each mode of communication.
Thus for example, car-to-car communications can
support low-data latency but it is not expected to
be widely available over-night, and therefore will
limit the class of applications (e.g. lane merging)
that would only work well if large percentage of
vehicles are equipped with the capability to detect
and transmit data in a car-to-car manner.
VII. OTHER IN-VEHICLE
NETWORKING STANDARDS
Whilst CAN is the most commonly found in-vehi-
cle network today, there are other network systems
which are being deployed to either inter-work or
supersede CAN. An example of the inter-working
network is the Local Interconnect Network (LIN)
(LIN, 2008). LIN standard precedes CAN and is a
single-wire bus system. A LIN network is typically
made up of a master node and one or more (up to
16) slave nodes connected serially to the master
node. A slave node can typically be a sensor such
as a tilt sensor or anti-theft alarm sounder. It can
also be an actuator in the form of an intelligent
electronic or electromechanical assembly which
comes under the control of a master node. The
master node controls the speed (normally 1 to
20 kbits/sec), sequence, as well as the frequency
of data transfer between itself and slaves nodes.
A number of LIN master nodes, in turn, can be
connected to a CAN ECU node.
Next, with increasingly sophisticated electron-
ics being added to the vehicles, a greater demand
is placed on the interconnecting network bus in
relation to data speed, time-determination and
fault tolerance. FlexRay (FlexRay, 2008) is a
98
new generation standard that has been designed
to meet these requirements better than CAN. For
the start, FlexRay is capable of delivering data rate
of 10 Mbits/sec that is almost 10 times more than
the fastest CAN available. FlexRay protocols also
provide better timing requirement by transmitting
messages in fixed time slots (verses the asyn-
chronous CAN), and superior fault-tolerance and
redundant message transmission on two separate
physical channels. Unfortunately, FlexRay had not
been designed to inter-work with CAN network
in mind, and therefore a complete replacement of
CAN is unlikely to happen anytime soon.
Finally, an emerging future vehicular network
standard that is gaining acceptance is the Media
Oriented Systems Transport (MOST) (MOST,
2008) networking standard. Given the expected
rise in in-car telematics and entertainment systems
such as digitized audio, video contents, loca-
tion-based applications and broadband Internet
connections, a much higher data transfer rate of
more than CAN’s maximum 1 Mbits/sec is needed.
For example, a transfer rate of at least 6 Mbits/sec
is required for the transmission of digital TV
signals and high-quality audio. For this reason,
MOST network is defined to handle transfer rate
in excess of 20 Mbits/sec. MOST networks make
use of the large available raw bandwidth in opti-
cal fibers, with optical transceivers connected to
each MOST node. Nodes are normally connected
in the form of a ring-bus.
Finally, wireless networking standards such
as Bluetooth (Bluetooth, 2008), Ultra-wideband
(UWB, 2008) and related wireless USB (USB,
2008) are expected to play roles in the future
in-vehicle data networking solution. Wireless
networks have the key advantages of easy instal-
lation and zero risk of cable-related damages.
However, more research is required to ensure that
these networking standards can adapt well in a
relatively noise vehicular environment. Another
key consideration is the protection of data being
transmitted by these means.
Enabling Secure Wireless Real-Time Vehicle Monitoring and Control
Figure 5. Multiple modes of remote data communications
(1)Request
(1)Response
Wireless Channels
(4)
(3)Request
V III. REMOTE DATE
COMMUNICATIONS
With the proliferation of wireless communication
infrastructure and technologies, plus inter-opera-
tor roaming arrangements, sending and receiving
data remotely is already a reality (IEEE 802.11p,
2008; DSRC, 2008; IEEE802.11, 2008). To sup-
port CAN data communications, many of the
standardized higher layer CAN protocols such as
CANopen, CAN Kingdom (see Section III) will
allow for seamless migration to IP-based cellular
data services (e.g. GPRS, HSDPA, EDGE), and
thus realize remote CAN data communications
easier. However, what is not so certain is the need
to ensure timeliness and reliable communications
of CAN data between the vehicle and say applica-
tion servers. As shown in Figure 5 above, there
are typically few different modes of communica-
tions between the moving vehicles and data server
(typically housed within a data centre), as well as
among the vehicles themselves.
1. The first ((1) in Figure 5) is server
requests and vehicle responses. This
mode of communication is achieved by
vehicle communicating via a road-side
communication infrastructure or cel-
lular phone system. It is typically used
by the server to configure, disseminate
control information to the vehicles, and
Application
Server
(2)Data
(3)Response
even availability test of the communica-
tion channel between the server and the
vehicle.
2. Using the same communication chan-
nels as the above, the second mode of
communications caters for periodic data
sent from the vehicles to the server and
vice-verse. This mode of communication
is also used for the communication of
monitored data including non-real time
as well as urgent/emergency data from the
vehicles to the server and vice verse.
3. In this mode, again using the same com-
munication channels as the first two, the
vehicles are able to query server for infor-
mation pertaining to server’s parameters,
as well as link test the return channel
availability – this is required as there is
often the case that wireless channels are
of asymmetric nature.
4. Finally, data can also be exchanged via
car-to-car communications (DSRC,
2008); IEEE802.11, 2008). This ad-hoc
communication technology is to be
implemented in each participating vehicle
and enable the autonomous exchange of
data between them in a speedily man-
ner without going through the road-side
infrastructure and server as in previous
three cases.
99
 Enabling Secure Wireless Real-Time Vehicle Monitoring and Control
Indeed, it should be obvious from the above
that depending on the mode of server-vehicle com-
munications, and the nature of CAN data, there
would be different transmission requirements
(e.g. tolerable delay, loss and bandwidth needed
etc.) for each of the communication modes above.
Whilst today cellular data services support roam-
ing and base-station handover effectively, with
some schemes even allow for seamlessly transfer
of access to other wireless broadband internet
technologies such as WIFI, EDGE, WIMAX
and vice verse. There are other constraints such
as cost, coverage and end-to-end delay of these
basic communication services that need to be taken
into consideration; another word, establishment of
basic data connectivity is required but it will not
satisfy the quality-of-service (QOS) requirements
for remote CAN data communications.
An enhanced solution would include additional
software functions such as a link QOS manager
and a connection manager to be added at the com-
munication infrastructure provider end, as well
as additional software components to be included
into the monitoring server ends to facilitate link
QOS measurements (latency, bandwidth, cost
etc) between the data server and vehicles being
monitored. Constant monitoring and proactive
handover and roaming actions will be initiated
by the QOS manager and connection manager to
ensure that data communication QOS between the
server and vehicles is satisfactory and cost effec-
tive. Furthermore, end users should be prompted
in the event where coverage of the network is
insufficient or the required QOS cannot be met
given the available communication channels.
Finally, standard for future car-to-car (i.e. ad-
hoc) and car-to-infrastructure (i.e. in the form of
dedicated short-range communications (DSRC)
and/or IEEE 802.11p) communications are also
being developed by the international community.
The deployment of DSRC can take many forms,
ranging from delay-tolerant “data filling” where
DSRC nodes are deployed every few kilometers
apart (e.g. on lamp posts) to communicate with
100
moving vehicles once they come within the cover-
age range, to hop-by-hop ad hoc vehicle-to-vehicle
communications. Whilst these emerging stan-
dards are not expected to substantially change the
way vehicular data is being remotely monitored,
one can expect improvement in cost as well as
communication availability once these standards
are fully adopted and implemented (Kosch, T.,
2004; Matheus, K. 2002).
IX. DATA SEURITY
With the emerging vehicular networks in the forms
of infrastructure-to-car (I2C), car-to-car (C2C),
car-to-infrastructure (C2I) communications, as
well as in-car CAN ECU described so far, a va-
riety of applications for safety, traffic efficiency,
driver assistance, as well as infotainment to be
incorporated into future automotive designs. At
the same time, networked CAN ECUs are increas-
ingly being deployed in cars to realize functions
such as engine management, air-bag deployment,
and even in intelligent brake systems. However,
the availability of these exciting automotive
applications can also lead to formidable set of
exploits and vulnerable to attacks, as more and
more data is being exchanged and communicated
in and out of a car. Therefore, one of the most
important aspects of remote vehicle monitor-
ing is to ensure the integrity of CAN data being
exchanged between the vehicle and monitoring
application server. Unlike other communication
networks, the vehicular network differs in size,
speed of the vehicles, as well as relevance of
their geographic position. Most importantly, the
sporadic connectivity between them also made
providing data security to vehicles especially
challenging.
There are some possible security threats and
attacks (Raya M., Hubaux J.P., 2005; Hubaux, J.P.,
Capkun, S., & Luo, J. 2004; Zarki M.E., Mehrotra
S., Tsudik G., & Venkatasubramanianm N., 2002;
Eymann T.,2006) due to the limited security:
Enabling Secure Wireless Real-Time Vehicle Monitoring and Control
1. Eavesdropping: Eavesdrop and record of a
warning message about emergency vehicle,
or diffuse wrong information in the network
to affect the behavior of other drivers (e.g.,
divert traffic from a given road and thus free
it).
2. Denial of service: Accessibility of a service
is restricted (e.g., channel jamming and ag-
gressive injection of dummy messages).
3. Bogus information: Faking of a warning
message.
4. Spoofing: Take over of the identity of an
authorized device.
5. ID disclosure of other vehicle: Surveillance
of the vehicle motions by using the C2C and
C2I infrastructure.
6. Cheating with sensory information: Alter
the perceived position, speed, direction, etc.
in order to escape liability notably in the
case of an accident.
What is required is an integrated info-security
solution to manage and protect the critical data
for intra-car, car-to-car and car-to-infrastructure
communications. Respective key information
security issues to be addressed can be listed as
follows:
1. Intra-car communications: Anti-theft and
confidentiality of real-time sensor and ECU
data;
2. Car-to-car communications: Messages
exchanged must have stringent constraints
on confidentiality, integrity, privacy and
time-sensitivity; finally
3. Car-to-infrastructure communications:
Integrity of traffic-related information, fu-
ture entertainment related data (e.g. songs
and video), and even liability-related mes-
sages must be protected. Other issues include
the availability of network infrastructures
to provide communications.
The above solutions should ensure that data
within and without the automotive is protected
from abuse and security attacks. In order to ad-
dress this holistically, one begins by identifying
the various security threats and attacks that vari-
ous vehicular applications and communication
networks are vulnerable to. Next security analysis
on these networks and applications needs to be
carried out in each of these cases. And finally,
proposed solutions and design will include three
major aspects: tamper-resistant software in em-
bedded devices for intra-car security, security
solutions for vehicle ad-hoc network (VANET)
and, last but not least, standard compliant.
X. An OVEALL ARCHITECTURE
In summary, Figure 6 shows an overall architec-
ture of the secure wireless real-time vehicle moni-
toring and control. The various components of this
architecture have been elaborated previously in
this chapter. The in-vehicle node is responsible
for the interfacing and monitoring of vehicle via
the in-vehicle networks such as CAN and MOST;
it is also capable of conveying control data from
the backend applications to the vehicle. Another
key function of this component is to communi-
cate securely with the other vehicles via its C2C
and/or C2I communication protocol stack. In
C2C mode of communication, it often involves
the exchange of real-time messages either with
one of the vehicle within its range, or broadcast
to all the vehicles within its range. The C2I mode
on the other hand, is with the stationary service
node where communication can be via ad-hoc
and/or cellular data service such as GPRS.
The main functions of the service node include
communicating with the on-the-road vehicles via
their in-vehicle nodes, as well as facilitating the
request, access and the delivery of one or more
car-telematic services and applications to the
vehicles. The service node therefore enables the
101
 Enabling Secure Wireless Real-Time Vehicle Monitoring and Control

provision and delivery of IT applications to the
on-the-road vehicles in a secure and transparent
manner. As shown in Figure 6, with the help of a
service node, services are made available to mobile
digital devices such as smart phones and wireless
laptops, as more and more of these devices are
being carried by drivers and passengers alike.
Finally in Figure 6, the various applications
and services are being offered via the backend
cloud service middleware to ensure that they
are available in a feasible (e.g. transferring from
smart phone to in-car display terminals), reliable
and secure manner between the backend hosted
services and vehicles. The details involved are be-
yond the scope of this chapter, but key middleware
functions here include secure single sign-on for
users, user profiles management, accounting and
charging, as well as flexible session monitoring
and control. It is important to note that a number
of vendor-based de-facto standards have started to

Figure 6. Overall architecture of secure wireless real-time vehicle monitoring and control

v ideo server
Location-
based
server

b ackend
v ehicular
Traffic Mgt
Telematic s ervices
and s ecurity data
Applications portal

s ervice
s ignaling

data

s ervice Node
C2I and Cellular
s ecure Communications

C2C
Mobile devices

In-vehicle Node

102
Enabling Secure Wireless Real-Time Vehicle Monitoring and Control
emerge (GEARS, 2008; LiveMesh, 2008), as well
as international standard bodies have defined and
are currently working on standards in these areas,
especially those which are related to networks and
communications (WAVE, 2008; C2C, 2008).
XI. FUTURE TRENDS
As automotive vehicles continue to play an impor-
tant role from a country’s economy to individual
life style, efforts in making a car better in terms
of comfort, safety, reliable and even fun to drive
will continue relentlessly. One of the key trends
along this development has been the proliferation
of embedded data monitoring and control network
systems within a vehicle (aka networked vehicle).
In this chapter, we elaborated one of the most
popular in-car networking technologies called
Controller Area Network (CAN). It is also worth
noting that newer and emerging in-car network
technologies such as FlexRay and Media-Oriented
System Transport (MOST) are also being deployed
in higher-end vehicles. The ability to monitor
and control a vehicle wirelessly for safety, main-
tenance, reliable and entertainment purposes is
fast becoming a reality in not too distance future.
Indeed, as more digitized contents find their way
to the vehicles in the form of down-loaded and
streamed audio and video infotainment-on-the-go
applications and services, the availability of higher
speed like MOST will become more urgent.
XII. CONCLUS
This chapter began by introducing the basic
CAN technologies and subsequently focused on
elaborating the potential applications and related
technical challenges in achieving secure remote
monitoring and control of vehicles via CAN.
Furthermore, the details of various modes of
wireless communications were presented. Last but
certainly not least, the importance of protecting
data and information between the vehicle and ap-
plication server was also discussed. In conclusion,
the author sees innovation in real-time monitoring
and control area being the key research platform
to kick-start a potentially fertile ground for future
automotive innovations and services.
RE
AALCAR. (2008). Automotive CAN Network
Diagnostics. Retrieved October 13, 2008, from
http://www.aa1car.com/library/can_systems.
htm
Audi CAN-bus. (2008). Trainer Information.
Edition 1
Bluetooth. (2008). Bluetooth standard. Retrieved
October 13, 2008, from http://www.bluetooth.
com
CiA. (2008). CAN in Automation. Retrieved Oc-
tober 13, 2008, from http://www.can-cia.org/
C2C. (2008). European based Car-to-Car com-
munication Consortium. Retrieved October 13,
2008, from http://www.car-to-car.org
DSRC. (2008). Dedicated Short Range Com-
munications. Retrieved October 13, 2008, from
http://grouper.ieee.org/groups/scc32/dsrc/
Eymann, T. (2006). The EASIS Security Architec-
ture Approach. 1st C2C-CC Security Workshop.
Berlin, November 2006.
FlexRay. (2008). FlexRay In-vehicle Network.
Retrieved October 13, 2008, from http://www.
interfacebus.com
GEARS. (2008). Google GEARS. Retrieved Oc-
tober 13, 2008, from http://gears.google.com
Hubaux, J. P., Capkun, S., & Luo, J. (2004). The
Security and Privacy of Smart Vehicles. IEEE
Security and Privacy Magazine 2(3) (2004),
49-55.
103
 Enabling Secure Wireless Real-Time Vehicle Monitoring and Control
IEEE 802.11p. (2008). Retrieved October 13,
2008, from http://en.wikipedia.org/wiki/IEEE_
802.11p
IEEE802.11. (2008). IEEE 802.11 working group.
Retrieved October 13, 2008, from http://grouper.
ieee.org/groups/802/11/
Johansson, K. H., Törngren, M., & Nielsen, L.
(2005). Vehicle applications of controller area
network. In D. Hristu-Varsakelis & W. S. Levine
(Ed.), Handbook of Networked and Embedded
Control Systems, Birkhäuser.
Jaman, G. (2007). Structural Monitoring Using
Wireless Sensors and Controller Area Network.
IEEE 5th Annual Conference on Communication
Networks and Services Research (CNSR’07).
Kosch, T. (2004). Technical Concept and Prereq-
uisites of Car-to-Car Communication. Retrieved
October 13, 2008, from http://www.car-to-car.
org/fileadmin/dokumente/pdf/C2C-CC_presen-
tation_3_BMW.pdf
Kvaser. (2008). CAN Kingdom. Retrieved October
13, 2008, from http://www.kvaser.com/index.
htm
LIN. (2008). Local Interconnection Network
(LIN). Retrieved October 13, 2008, from
http://www.interfacebus.com/Design_Connec-
tor_LIN_Bus.html
LiveMesh. (2008). Microsoft Live Mesh. Re-
trieved October 13, 2008, from http://news.cnet.
com/8301-10784_3-9926578-7.html
104
Matheus, K. (2002). Car-to-Car Communication
– Market Introduction and Success factors. Re-
trieved October 13, 2008, from http://www.car-to-
car.org/fileadmin/dokumente/pdf/C2C-CC_pre-
sentation_4_VW_Audi_Carmeq.pdf
MOST. (2008). Media Oriented System Trans-
port (MOST). Retrieved October 13, 2008, from
http://www.interfacebus.com/MOST_Bus_In-
terface.html
Raya, M., Hubaux, J.P. (2005). The Security of
Vehicle Ad Hoc Networks. Proceedings of the
ACM Workshop on Security in Ad hoc and Sen-
sor Networks (SASN’05), pp. 11-21, Alexandria,
Virginia, USA, November 2005.
TTCAN. (2008). Time Triggered Controller Area
Network (TTCAN). Retrieved October 13, 2008,
from http://www.ttcan.com/
USB. (2008). Wireless USB. Retrieved October
13, 2008, from http://www.usb.org/wusb/
UWB. (2008). Ultra-wideband Technolgy. Re-
trieved October 13, 2008, from http://www.intel.
com/technology/comms/uwb/
WAVE. (2008). IEEE Wireless Access in Vehicu-
lar Environment (WAVE), IEEE 1609. Retrieved
October 13, 2008, from http://standards.ieee.
org/announcements/pr_P1609.0_new.html
Zarki, M. E., Mehrotra, S., Tsudik, G., & Ven-
katasubramanianm, N. (2002). Security Issues
in Future Vehicular Network. Proceedings of
European Wireless Conference 2002.
 105

Chapter VI
MAC and Routing Protocols for
Vehicle to Vehicle
Communications
Xiaobo Long
Rensselaer Polytechnic Institute, USA

Biplab Sikdar
Rensselaer Polytechnic Institute, USA

ABSTRACT

Numerous efforts are currently under progress to enhance the safety and efficiency of vehicular traffic
through intelligent transportation systems. In addition, the growing demand for access to data and in-
formation from human users on the go has created the need for advanced vehicle-to-vehicle and vehicle-
to-roadside communication systems capable of high data rates and amenable to high degrees of node
mobility. Vehicular communications and networks are expected to be used for a number of purposes such
as for enabling mobile users to transfer data and information from other networks such as the Internet
and also for implementing services such as Intersection Decision Systems (IDS), Automated Highway
Systems (AHS), and Advanced Vehicle Safety Systems (AVS). In this chapter the authors describe medium
access control (MAC) and routing protocols for vehicular networks and the various factors that affect
their design and performance.

Introduction and tion between road-side beacons and individual

Background
PHY Layer Issues
Vehicle-to-Vehicle (V2V) communication sys-
tems, as part of future Intelligent Transportation
Systems (ITS), require bidirectional communica-
vehicles. A recent standard for vehicle-to-vehicle
communication is the 5.9 GHz Unlicensed Na-
tional Information Infrastructure (UNII) been
developed to support high mobility in IEEE
802.11 protocol. This emerging 802.11 standard
for 5.9 GHz is generally referred to as Wireless
Access in the Vehicular Environment (WAVE),

Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.
 MAC and Routing Protocols for Vehicle to Vehicle Communications
which provides Dedicated Short Range Com-
munication (DSRC) for future vehicle-to-vehicle
communications. The radio link of these commu-
nication systems has a range varying between a
few meters and hundred meters (Eichler, 2007).
Wireless LAN infrastructure is employed to build
inter-vehicle communication based applications.
WAVE networks consist of seven 10MHz wide
channels, with Channel 178 being a control chan-
nel and the other 6 being service channels. The
channel switching mechanism in IEEE 802.11h is
adapted to perform the multi-channel operation.
The physical (PHY) and Medium Access Control
(MAC) layers of 5.9 GHz DSRC are based on IEEE
802.11a PHY, which uses an OFDM (Orthogonal
Frequency Division Multiplexing) waveform and
the IEEE 802.11 MAC respectively. However, the
vehicular environment and the requirements for
fast access introduce unique requirement on the
MAC functionality. While IEEE 802.11 is intended
for stationary or low mobility environment, com-
munication in mobile WLAN (wireless local area
network) must be able to transfer messages to and
from vehicles at speed up to 120 miles per hour.
Vehicular traffic scenarios pose greater chal-
lenges than the indoor WLAN applications, due
to associated driving speed, varying vehicular
traffic patterns and driving environments. The
new standard is expected to establish require-
ments for packet error rate performance in these
challenging environments. The IEEE is currently
working on the IEEE 802.11p Wireless Access in
Vehicular Environments (WAVE) standard. There
are two types of inter-vehicle communication
in WAVE. One is inter-vehicle communication
using AP (Access Point). Access points may be
provided or co-located at street corners, traffic
lights, emergency phones, parking lots, rest ar-
eas, gas stations or other shops in service areas.
This approach requires development of exclusive
infrastructure and is cost expensive. The other is
inter-vehicle communication supported by wire-
less ad-hoc network. In this mode, there is no
distributed 802.11 beaconing mechanism. Only
106
the basic carrier sense mechanisms are used, and
the RTS (request to send)/CTS (clear to send) ac-
cess is prohibited, with the purpose to facilitate
fast access to vehicles. In ad-hoc network, each
mobile vehicle detects other’s position and rout-
ing path.
The other potential candidate of vehicle-to-ve-
hicle communication protocol is the IEEE 802.16e
Wireless Metropolitan Area Network (WMAN)
standard, operating in the 2-6 GHz frequency
band. In Europe, a microwave frequency of 5.8
GHz is recommended. Two or four channels
with 5MHz bandwidth each and a maximum
Equivalent Isotropically Radiated Power (EIRP)
of 3 db Watt are available (Baranowski, Lienard
& Degauque, 1991).
In vehicle-to-vehicle communication, multipa-
th fading is a major problem for short range mobile
microwave links. Multipath fading is caused by
reflected waves from the road, vehicles, traffic
signs and buildings, which interfere with the direct
signal. As a result, the received signal strength is
position dependent and time varying. The interfer-
ence of reflected waves can be destructive to the
received radio power and lead to signal outages.
Usually the reflected waves have less amplitude
than direct wave in vehicle communications and
the wave components can be absorbed, scattered
or reflected. The received power is determined
by the radiating characteristics of beacon and
the vehicle antenna as well as statistically dis-
tributed reflectors. To study the link performance
in multipath fading channels, both deterministic
channel models and statistical channel models are
developed in literature. A deterministic channel
model was derived in (Pfeiffer & Schrei, 1994)
for short range radio link,
λ0 1
Pr = PG
t r Gt ( ) | T |2 (1.1)
4π nt
where the subscript t and r denote the transmitter
and the receiver antenna, respectively. P indicates
MAC and Routing Protocols for Vehicle to Vehicle Communications
the signal power, G is the antenna gain and λ0 is
the wavelength in free space. nt is the number
of transmitter antenna elements. Parameter T is
determined by many factors, including the number
of transmitter and receiver antenna elements, the
number of propagation paths and their length, the
radio radiating patterns and the reflection factors
of the reflecting surface.
In statistical models, the propagation channel
can be modeled as a time varying linear filter,
the impulse response given by (Wang, Sen &
Matolak, 2007),
L −1 L −1
h(τ,t) = ƒ zl (t )α l (t )e jθl ( t ) = ƒ hl (t )δ(τ − τl (t ))
0 0
(1.2)
Where αl(t) and θl(t) represents the amplitude
and phase of the lth received wave component,
which has a time-varying time delay τl(t). The
δ function is the Dirac delta. Compared to con-
ventional wireless channels, vehicle-to-vehicle
environments can have frequent and rapid channel
changes due to mobility and low transmitting and
receiving antenna heights. Thus multipath com-
ponents are modeled as on/off process in (Wang,
Sen & Matolak, 2007). z(t) in Equation (1.2) is the
persistence process which is used to account for
the finite lifetime of propagation paths. The on/off
process z(t) is modeled by a Markov chain. The
received signal strength for the empirical channel
is modeled using Weibull distribution given by
(Wang, Sen, & Matolak, 2007) for channels in
802.16e systems where distance between trans-
mitter and receiver is large,
Pw ( x) = β / αβ xβ−1 exp(−( x / α)β ) Environments and Mobility
(1.3)
Where β is a shape factor that determines fad-
ing severity, α = E ( x 2 ) / Γ(2 / β + 1) is a scale
parameter, and Γ is the gamma function. β = 2
yields well known Rayleigh distribution, and β < 2
implies more severe fading. In vehicle-to-vehicle
802.11 or WAVE systems, a direct transmission
path between receiver and transmission is usually
required. Rician fading channel, with a direct line-
of-sight (LOS) component and a strong ground-
reflected component, is used to model vehicle to
vehicle channel in many scenarios. The dominant
component in Rician fading channel is likely to be
relatively strong compared to the reflected signal,
which lead to large Rician κ factor. The delay
spread is likely to be relatively small since reflec-
tions usually occur in the immediate vicinity of the
transmitter and receiver antenna. However, when
the antenna height is low and propagation distance
is large, the LOS component is not presented in
the received signal. In this case, Rayleigh fading
is a reasonable model.
To achieve reliable communication against
multipath fading and the associated Doppler shift
as well as interference from other links using the
same frequency, diversity techniques, includ-
ing antenna diversity and frequency diversity,
are employed to improve the link performance
(Grabow, & Schrei 1992). In IEEE 802.11e stan-
dard, Orthogonal Frequency Division Multiple
Access (OFDMA) scheme provides permutation
subcarrier allocations that reduce the probability
of experiencing bad fading on all subcarriers of
each subchannel. Other multiple access schemes
such as Time Division Multiple Access (TDMA),
Direct-Sequence Code Division Multiple Access
(DS CDMA) are also investigated. Packet Error
Rate (PER) or Bit Error Rate (BER) are often
used to measure the performance of the various
multiple access schemes for a given spectrum
allocation.
A vehicle-to-vehicle communication system might
encounter different propagation environments
that vary with traffic density and geography.
Vehicle-to-vehicle communication scenario can
be classified as follows (Wang, Sen, & Matolak,
2007; Singh, Bambos, Srinivasan, & Clawin,
107
 MAC and Routing Protocols for Vehicle to Vehicle Communications
2002). The first scenario is urban environment,
where vehicular speeds limit at around 40 miles
per hour and driving environment has a few build-
ing structures and roadside tree plantations. The
vehicles stop at traffic lights but not frequently.
This urban scenario can be further classified into
urban-antenna inside car (UIC) and urban-antenna
outside car (UOC). The second scenario is city
environment, which has speed limits of 25 miles
per hour amidst roadside building constructions.
The traffic scenario can be assumed to be the
rush hour traffic, with vehicles often stopping
at traffic lights and in jams. The third scenario
is freeway, which is an open environment with
scarce roadside vegetation and driving speed
limit of 65 miles per hour. There can be vehicles
traveling between and along the communica-
tion test vehicles. This freeway scenario can
be further classified into open area-low traffic
density (OLT) and open area-high traffic density
(OHT). Several analytical and empirical models
have been proposed to model vehicle-to-vehicle
channel under varying conditions. Geographical
conditions, vehicle speed and traffic patterns,
such as the distance between moving vehicles
and relative velocity and directions all affect the
network performance. Measurements show that
the quality of the wireless communication channel
degrades with increasingly stressful communica-
tion scenarios, which lead to degrading in network
performance including decreasing throughput and
increasing PER or BER. The quality of wireless
channel can be indicated using Signal to Noise
Ratio (SNR) at the receiver or received signal
strength (RSSI). The link quality usually degrades
with increasing distance between transmitter and
receiver. Also the link quality often deteriorates
with increase in relative and average velocities
of the vehicles. Measurements indicate that for
vehicle-to-vehicle ad-hoc WLAN networks, the
sub-urban environment is the most favorable and
the urban driving conditions are the most hostile
for inter-vehicle communication. The freeway
environment link quality lies in between the
108
two. Early and accurate knowledge of rapidly
time variant fading channel conditions enables
adaptive data transmission in vehicle-to-vehicle
communication systems, which in turn improves
the quality of service for end users and reduces
the power consumption for data transmissions. Ef-
ficiency in data transmission in vehicle-to-vehicle
systems can be enhanced through design of MAC,
routing and other application layer protocols that
adapt to changing link conditions. For example,
in Mobile Ad-hoc Network (MANET), switch-
ing the packet size in certain scenarios can help
enhance the network performance. An increase in
packet size is observed to increase the throughput
when the vehicle to vehicle distance is small and
link quality is good (Singh, Bambos, Srinivasan,
& Clawin, 2002). However, for larger distances
between transmitter and receiver and poor link
quality, the throughput is better for smaller packet
size. For adverse scenarios with vehicles separated
by a large distance, a larger packet size would lead
to a larger packet error rate and hence a lower
throughput. It is therefore desirable to switch
the packet size beyond a specific inter-vehicle
separation. Similar work on adaptive frame length
control for optimizing performance is discussed
in (Lettieri & Srivastava, 1998).
To model the mobility of vehicles, a set of pa-
rameterized mobility models is suggested in (Bai,
Sadagopan & Helmy, 2003) to capture the mobility
characteristics of ah-hoc vehicle-to-vehicle com-
munication scenarios. Four models are proposed
including Random Waypoint, Group Mobility,
Freeway and Manhattan models. For Random
Waypoint model, it does not be able to capture
the mobility characteristics of spatial dependence,
temporal dependence and geographic restrictions
in the mobile network. In Group Mobility model,
each group has a group leader that determines the
group motion behavior. Initially, each member of
the group is uniformly distributed in the neighbor-
hood of the group leader. Then, at each instant,
every node has a random speed and direction
deviating from that of the group leader. Group
MAC and Routing Protocols for Vehicle to Vehicle Communications
Mobility model can be used in military battlefield
communications where the commander and sol-
diers form a logical group. In Freeway Mobility
model, each mobile node is restricted to its lane
on the freeway, and the velocity of mobile node
is temporally dependent on its previous velocity.
If two mobile nodes on the same freeway lane
are within the Safety Distance (SD), the velocity
of the following node cannot exceed the veloc-
ity of preceding node. In Manhattan Mobility
model, the mobile node is allowed to move along
the grid of horizontal and vertical streets on the
map. At an intersection of a horizontal street
and a vertical street, the mobile node can turn
left, right or go straight. The mobility pattern
influences the connectivity of MANET, which in
turn influences the routing protocol performance.
Different mobility models may drastically affect
the performance of MANET routing protocols
including DSR (Dynamic Source Routing), AODV
(Ad Hoc On-Demand Distance Vector Routing)
and DSDV (Destination Sequenced Distance
Vector routing). It is observed that DSR, DSDV
and AODV achieve the highest throughput and
the least overhead with Group Mobility model
and incur high overhead and low throughput
with both Freeway and Manhattan models (Bai,
Sadagopan & Helmy, 2003).
MAC Protocol for VEHhicl
to VEHhicl NETtwork
The fundamental task of MAC protocols is to
schedule transmissions from stations sharing
a common channel with the primary goal of
preventing collisions, thereby leading to higher
throughput and bandwidth utilization. The mo-
bility and speed of the communicating nodes in
vehicular networks adds another dimension to the
challenges faced by the MAC protocols in addition
to their requirements of reliability and efficiency.
In this section we review some of the existing
MAC protocols for vehicular network.
Basic MAC Protocols
A number of protocols have been proposed
for mobile ad hoc networks in the recent past,
some of which may find application in vehicular
networks with varying degrees of success. The
earliest MAC protocol for wireless networks,
ALOHA (Abramson, 1970) is a decentralized,
randomized protocol that can be easily applied
to vehicular networks due to its simplicity. In
this protocol, any node with a packet to transmit
initially goes ahead with the transmission without
regard to other ongoing transmissions. In case
of a collision, the node goes into a backoff and
waits for a random duration before attempting a
retransmission. While the protocol is simple, its
throughput is limited to 18.4 percent for fixed
packet lengths. As an improvement over this
basic access mechanism, a slotted version of the
protocol, S-ALOHA (Roberts, 1975) may be used
where time is divided into slots and a node may
transmit only at the beginning of a slot. The slot-
ted operation reduces by half the period during
which an ongoing transmission may encounter
a collision, thereby increasing the maximum
throughput to 36.8 percent.
The primary source of degradation in the
throughput of MAC protocols is from the col-
lisions that result when the transmissions from
two nodes overlap in time. To further reduce the
incidence of collisions as compared to ALOHA
and S-ALOHA, Carrier Sense Multiple Access
(CSMA) has been proposed and widely used by
many MAC protocols (Kleinrock & Tobagi, 1975).
With CSMA, a node with a packet to send first
senses the channel for activity before initiating
its transmission. If the channel is idle, the node
proceeds to transmit its packet immediately. On
the other hand, if the channel is busy, the node
defers its transmission till the current transmis-
sion ends. At the end of the current transmission
the node may probabilistically decide when to
attempt its transmission. However, collisions are
still possible if two nodes decide to transmit at the
109
 MAC and Routing Protocols for Vehicle to Vehicle Communications
same time following the end of an ongoing trans-
mission. To further improve the MAC protocol’s
performance and reduce overhead due to colli-
sions, CSMA protocols with Collision Detection
(CD) mechanisms have also been proposed for
MAC protocols (Metcalfe & Boggs, 1976). While
effective in wired networks, collision detection is
difficult to implement in wireless transceivers.
A major contributor to the MAC layer collisions
in wireless networks is the presence of “hidden”
nodes. Hidden terminals occur when a node that
is in the neighborhood of the receiver of an on-
going transmission but outside the range of the
transmitter, initiates its own transmission thinking
that the channel is idle, thereby causing a colli-
sion at the receiver of the ongoing transmission.
The incidence of hidden nodes is expected to be
higher in vehicular networks due to node mobil-
ity. Multiple Access with Collision Avoidance
(MACA) uses a two-step handshake mechanism
to avoid hidden terminals (Karn, 1990). When a
node wants to send a packet, the transmitter initi-
ates the handshake by first sending a RTS packet
to the destination with the intention of reserving
the channel for the transmission of the packet. All
nodes in the vicinity of the transmitter overhear
this RTS packet and defer their transmission for
this duration. If the RTS packet is successfully
received by the destination and it is available for
receiving the packet, it sends a CTS packet to
the transmitter node. All nodes in the vicinity
of the receiver that receive the CTS packet are
thus aware of the impending transmission and
defer their own transmissions. On receiving the
CTS packet, the transmitter proceeds with the
transmission of the data packet while avoiding
any hidden terminals.
Hidden terminals may also be suppressed
through the Busy Tone Multiple Access (BTMA)
MAC protocols (Haas & Deng, 2002). The solution
requires the splitting of the available bandwidth
into two channels, one for data transmissions
and the other for control signals. A node with
a data to transmit that uses the BTMA protocol
110
first senses the control channel for activity before
transmitting its packet. If the control channel is
idle, the node proceeds to transmits its packet
in the data channel while simultaneously trans-
mitting a busy tone in the control channel. All
neighbors of the transmitter that sense the busy
tone also transmit the busy tone. Thus two-hop
neighbors of the transmitter also receive the busy
tone and defer transmitting on the data channel,
thereby preventing transmissions from hidden
terminals. The Dual Busy Tone Multiple Access
(DBTMA) protocol extent the BTMA protocol by
using two busy tones or control channels (Haas
& Deng, 2002). With this extension, the sender
uses one tone to inform its neighbors about the
ongoing transmission while the receiver uses the
second busy tone to inform its neighbors about
its reception.
Finally, time division multiple access (TDMA),
frequency division multiple access (FDMA) and
code division multiple access (CDMA) have also
been proposed for mobile networks (Jurdak, Lopes
& Baldi, 2004). In the TDMA based protocols,
time is split into fixed frames which are further
divided into slots. Each node transmits in the
slots that it is assigned. In FDMA, the medium
is split into non-overlapping frequency bands.
Each user is allocated a separate band and users
may transmit simultaneously without interfering
with each other. In CDMA, each user encodes its
messages with one of several orthogonal codes
specified by the system. Since the codes are
orthogonal, nodes may transmit in parallel. The
main drawback of these protocols in the context
of vehicular networks is that they tend to be
centralized, making it difficult to adapt easily to
changing network conditions.
The IEEE 802.11 MAC Extension for
Vehicular Networks
The IEEE 802.11 standard covers both the physical
(PHY) and the MAC layer of wireless networks.
The PHY layer can use any of the following
MAC and Routing Protocols for Vehicle to Vehicle Communications
modes: (1) Direct sequence spread spectrum;
(2) Frequency hopping spread spectrum; (3) In-
frared pulse modulation. The channel data rate
can either be from 1Mbps to 56Mbps depending
on the version of the protocol, with higher rates
expected to be supported in future versions. In
addition to the physical layer specifications, the
IEEE 802.11 standard specifies two methods for
medium access: Distributed Coordination Func-
tion (DCF) and the Point Coordination Function
(PCF). While the implementation of the DCF is
mandatory and is based on a distributed, backoff
based channel access mechanism, the PCF is
optional and in this access method, the nodes
are polled by a “master” residing within the base
station. The channel access mechanism alternates
between the DCF and PCF modes when PCF is
implemented. The duration of time that the DCF
is used for channel access is termed the contention
period (CP) and the polled duration is called the
contention-free period (CFP). The lengths of the
CP and the CFP are controlled explicitly by the
contention free period repetition interval (CFPri).
In vehicular networks, the DCF may be used for
vehicle-to-vehicle communications while the
PCF may be used for vehicle-to-infrastructure
communications.
Each CFP begins with a beacon frame and
the CFPs occur at a defined repetition rate as
determined by the CFPrate parameter. With PCF,
the access to the channel is determined centrally
by the base station, usually referred to as the
Point Coordinator (PC) that provides a conten-
tion free transfer service. The PC gains control
of the medium at the beginning of the CFP, and
maintains control for the entire CFP by waiting
for a shorter time between transmissions than the
stations using the DCF access mode. All stations
other than the PC set their network allocation
vectors (NAVs) to the CFPMaxDuration at the
start of each CFP. The PC transmits a CF-End
or CF-End+ACK frame at the end of each CFP
and on receiving either of these frames a station
resets its NAV. During the CFP, the base sta-
tion polls the nodes for a single pending frame
transmission according to a list ordering of their
association with the base station, known as the
polling list. The PC starts CF transmissions at
a SIFS (Short Inter-Frame Space) interval after
the beacon frame by sending a CF-Poll (no data),
Data or Data+CF-Poll frame. If a station receives
a CF-Poll (no data) frame from the PC, the station
can respond to the PC after a SIFS interval with
a CF-ACK (no data) or a Data+CF-ACK frame.
If the PC receives a Data+CF-ACK frame from
a station, it can send a Data+CF-ACK+CF-Poll
frame to a different station where the CF-ACK
part is used to acknowledge receipt of the previ-
ous data frame. If the PC transmits a CF-Poll (no
data) frame and the destination station does not
have any data to transmit, the station sends a Null
Function (no data) frame back to the PC. If the
PC fails to receive an ACK for a transmitted data
frame, it waits for a SIFS interval and moves on
to the next station in the polling list.
The DCF mode of operation of the IEEE 802.11
protocols is based on CSMA with collision avoid-
ance (CA). In a network with a CSMA/CA MAC
protocol, each node with a packet to transmit first
senses the channel to ascertain whether it is in use.
If the channel is sensed to be idle for an interval
greater than the Distributed Inter-Frame Space
(DIFS), the node proceeds with its transmission.
If the channel is sensed as busy, the node defers
transmission till the end of the ongoing transmis-
sion. The node then initializes its backoff timer
with a randomly selected backoff interval and
decrements this timer every time it senses the
channel to be idle. The timer has the granularity
of a backoff slot and is stopped in case the channel
becomes busy and the decrementing is restarted
when the channel becomes idle for a DIFS again.
The node is allowed to transmit when the backoff
timer reaches zero. Since the backoff interval
is chosen randomly, the probability that two or
more stations will choose the same back off value
is expected to be low. Along with the Collision
Avoidance, 802.11 uses a positive acknowledg-
111
 MAC and Routing Protocols for Vehicle to Vehicle Communications
ment (ACK) scheme. All the packets received
by a node implementing 802.11 MAC must be
acknowledged by the receiving MAC. After
receiving a packet, the receiver waits for a brief
period, called the Short Inter-Frame Space (SIFS),
before it transmits the ACK. IEEE 802.11 also al-
lows the use of the two-step handshake mechanism
of MACA to avoid hidden terminals.
IEEE 802.11p is an amendment to the IEEE
802.11 standard to add wireless access in vehicular
environments (WAVE). DSRC is a general pur-
pose communications link between the vehicle
and the roadside (or between vehicles) using the
802.11p protocol. The MAC layer operation of
IEEE 802.11p uses the same CSMA/CA mecha-
nism as used in the legacy IEEE 802.11 systems.
However, IEEE 802.11p allows the use of service
classes as defined in the IEEE 802.11e standards
to provide service quality differentiation. At the
PHY layer, the use of an OFDM system is expected
to allow vehicle-to-vehicle as well as vehicle-to-
infrastructure communications for distances of
up to 1000m and relative speeds of 200 km/hr.
The protocol may operate either in the 10MHz
or the 20MHz channels, allowing data rates from
3 to 56 Mbps.
Other MAC Protocols for Vhicular
Networks
The ADHOC-MAC is a slotted MAC protocol
for vehicular networks that is based on the reli-
able reservation ALOHA (RR-ALOHA) protocol
(Borgonovo, Capone, Cesana, & Fratta, 2004).
The RR-ALOHA protocol is an extension of
the reservation ALOHA protocol (Crowther,
Rettberg, Walden, Ornstein & Heart, 1973). The
reservation ALOHA protocol is a combination of
slotted ALOHA and time division multiplexing
(TDM) where a given number of slots constitute
a group or frame. An idle slot in a group or frame
is considered to be available for all other nodes.
Thus in subsequent frames, all nodes may contend
for transmission in the slot. If a node is success-
112
ful in a slot, that particular slot is reserved for
that user in subsequent frames. Once the node
finishes transmitting all its data, it leaves the
slot idle, allowing other nodes to contend for this
slot. The RR-ALOHA protocol extends this basic
protocol to achieve slot reservation and dynamic
TDMA in a distributed way while all nodes have
a global view of the transmissions in its two-hop
neighborhood, in order to avoid hidden terminals.
Again, time is slotted and a number of slots are
grouped together to form frames. To achieve
this, each node is required to transmit a vector
that indicates how the node sensed the status of
the slots in the previous frame (i.e. information
on which node transmitting in which slot). The
transmission is done in the slot assigned to the
node. Each node receiving this message updates its
view of the system state based on the status-vector
received from each node. When a new node or
vehicle comes into an area, it first listens for the
entire frame to select a free slot in which it may
transmit. Once the node attempts a transmission
in a free slot, it observes the channel-state vectors
received from other nodes in the next frame. If all
the neighboring nodes mark the slot in which the
new node transmits with the ID of the new node, it
implies that this slot is reserved for the new node
in the two-hop neighborhood and may be used by
the new node in subsequent frames.
While the protocols described above consider
omni-directional transmissions, the fact that the
movement of vehicles is restricted by the orienta-
tion of the roads and driving restrictions may be
used by directional transmissions. The Directional
MAC (D-MAC) protocol is based on the IEEE
802.11 MAC but requires that each node be aware
of the geographical location of its neighbors as
well as itself (Ko, Shankarkumar & Vaidya, 2000).
In this protocol, the RTS packet is sent either
directionally or omni-directionally, based on the
ongoing transmissions in the node’s neighborhood.
Each directional antenna on a node that receives
a RTS or CTS packet updates its NAV accord-
ingly and does not transmit on that antenna for
MAC and Routing Protocols for Vehicle to Vehicle Communications
the duration specified in the RTS or CTS packet.
However, the node may transmit on antennas that
do not have any ongoing transmissions in their
direction. The use of directional antennas adds
complexity to the system but can improve the
spatial reuse while reducing collisions.
Routing Protocol
Routing Performance Requirements
Future Intelligent Transportation Systems re-
quire fast and reliable communication between
cars (vehicle-to-vehicle) or between a car and a
road side unit (vehicle-to-infrastructure). In Ad
hoc vehicle-to-vehicle communication where no
supporting infrastructure is required, vehicles
communicate when they are within the radio
range of each other, or when multiple hop relay
via other vehicles are available. Messages need
to be routed from the source to one or several
destinations. Desirable characteristics of routing
protocols include:
• Minimal control overhead
• Loop-free routing paths
• Low complexity
• Multicast capabilities
Beside the above requirements, the vehicular
environment poses new challenging requirements
to vehicle-to-vehicle routing protocol design,
including:
• Adapting routing information in highly
mobile topologies
• Short convergence time of the routing algo-
rithms
• Short delay for neighbor discovery
• Scalability
Compared to general wireless ad hoc networks,
the roadway traffic environment in vehicle-to-ve-
hicle communication consists of uncoordinated
vehicles whose neighboring vehicles constantly
change and do not have a priori relationships
with each other (Chennikara-Varghese, Chen,
Altintas & Cai, 2006). As a result, the network
topology in vehicle-to-vehicle communication
changes very fast. Dynamic topology maintenance
techniques are needed to adapt routing informa-
tion in highly mobile topologies. Mobile Ad Hoc
NETworks (MANET) protocols face several
challenges when applied to vehicle networks.
In MANET, nodes usually move together as a
group and have synchronous movement which
ensures that the nodes have a priori knowledge
of each other. Although these protocols work
well for coordinated nodes, it may not work well
also for vehicles on a highway where nodes are
generally unknown and unrelated. For example,
MANET protocols can incur high latencies in
configuration an operation time with uncoordi-
nated vehicles. As vehicle mobility increases, the
available routes will change frequently which lead
to even larger delay in routing algorithms opera-
tion. Furthermore, the collision and contention of
the underling MAC protocols such as 802.11 may
increase routing convergence time and delay for
neighbor discovery (Chennikara-Varghese, Chen,
Altintas & Cai, 2006). This is because many
nodes within radio coverage tend to advertise
their routing information at the same instant. In
an ad hoc vehicle network environment where
node density is high, such as traffic congested
area, this is especially true. Beside the above
issues, scalability is also a challenging issue in
routing protocol design in vehicle networks where
the network can go from scarce to dense in a very
short time. Parameters from other layers, such
as wireless channel quality, power efficiency or
application specific requirements may also affect
the routing decisions and can be considered in
aiding the design of routing protocols.
113
 MAC and Routing Protocols for Vehicle to Vehicle Communications
Ad-Hoc Routing Techniques
Ad hoc unicast routing schemes can be divided into
two categories: topology-based routing and posi-
tion-based routing. Topology-based schemes use
a proactive or reactive approach to create routes.
Position-based routing schemes do not use routing
tables to store route. Instead, they use position
information about neighboring and destination
nodes to determine the next hop forwarding to
destination. The routing techniques can also be
categorized as flat routing and hierarchical routing.
In flat routing algorithm, each node has the same
responsibility in the routing process. Flat routing
techniques suffer from scalability issues but still
are the most popular ad hoc routing techniques.
In hierarchical routing, responsibility is only as-
signed to a certain number of nodes that provide
required routing information to the rest of the
nodes (Chennikara-Varghese, Chen, Altintas &
Cai, 2006; Böhm, 2007).
Topology-Based Routing
In flat topology-based routing schemes, two
general approaches are used. One is proactive
routing, which is also called table-driven routing.
The other is reactive or on-demand routing. In
proactive routing protocols, route information is
stored in routing tables, and nodes exchange route
updates periodically to reflect any changes to the
network topology. The routing table updates can
be periodic or event-driven which quickly react
to network changes. The main advantage of pro-
active routing is that a route can be immediately
selected from the routing table when it is needed.
However, additional control traffic is needed to
update unused route entries repeatedly. When
the routes need to be updated very frequently,
network traffic congestion may occur, which
in turn triggers off delays, retransmissions and
packet losses. Examples of proactive protocols
include DSDV, Optimized Link State Routing
(OLSR), Fisheye State Routing (FSR) which is a
114
link state behavior protocol and Wireless Rout-
ing Protocol (WRP), a path finding Algorithm.
However, the existing proactive routing solutions
for mobile ad hoc networks are not really suitable
for inter-vehicle communication because of the
quickly changing topology. In a vehicular ad hoc
network, the routes are required to be updated
very frequently, which adds control overhead and
influences the total throughput over the network.
Besides, the risk of using outdated routes would
be high even with very frequent route updates
(Böhm, 2007).
Reactive routing schemes establish routes only
when applications need to send packets to the
destination. These routes expire after some time.
A route discovery phase is needed each time an
application wants to send data over the network. In
the route discovery phase, route request messages
are partially or entirely flooded in the network to
find the shortest path to the destination. In reac-
tive routing methods, the communication delay
is usually larger than proactive routing, but the
control overhead is smaller. Besides, scalability
in reactive routing is improved since routing
information is only exchanged when needed. A
node can be certain to use the most recent routing
information, which leads to a higher transmis-
sion success than with proactive routing. Among
various reactive ad hoc routing protocols, AODV
and DSR are the two that gain most attention.
Compared to AODV, DSR has higher through-
put and considerably smaller control overhead.
DSR handles the quick changes of high mobility
better than AODV. On the other hand, this intro-
duces a higher delay due to the route discovery
phase (Böhm, 2007). Other flat, reactive routing
algorithms for mobile ad hoc networks include
Temporally Ordered Routing Algorithm (TORA)
and Associativity Based Routing Algorithm
(ABR), etc. When reactive routing protocols are
employed in inter-vehicle communication, some
real-time applications require flooding short warn-
ing messages between vehicles. In these cases,
flooding the network with route request messages
MAC and Routing Protocols for Vehicle to Vehicle Communications
is only meaningful if the actual data to be sent
is considerably larger than the request message.
Otherwise, it is more bandwidth efficient to flood
the data itself (Böhm, 2007).
Pure proactive and reactive protocols work
well for a small number of nodes. As the number
of nodes increases in the networks and clusters
of nodes are formed, hierarchical protocols can
be used for better performance. In hierarchical
routing, the network is divided into clusters of
nodes. Routing decisions are made by certain
nodes (cluster heads), while the rest of the nodes
in a cluster receive or relay data via the cluster
head. Hierarchical protocols attempt to adopt the
advantages of both proactive and reactive proto-
cols. Therefore, two different routing protocols
are used, one between clusters and another within
each cluster. Information on topology changes will
only be propagated among the nodes in the same
cluster. Nodes in other clusters only need to know
the route to the destination cluster. Clusters can be
combined into super clusters, forming a hierarchy
of clusters. One or more nodes in a cluster may
serve as a router for traffic passing in and out of
that cluster. Examples of hierarchical protocols
include Cluster Based Routing Protocol (CBRP),
Core Extraction Distributed Ad-hoc Routing
(CEDAR) and Zone Routing Protocol (ZRP).
In hierarchical routing, both routing table size
and the size of control packets for route discovery
or maintenance are reduced considerably. Better
scalability is achieved as only the cluster heads
need to concern about nodes entering or leaving
the network. Hierarchical routing techniques also
lead to shorter response time as the flooding of
routing information is restricted to the cluster
heads and routes are discovered more quickly.
On the other hand, in highly mobile vehicular-
to-vehicle networks, the hierarchical partition
of the network needs to be frequently updated
and renegotiated, which complicates the routing
process and adds delay.
Position-Based Routing
Flooding the whole network with route requests
or routing updates is a waste of resources. If
the position of the destination node is known,
integrating this position information in the rout-
ing process would enhance the routing protocol
performance. A number of position-based (other
frequently used names are georouting, location-
based or geographical routing) routing protocols
have been proposed in MANET and for vehicle-
to-vehicle networks. Vehicles equipped with a
GPS (Global Positioning System) receiver have
access to their coordinates in a globally known
reference system. But it is not always possible
to reach a sufficient number of GPS satellites to
produce a correct position, for example, in tunnels
or in urban areas. Other ways of positioning can
be used such as triangulation or local reference
systems. Position-based routing consists of two
stages. First a location service is responsible for
providing a source with the position of the desired
destination. In a vehicular network, information
about a vehicle’s speed and direction (maybe
even in combination with knowledge of speed
limits, road type and other map-based informa-
tion (Böhm, 2007)) could be included in location
services. Popular location services in position-
based routing protocols are Distance Routing
Effect Algorithm for Mobility (DREAM) and Grid
Location Service (GLS). The second stage is the
forwarding phase, where the route from source
to destination is determined. There are a number
of position-based forwarding strategies, includ-
ing Greedy Packet Forwarding (GPF), Greedy
Parameter Stateless Routing Protocol (GPSR)
and Restricted Directional Flooding (RDF). In
restricted flooding, the message is flooded to all
nodes within a given region. There are different
ways to determine such a region: the DREAM
protocol, which is mentioned before as a location
service, and Location Aided Routing (LAR).
However, none of the strategies scales very well
115
 MAC and Routing Protocols for Vehicle to Vehicle Communications
to large networks but is very robust against node
failure and position inaccuracies.
Many of the position-based algorithms are ap-
plicable to the requirements of inter-vehicle com-
munication (Böhm, 2007). It is usually sufficient
for a node to know its own position, the position
of its immediate neighbors and the position of the
destination node acquired through the location
service. Therefore, no routes have to be established
or maintained and most of the routing overhead
is avoided. This makes most position-based ap-
proaches scale well and adapt quickly to a highly
mobile network. In a vehicular network with very
high node mobility, position-based routing can
avoid extensive flooding of routing information,
especially when traffic patterns, speed etc. are
known, which can be used to determine the re-
stricted flooding areas. The main disadvantage is
that the position of the destination node has to be
known and location information has to be spread
through the network. Maintaining location service
adds overhead to the protocols, in compensation
that low routing overhead is achieved. The design
of location update schemes is often more difficult
than the routing decisions themselves. Apart from
that, the position offered by a location informa-
tion receiver, such as GPS can be inaccurate or
unavailable due to obstacles. To work properly,
all position-based algorithms rely on correct posi-
tions. Besides, position-based routing protocols
have problems handling “topology holes”. In areas
with scarce traffic or in a situation where only
few of the vehicles are equipped with communi-
cation devices, this might become a problem for
inter-vehicle communication scenarios. In high
density networks, geographic forwarding works
well. But when there are frequent topology holes,
the stateless geographic forwarding works poor,
since it will initiate a routing recovery process
for packet reaching, which seriously degrades the
packet routing performance.
To cope with topology holes, a context aware
protocol, Context Assisted Routing (CAR), is
proposed by Dumitrescu and Guo (2005). CAR
116
can assist the drivers to make efficient route plan-
ning decisions and better balance traffic around
highly populated urban areas. It is an extension to
both location services and the greedy forwarding
strategy by introducing context-assisted graphs.
The method collects, disseminates and stores
traffic context data including road conditions,
vehicle density, average vehicle speed and roadside
facilities, and uses these information to aid the
design at the network layer by using a weighted
oriented graph.
Spatially Aware Routing (SAR) is another
proposed routing algorithm to overcome the
problem of topology holes. As the occurrence of
topology holes is often closely related to the road
infrastructure, relevant spatial information is
extracted from a Geographic Information System
(GIS) to build a graph of intersections and road
segments. In the graph, vertices represent the road
intersections and edges are the road segments.
Only major roads and intersections can be used
depending on the memory capability. But the
information can be extended to include additional
data on the length of the road segment, speed
limits, number of lanes etc., which can be used
to assign weights to the edges of the graph. The
sender uses the graph to find the shortest path to
the destination and the packet forwarding is done
via vehicles along the path. In case no suitable
intermediate node can be found, an alternative
graph is computed or the packet is suspended in
a buffer until a forwarding vehicles comes along.
The authors compared the SAR protocol and a
version with suspension buffer (called SARB) to
standard GPSR and showed that SAR gives the
shortest delay and highest packet delivery ratio
(Tian, Han & Rothennel, 2003).
Position-based Quality of Service (QoS)
Routing is another example of real-world routing
algorithm that can be used for vehicle-to-vehicle
communication. It uses a grid-based hierarchy
with cluster heads maintaining not only location
information, but also information on available
bandwidth and the link state in the network. Up-
MAC and Routing Protocols for Vehicle to Vehicle Communications
dates are sent to the cluster head periodically or
on demand within each cell of the grid and then
forwarded to other cluster heads in the network.
Thereby, a global knowledge is achieved and a
cluster head can provide the sender with infor-
mation about both the position of destination and
possibility of setting up a QoS link. This routing
algorithm requires the support from other layers,
such as Medium Access Control layer, to give
the QoS guarantees (Erbas, Garcia & Jobmann,
2004).
Multicast
In many applications, the information sent out
from one source often concerns not only one
destination node but all nodes in a certain alert
area, for example, the vehicles alert messages.
Multicast is therefore an important feature to
incorporate in routing protocol design. Current
multicast protocols that can be used in vehicle to
vehicle networks include: Position-based Multi-
cast (LBM), GeoGRID, Unicast Routing with
Area Delivery and Inter-Vehicle Geocast (Böhm,
2007).
LBM is based on the concepts of position-based
routing and directional flooding. It defines a des-
tination region and a flooding zone and discards
packets outside this zone. The flooding zone is
a rectangle including at least the sender and the
destination area. Another approach is to only allow
intermediate nodes that are closer than a certain
distance threshold from the destination area. A
node calculates its geographical distance from the
center of the destination region and determines if it
is to forward the packet or to drop it. The distance
is decreased with each hop. GeoGRID enhances
the LBM approach by dividing the forwarding
region into grids where only preselected gateway
nodes are relaying the packets until the destina-
tion region is reached. In Unicast Routing with
Area Delivery protocol, any unicast georouting
protocol of choice is used to reach the destination
area and simply flood the region.
The Inter-Vehicle Geocast Protocol (IVG) is
developed for relaying emergency messages in a
highway accident situation. A broken down ve-
hicle sends out an alarm to approaching vehicles.
The packets are only relayed among the vehicles
approaching the accident site, not by vehicles
driving away from the site, so the packets are not
spread to vehicles that do not need the informa-
tion. A node near the accident site waits a given
amount of time until it broadcasts the message
in its transmission range. If during this time, it
receives the same message from a node behind
it, it knows that another node has taken over the
role as relay node and thus drops the packet. Oth-
erwise, the node rebroadcasts the message with
a certain time interval. There is no time-to-live
limit to prevent infinite broadcasting in order to
keep messages alive in sparse network such as a
highway at night time.
ConcluSIOion
In this section, we present an overview of routing
techniques in vehicle ad hoc networks. While
a number of schemes for handling information
dissemination in vehicle-to-vehicle networks
exist, only few may be able to handle the require-
ments of safety applications, due to overhead in
discovering and/or maintaining routes in highly
mobile networks of uncoordinated vehicles. In a
vehicular ad hoc network, the routes are required
to be updated very frequently, which adds control
overhead and influences the total throughput
over the network. Proactive routing solutions for
mobile ad hoc networks are not really suitable
for inter-vehicle communication because of the
quickly changing topology. In reactive routing
methods, the control overhead is smaller than
proactive routing, but the communication delay
is usually larger.
As the number of nodes increases in the net-
works, hierarchical protocols can be used for better
performance, where both routing table size and
117
 MAC and Routing Protocols for Vehicle to Vehicle Communications
the size of control packets are reduced consider-
ably. Better scalability is achieved as only the
cluster heads need to concern about nodes enter-
ing or leaving the network. Hierarchical routing
techniques also lead to shorter response time as
the flooding of routing information is restricted
to the cluster heads. However, in highly mobile
vehicular-to-vehicle networks, the hierarchical
partition of the network needs to be frequently
updated and renegotiated, which makes the rout-
ing process complicated.
Position-based algorithms are promising ap-
proaches for routing in vehicular environments.
In position-based algorithms, a node only needs
to know the position of its own, its immediate
neighbors and the destination node. Therefore,
no routes have to be established or maintained
and most of the routing overhead is avoided. This
makes most position-based routing approaches
adapt quickly to a highly mobile network and scale
well. Also, in a vehicular network with very high
node mobility, position-based routing can avoid
extensive flooding of routing information.
The main disadvantage of position-based
routing is that the destination node’s position in-
formation has to be spread through the network.
Maintaining a position service adds overhead,
which takes away much of the gain from low
routing overhead. The design of location update
schemes is often more difficult than the routing
decisions themselves. Besides, the position of-
fered by e.g. a GPS receiver can be inaccurate
or unavailable due to obstacles. Position-based
routing protocols have problems handling topol-
ogy holes. CAR and SAR are recently proposed
routing algorithms to overcome the problem of
topology holes. Overall, routing for vehicle safety
communications remains a challenging topic.
REFERENCES
Abramson, N. (1970). The ALOHA system – An-
other alternative for computer communications,
AFIPS Conference (pp. 295-298).
118
Bai, F., Sadagopan, N., & Helmy, A. (2003, April).
Important: A framework to systematically ana-
lyze the Impact of Mobility on Performance of
Routing Protocols for Ad-hoc Networks. IEEE
INFOCOM, 2, 825-835.
Baranowski, S., Lienard, M., & Degauque, P.
(1991, February). Beacon-vehicle link in the 1-10
GHz frequency range. DRIVE Conference (pp.
194-217).
Böhm, A. (2007). State of the art on network layer
aspects for inter-vehicle communication (Tech.
Rep. IDE 0748).
Borgonovo, F., Capone, A., Cesana, M., & Fratta,
L. (2004, July) ADHOC MAC: A new MAC ar-
chitecture for ad hoc networks providing efficient
and reliable point-to-point and broadcast services.
ACM Wireless Networks, 10, 359-366.
Chennikara-Varghese, J., Chen, W., Altintas, O.,
& Cai, S. (2006, July). Survey of routing protocols
for inter-vehicle communications. IEEE Workshop
on Mobile and Ubiquitous Systems Conference
(pp. 1-5).
Crowther, W., Rettberg, R., Walden, D., Ornstein,
S., & Heart, F. (1973). A system for broadcast
communication: Reservation-ALOHA. Hawaii
International Conference on System Sciences.
Dumitrescu, V., & Guo, J. (2005, June). Context
assisted routing protocols for inter-vehicle wireless
communication. IEEE Symposium on Intelligent
Vehicles (pp. 594-600).
Eichler, S. (2007, October). Performance Evalua-
tion of the IEEE 802.11p WAVE Communication
Standard. IEEE Vehicular Technology Conference
(pp. 2199-2203).
Erbas, F., Garcia, J.-E., & Jobmann, K. (2004).
Position-based QoS routing in mobile ad-hoc net-
works: problem statement and a novel approach.
IEEE Performance, Computing, and Communica-
tions Conference (pp. 619- 623).
MAC and Routing Protocols for Vehicle to Vehicle Communications

Grabow, W., & Schrei, A. (1992, October). A study wireless link throughput, range, and energy ef-
on diversity techniques in a 5.8 GHz microwave ficiency. IEEE INFOCOM, 2, 564-571.
link for traffic management systems. IEEE Sym-
Metcalfe, R., & Boggs, R. (1976, July). Ethernet:
posium on Personal, Indoor and Mobile Radio
Distributed packet switching for local computer
Communications (pp. 556-560).
networks, Communication of the ACM, 19, 395-
Haas, Z., & Deng, (2002, June). Dual busy tone 404.
multiple access (DBTMA)-a multiple access con-
Pfeiffer, G., & Schrei, A. (1994, June). Investi-
trol scheme for ad hoc networks. IEEE Transac-
gation on propagation characteristics of a short
tions on communications, 50, 975-985.
range radio link with statistical distributed reflec-
Jurdak, R., Lopes, C., & Baldi, P. (2004). A tors. IEEE Vehicular Technology Conference, 3,
survey, classification and comparative analysis 1728-1732.
of medium access control protocols for wireless
Roberts, L. (1975, April). ALOHA packet systems
networks. IEEE Communication Surveys and
with and without slots and capture, Computer
Tutorials, 6.
Communications Review, 5.
Karn, P. (1990, September). MACA – A new chan-
Singh, J. P., Bambos, N., Srinivasan, B., & Clawin,
nel access method for packet radio, ARRL/CRRL
D. (2002, December). Wireless LAN performance
Amateur Radio Computer Networking Conference
under varied stress conditions in vehicular traffic
(pp. 134-140).
scenarios. IEEE Vehicular Technology Confer-
Kleinrock, L., & Tobagi, F. (1975, December). ence, 2, 743-747.
Packet switching in radio channels: Part I – Carrier
Tian, J., Han, L., & Rothennel, K, (2003, Octo-
sense multiple-access modes and their through-
ber). Spatially aware packet routing for mobile
put-delay characteristics. IEEE Transactions on
Ad Hoc inter-vehicle radio networks. IEEE
Communications, 23, 1400-1416.
Intelligent Transportation Systems Conference,
Ko, Y., Shankarkumar, V., & Vaidya, N. (2000). 2, 1546- 1551.
Medium access control protocols using directional
Wang, B., Sen, I., & Matolak, D. (2007, October).
antennas in ad hoc networks. IEEE INFOCOM
Performance evaluation of 802.16e in vehicle to
(pp. 13-21).
vehicle channels. IEEE Vehicular Technology
Lettieri, P., & Srivastava, M. B. (1998, April). Conference (pp. 1406-1410).
Adaptive frame length control for improving

119
120

Chapter VII
Inter-Vehicular
Communications Using Wireless
Ad Hoc Networks

Raúl Aquino-Santos Omar Álvarez-Cardenas

University of Colima, Mexico University of Colima, Mexico

Víctor Rangel-Licea Arthur Edwards-Block

National Autonomous University of Mexico, University of Colima, Mexico
Mexico
Margarita G. Mayoral-Baldivia
Miguel A. García-Ruiz University of Colima, Mexico
University of Colima, Mexico
Sara Sandoval-Carrillo
Apolinar González-Potes University of Colima, Mexico
University of Colima, Mexico

ABSTtract

This chapter proposes a new routing algorithm that allows communication in vehicular ad hoc networks.
In vehicular ad hoc networks, the transmitter node cannot determine the immediate future position of the
receiving node beforehand. Furthermore, rapid topological changes and limited bandwidth compound
the difficulties nodes experience when attempting to exchange position information. The authors first
validate their algorithm in a small-scale network with test bed results. Then, for large-scale networks,
they compare their protocol with the models of two prominent reactive routing algorithms: Ad-Hoc
On-Demand Distance Vector and Dynamic Source Routing on a multi-lane circular dual motorway,
representative of motorway driving. Then the authors compare their algorithm with motorway vehicu-
lar mobility, a location-based routing algorithm, on a multi-lane circular motorway. This chapter then
provides motorway vehicular mobility results of a microscopic traffic model developed in OPNET, which
the authors use to evaluate the performance of each protocol in terms of: Route Discovery Time, End
to End Delay, Routing Overhead, Overhead, Routing Load, and Delivery Ratio.

Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.
Inter-Vehicular Communications Using Wireless Ad Hoc Networks
INTRODUCTION
How to best optimize traffic flow is one of the
primary challenges of specialists studying con-
gestion and safety on streets and motorways
because of the economic, health, and safety issues
related to inefficient traffic circulation. Propos-
als to mitigate traffic congestion, caused in part
by inefficient traffic flow, have often included
expensive construction projects. These projects,
however, have had only limited success. In the
United States, for highway travel congestion of
personal vehicles, the Transportation Statistics
Annual Report (2006) states that “highway travel
times increased between 1993 and 2003 in all but
3 of the 85 urban areas (98 percent)”, and “it took
37 percent longer, on average, in 2003 to make a
peak period trip (from 6 to 9 a.m. and 4 to 7 p.m.)”.
Additionally, this same report reveals that “there
were nearly 45,000 fatalities in transportation
accidents in the United States in 2004, of which
95 percent involved highway motor vehicles”.
Furthermore, although the number of fatalities
slightly decreased, “in 2005, 43,443 motorists and
non-motorists were killed in crashes involving
motor vehicles, up 1% compared with 2004, and
about 2.7 million people were injured”. Finally,
the report mentions that “there were 1.47 fatalities
per 100 million vehicle-miles of highway travel
in 2005”.
Although passive safety systems such as seat
belts and air bags have been used to significantly
reduce the total number of major injuries and
deaths due to motor vehicle accidents, they do
nothing to actually improve traffic flow or lower
the actual number of automobile collisions. In
order to reduce the number of vehicular accidents,
computer and network experts propose active safe-
ty systems, including Intelligent Transportation
Systems (ITS) that are based on Inter-vehicular
Communication (IVC) and Vehicle-to-Roadside
Communication (VRC). Presently, technologies
related to these architectures and their related
technologies may, in the future, more efficiently
administer traffic flow, which, in turn, can have
important safety, ecological, and economic
ramifications.
Active vehicular systems employ wireless
ad hoc networks and Global Positioning System
(GPS) to determine and maintain inter-vehicular
distances to insure the one-hop and multi-hop
communications network needed to maintain ve-
hicle spacing. Location-based routing algorithms
may help in the development of Vehicular Ad Hoc
Networks (VANETs) because their flexibility and
efficiency provide the ad hoc architecture neces-
sary for inter-vehicular communication. Although
several location-based algorithms already exist,
including Grid Location Service (GLS), Location
Aided Routing (LAR), Greedy Perimeter Stateless
Routing (GPSR), and Distance Routing Effect
Algorithm for Mobility (DREAM), to name just a
few, we propose a Location-Based Routing Algo-
rithm with Cluster-Based Flooding (LORA-CBF)
as an option for present and future automotive
applications (Santos et al., 2005).
Ad Hoc routing protocols have the design goals
of network optimality, simplicity, low overhead,
robustness, stability, rapid convergence, and
flexibility. However, since mobile nodes suffer
from significantly less available power, process-
ing speed, and memory, low overhead becomes
even more important than in conventional fixed
networks. The high mobility present in vehicle-
to-vehicle communication also places great
importance on rapid convergence. Therefore,
it is imperative that ad hoc routing protocols
effectively compensate for any inherent delays
in the underlying technology, adapt to varying
degrees of mobility, and be sufficiently robust to
deal with potential transmission loss due to drop
out. Additionally, such protocols must more ef-
fectively route packets than traditional network
algorithms in order to effectively compensate for
limited bandwidth resources.
Several routing algorithms for ad hoc networks
have emerged recently to address difficulties
related to unicast routing. Such algorithms can
121
 Inter-Vehicular Communications Using Wireless Ad Hoc Networks
be categorized as either proactive or reactive,
depending on their route discovery mechanism.
This chapter presents a set of performance predica-
tions for ad hoc routing protocols used in highly
mobile vehicle-to-vehicle multi-hop networks as
part of the extensive research and development
effort which will be undertaken in the next decade
to incorporate wireless ad hoc networks in the
automobile industry.
In proactive algorithms, each node continu-
ously updates the routes to all other nodes in the
network by periodically exchanging control mes-
sages. Consequently, the route is immediately
available when a node needs to send a packet to
any other node in the network. The main advan-
tage of proactive algorithms is that they have a
shorter delay. Examples of proactive algorithms
include Optimized Link State Routing Protocol
(OLSR) and Topology Dissemination Based on
Reverse-Path Forwarding (TBRPF).
The disadvantage of the OLSR and TBRPF
protocols, however, is their link state routing
dissemination strategy. Recognized link changes
cause nodes to flood control packets across the
entire network, which taxes network resources
(Zou, Ramamurthy, Magliveras, 2002). Vehicular
ad hoc networks consist of many highly mobile
nodes whose mobility is defined, in part, by the
physical limitations of the motorway and the
actual relative positions of the vehicles circulat-
ing on it, as well as the significant number of
vehicles simultaneously entering and exiting the
motorway. As a consequence of these factors, the
dissemination of routing information increases
the demand for often limited bandwidth resources
and computation time.
Conversely, reactive algorithm nodes dis-
cover routes on demand and maintain smaller
active route tables. Thus, a route is discovered
whenever a source node needs to communicate
with a destination node for which a route has not
already been established. Discovery is based on
flooding, which can be total, as in AODV and
DSR, or limited, as in OLSR and TBRPF. In
122
these scenarios, source nodes broadcast a route
request message to all immediate neighbors, and
these in turn, re-broadcast the route request to
their neighbors. When the route request reaches
either the destination or a node that has a valid
route to the destination, a route reply message
is generated and transmitted back to the source.
Therefore, as soon as the source receives the route
reply, a route is created from the source to the
destination. The advantage of reactive algorithms
is that there are no control messages for non-ac-
tive routes. The major drawback is the latency
in establishing transmission routes. Examples
of reactive algorithms include AODV (Perkins,
Belding-Royer & Das, 2003) and DSR (Johnson,
Maltz, & Hu, 2007).
Previously, proactive algorithms were not con-
sidered suitable for highly mobile environments
because they tend to have poor route convergence
and low communication throughput (Royer &
Toh, 1999).
Some Cluster-Based Flooding strategies for
routing in wireless ad hoc networks have been
reported in literature (Boris and Arkady, 1999;
Krishna et al., 1997; Bevas et al., 1997; Raghu-
pathy et al., 1998; Ching-Chuan et al., 1997). The
main contribution of this work is the re-broadcast
and gateway selection mechanisms. The cluster
formation in Boris and Arkady (1999) is based on
the Link Cluster Algorithm (LCA) (Baker et al.,
1984) and the algorithm is based on Link State
Routing protocol, where all nodes in the cluster
are expected to acknowledge the Link State Up-
date (LSU). If one of the nodes does not send an
acknowledgement, the cluster head retransmits
the LSU to that particular node. Flooding is
transmitted from the source to a destination via
cluster heads and gateways. In (Krishna et al.,
1997) and (Bevas et al., 1997), the re-broadcast
is handled by the boundary nodes. Nodes other
than boundary nodes just listen and update their
tables. In (Raghupathy et al., 1998), there are two
types of routing strategies: Optimal Spine Rout-
ing (OSR) and Partial-knowledge Spine Routing
Inter-Vehicular Communications Using Wireless Ad Hoc Networks
(PSR). OSR uses full and up-to-date knowledge
of the network topology, permitting the source
to determine the route to the destination. On the
other hand, PSR uses partial knowledge of the
network topology and takes a greedy approach
to compute the shortest path from the source to
the destination. In (Ching-Chuan et al., 1997),
the authors use a cluster head controlled token
protocol (like polling) to allocate the channel
among competing nodes. We have implemented
a re-broadcast strategy where only gateways
belonging to different cluster heads re-broadcast
the location request packets, resulting in improved
routing overhead.
Inter-Vhicular and
Vehicle-to-Roadside Communications
The interest in inter-vehicular and vehicle-
to-roadside communication has significantly
increased over the last decade, in part, because
of the proliferation of wireless networks. Most
research in this area has concentrated on vehicle-
to-roadside communication, also called beacon-
vehicle communication (Rokitansky & Wietfeld,
1993), (Brasche, Rokitansky & Wietfeld, 1994),
(Wietfeld & Rokitansky, 1994), (Rokitansky &
Wietfeld, 1995), (Wietfeld & Rokitansky, 1995),
in which vehicles share the medium by accessing
different time slots (Time Division Multiple Ac-
cess, TDMA), the beacon (down-link direction),
and the vehicles (up-link direction). The beacon
arranges up-link time slots (so called windows
allocations) for the vehicles; the vehicles are not
allowed to access the medium without a window
allocation sent to them by the beacon. The beacon,
as the primary station, offers two different types
of windows to the vehicles: public and private
windows. A public window is a time slot that
can be accessed by any vehicle within the com-
munication zone. A private window allocation
reserves a time slot for one specific vehicle, thus
protecting it against data collisions.
A typical communication process between
a beacon and a vehicle can be divided into two
phases: the connection establishment phase and
the transaction phase. When a vehicle enters the
beacon communication zone, the vehicle address
is not known to the beacon, so the beacon periodi-
cally broadcasts a request for identification and
offers a public up-link window before establishing
the actual connection. When a specific vehicle
responds to an identification request, it records
the address in the transmitter node’s table and
individual addressing becomes possible. The
beacon then opens a private uplink window to
a specific vehicle for data exchange during the
transaction phase.
Other mechanisms to communicate vehicles
in vehicle-to-roadside communication have
been proposed by (Kwak & Jae, 2004), (Dobias
& Grabow, 1994), (Zandbergen & Van der Ree,
1992), (Abdulhamid, Abdel-Raheem & Tepe,
2007), (Matthaiou, Laurenson, & Cheng-Xiang,
2008), (Cottingham, Wassell & Harle, 2007),
(Mussa & Upchurch, 2000), (Bantli, Ring & Goff,
1997), and (Dickey, Huang & Guan, 2007).
Some applications for vehicle-to-roadside
communication, including Automatic Payment,
Route Guidance, Cooperative Driving, and Park-
ing Management have been developed to function
in limited communication zones of less than 60
meters. However, the IEEE 802.11 Standard has
led to increased research in the areas of wireless
ad hoc networks and location-based routing algo-
rithms, (Morris et. al., 2000), (Da Chen, Kung, &
Vlah, 2001), (Füßler, et. al., 2003), (Lochert, et.
al., 2003), (Kosh, Schwingenschlögl, & Ai, 2002).
Applications for inter-vehicular communication
include Intelligent Cruise Control, Intelligent
Maneuvering Control, Lane Access, and Emer-
gency Warning, among others. In (Morris et. al.,
2000), the authors propose using Grid (Li, et.
al., 2000), a geographic forwarding and scalable
distributed location service, to route packets from
car to car without flooding the network. The au-
123
 Inter-Vehicular Communications Using Wireless Ad Hoc Networks

thors in (Da Chen, Kung, & Vlah, 2001) propose Figure 1. Inter-vehicular communication system
relaying messages in low traffic densities, based using wireless ad hoc networks
on a microscopic traffic simulator that produces
accurate movement traces of vehicles traveling
on a highway, and a network simulator to model
the exchange of messages among the vehicles.
Da Chen et. al., employ a straight bidirectional
highway segment of one or more lanes. The mes-
sages are propagated greedily each time step by
hopping to the neighbor closest to the destination.
The authors in (Füßler, et. al., 2003), compare a
topology-based approach and a location-based
routing scheme. The authors chose GPSR (Karp &
Kung, 2000) as the location-based routing scheme
and DSR (Johnson, Maltz, & Hu, 2007) as the
topology-based approach. The simulator used in
(Füßler, et. al., 2003) is called FARSI, which is
nologies. The major goals are to provide increased
a macroscopic traffic model. In (Lochert, et. al.,
automotive safety, achieve smoother traffic flow,
2003), the authors compare two topology-based
and improve passenger convenience by provid-
routing approaches, DSR and AODV (Perkins,
ing information and entertainment. In order to
Belding-Royer & Das, 2003), versus one position-
avoid communication costs and guarantee the
based routing scheme, GPSR, in an urban envi-
low delays required to exchange safety-related
ronment. Finally, in (Kosh, Schwingenschlögl, &
data between cars, inter-vehicular communica-
Ai, 2002), the authors employ a geocast routing
tion (IVC) systems, based on wireless ad hoc
protocol that is based on AODV.
networks, represent a promising solution for fu-
In inter-vehicular communication, vehicles are
equipped with on-board computers that function ture road communication scenarios. IVC allows
as nodes in a wireless network, allowing them to vehicles to organize themselves locally in ad hoc
contact other similarity equipped vehicles in their networks without any pre-installed infrastructure.
vicinity. By exchanging information, vehicles Communication in future IVC systems will not
can obtain information about local traffic condi- be restricted to neighboring vehicles traveling
tions, which improves traffic control and provides within a specific radio transmission range. As in
greater driver safety and comfort. typical wireless scenarios, the IVC system will
In this chapter, we will focus on inter-vehicu- provide multi-hop communication capabilities by
lar communication because vehicle-to-roadside using “relay” vehicles that are traveling between
communication has been already proposed for the sender and receiver. Figure 1 illustrates this
standardization in Europe (CEN TC 278 WG 9) basic idea. In this particular example, the source
and North America (IVHS). vehicle is still able to communicate with the desti-
nation vehicle, although the destination vehicle is
Issues Concerning Inter-Vhicular not in source vehicle’s immediate communication
Communication Using Wireless Ad range. Vehicles between the source-destination
Hoc Networks act as intermediate vehicles, relaying data to the
receiver. As a result, the multi-hop capability of
Future developments in automobile manufactur- the IVC system significantly increases the virtual
ing will also include new communication tech-

124
Inter-Vehicular Communications Using Wireless Ad Hoc Networks
communication range, as it enables communica-
tion with more distant vehicles.
LOCATION ROUTING ALGORITHM
WITH CLUSTER-BASED FLOODING
(LORA-CBF)
In this section, we present a Location Routing
Algorithm with Cluster-Based Flooding (LORA-
CBF), which is formed with one cluster head, zero
or more members in every cluster, and one or
more gateways to communicate with other cluster
heads. Each cluster head maintains a “Cluster
Table”. A “Cluster Table” is defined as a table that
contains the addresses and geographic position
of the member and gateway nodes.
When a source attempts to send data to a
destination, it first checks its routing table to
determine if it knows the location of the destina-
Figure 2. Flow diagram for LORA-CBF
tion. If it does, it sends the packet to the closest
neighbor to the destination (Figure 2). Otherwise,
the source stores the data packet in its buffer,
starts a timer, and broadcasts Location Request
(LREQ) packets. Only gateways and cluster heads
can retransmit the LREQ packet. Gateways only
retransmit a packet from one gateway to another
in order to minimize unnecessary retransmissions,
and only if the gateway belongs to a different
cluster head.
Upon receiving a location request, each cluster
head confirms that the destination is a member
of its cluster. Success triggers a Location Reply
(LREP) packet that is returned to the sender us-
ing geographic routing. This is possible because
each node knows the position of the source and
the closest neighbor, based on the information
received from the LREQ and the Simple Location
Service (SLS). Failure triggers retransmissions by
the cluster head to adjacent cluster-heads (Reac-
125
 Inter-Vehicular Communications Using Wireless Ad Hoc Networks
tive Location Service, RLS) and the destination
address is recorded in the packet. Cluster heads
and gateways, therefore, discard request packets
they have previously seen.
Once the source receives the destination’s
location, it retrieves the data packet from its buf-
fer and forwards it to the neighbor closest to the
destination.
Basically, the algorithm consists of five
stages:
1. Cluster formation
2. Location discovery (LREQ and LREP)
3. Routing of data packets
4. Maintenance of location information
5. Short-term geographic positioning predictive
algorithm
Cluster Formation
All nodes maintain neighbor information in their
respective tables to enable cluster formation and
maintenance.
Let t be the period of time between the Hello
broadcasts. When a node first switches on, it first
listens to Hello packets on the broadcast chan-
nel. If any other node on the broadcast channel is
already advertising itself as a cluster head (status
of node = cluster-head), the new node saves the
heard cluster-head ID in its cluster-head ID field
and changes its status to member. At any point in
time, a node in the mobile network can associate
itself with a cluster head. The cluster heads are
identified by the cluster-head ID. Otherwise, the
new node becomes a cluster head. Cluster heads
are responsible for their clusters and periodically
send Hello Messages.
When a member of a cluster receives a Hello
message, it registers the cluster head and responds
with a reply Hello message. The cluster head then
updates the cluster table with the address and
position (longitude and latitude) of every cluster
member.
When a member receives a Hello packet from
126
a different cluster head, it must first register it, but
the member does not actually modify its cluster-
head ID until the expiration time for the field has
expired. Before the member rebroadcasts the new
information, it must change its status to a gateway.
After receiving the Hello packet, the cluster heads
update their cluster tables with the new gateway
information.
If the source broadcasts a message to the
destination, it must first check its routing table to
determine if it has a “fresh” route to the destination.
If it does, it first seeks its cluster table to identify
the closest neighbor to the destination. Otherwise,
it starts the location discovery process.
Location Discovery Process
When the source of the data packet transmits to
a destination that is not included in its routing
table or if its route has expired, it first places the
data packet in its buffer and broadcasts a Loca-
tion Request (LREQ) packet (Reactive Location
Service, RLS).
When a cluster head receives a LREQ packet,
it checks the packet identification field to deter-
mine if it has previously seen the LREQ packet.
Previously seen packets are discarded, but if the
destination node is a member of the cluster head,
it unicasts the Location Reply (LREP) packet to
the source node.
If the destination node, however, is not a member
of the cluster head, it first records the LREQ packet
address in its routing table and rebroadcasts the
LREQ packet to neighboring cluster heads.
Each cluster head node forwards the packet
once. The packets are broadcasted only to the
neighboring cluster head by means of an omni-di-
rectional antenna that routes them via the gateway
nodes. Gateways only retransmit a packet from
one gateway to another to minimize unnecessary
retransmissions, and only if the gateway belongs
to a different cluster head. When the cluster head
destination receives the LREQ packet, it records
the source address and location. From this, the
Inter-Vehicular Communications Using Wireless Ad Hoc Networks
destination cluster head locates the source node.
The destination then sends a LREP message back
to the source via its closest neighbor.
Finally, the packet reaches the source node that
originated the request packet. If the source node
does not receive a LREP after transmitting a LREQ
for a set period of time, it goes into an exponential
back off before re-transmitting the LREQ. Hence,
only one packet is transmitted back to the source
node. The reply packet does not have to maintain
a source/destination routing path, and the path is
determined from the location information given
by the source node. Importantly, the path followed
by the LREQ may be different from that traversed
by the LREP.
Routing of Data Packets
The actual data packet routing is then based on
source, destination, and neighbor location.
Since the protocol is not based on source rout-
ing, packets travel the path from the source to the
destination, based on their relative locations. The
packets find paths to their destinations individually
each time they transmit data between the source and
the destination. Packets are transmitted between
nodes based on their knowledge of the position
information in their tables. Moreover, since the
transmission is in the direction of the destination
node, the path found will be shorter than other rout-
ing mechanisms (non-positional-based). In non-po-
sitional-based routing strategies, the shortest path is
measured in hops, meaning the path discovered may
not be the shortest. However, the path found using
location information will be significantly shorter.
If the source of the data packet does not receive the
acknowledgement packet before its timer expires,
it will retransmit the data packet again. This might
occur particularly during packet loss because of
drop-out or network disconnection.
LORA-CBF uses the Most Forward within
Radius (MFR) forwarding strategy. In MFR, the
packet is routed via the neighbors that most reduce
the distance to the destination. The advantage of this
forwarding strategy is that it decreases the prob-
ability of collision and end-to-end delay between
the source and the destination.
Maintenance of Location Information
The LORA-CBF algorithm is suitable for networks
with very fast, mobile nodes because it maintains
and updates the location information of the source
and destination every time the pairs send or receive
data and acknowledgment packets. The source up-
dates its location information before sending each
data packet. When the destination receives the data
packet, its location information is updated and an
acknowledgment packet is sent to the source.
Sort-Term Geographic Positioning
Predictive Algorithm
In highly mobile environments, precise knowledge
of neighbor positions, to a great degree, determines
the routing efficiency of any algorithm. LORA-CBF
predicts the immediate future location of every
neighbor node, based on its short-term geographic
positioning predictive algorithm. After predicting
the position of all neighbor nodes, LORA-CBF
sends the packet to the neighbor node with shorter
distance to the destination node (MFR).
Mobility and contention of wireless media may
cause packet loss during transmission, which is
a very important phenomenon to consider when
developing predictive algorithms. Our algorithm
also predicts the future probable location of a
node, despite data stream time gaps caused by
collisions.
Assuming the data follows a linear trend,
our short-term geographic positioning predictive
algorithm attempts to extrapolate the position of
the next hop k ahead in time, according to the fol-
lowing equation:
Pj + k = Pj + ∆P * ℮ (1)
127
 Inter-Vehicular Communications Using Wireless Ad Hoc Networks
Where:
Pj + k future position of the next hop
Pj current position of the next hop
∆P interval between current position and
previous position of the next hop
℮ factor indicating the gap between packets
received
Short-term geographic positioning predictive
algorithms are useful in highly mobile conten-
tion-based networks. LORA-CBF periodically
broadcasts Hello messages to locate nodes en-
tering or leaving the ad hoc network. Location
information received from a Hello message helps
neighbor nodes calculate the future nodal position
of the sending node throughout the transmis-
sion range by calculating its immediate future
position. When a node receives a packet, before
forwarding it to a particular destination, it first
checks its routing table to determine if it possesses
the location information of the destination node.
If it does, it triggers the short-term geographic
positioning predictive algorithm to calculate the
future position of the destination. If the sender
node can directly reach the destination node, it
forwards it directly. However, if the sender node
cannot forward the packet itself, it must predict
the locations of the receiving node’s neighbors,
and send the packet to the closest neighbor node
to the predicted destination.
MICROSCOPIC TRAFFIC
STION MODEL
Vehicular traffic models may be categorized into
four classifications according to their level of
detail: sub-microscopic, microscopic, mesoscopic
and macroscopic (Hoogendoorn & Bovy, 2001).
Sub-microscopic models describe the characteris-
tics of individual vehicles in the traffic stream and
the operation of specific parts (sub-units) of the
vehicle. Microscopic models simulate individual
128
driver behavior and the interaction among drivers.
This type of model is very detailed and explicitly
tracks the space-time trajectory of each vehicle
(Festa et. al., 2001). Mesoscopic models represent
transportation systems and analyze groups of driv-
ers who display homogeneous behaviors. Finally,
macroscopic models describe traffic at a high level
of aggregation as a flow, without distinguishing
its basic parts (Cvetkovski & Gavrilovska, 1998).
Because we are interested in the space-time
trajectory of each vehicle governed by the lead
vehicle, our attention will focus on microscopic
traffic models.
A large number of microscopic traffic simu-
lation models have been developed. Basically,
these models describe the time-space behavior
of vehicles in the traffic system.
The microscopic traffic simulation model
used in this work to evaluate the performance
of the three algorithms is based on Simone 2000
(Minderhoud, 2002). The Simone 2000 model is
a sophisticated microscopic traffic flow model
that simulates a wide range of user-classes. The
model distinguishes longitudinal (car-following)
and lateral (lane-changing) driver behaviors. The
longitudinal distance controller is one of the main
elements of a microscopic simulation model for
traffic flow. It describes how a vehicle progresses
in a single lane, focusing on the car immediately
in front of it. We have implemented this model in
OPNET to simulate the mobility of the vehicles
on a motorway.
Basically, the simulation model is divided into
two functions:
Desired Gap Function
With this function, the longitudinal controller
determines the acceleration (positive or negative)
needed to obtain a desired minimum distance
from the leader.
si (t ) = li + ηi (t ) ∂( z 0i + z1i ∂vi (t ) + z 2i ∂vi (t ) 2 )
 Inter-Vehicular Communications Using Wireless Ad Hoc Networks

Where: VALIDATION OF THE MODE L

S (t) = desired gap distance (from rear follower
I to rear leader) (m), i = index vehicle, l = length
of vehicle i, η = congestion factor, z0 = margin
parameter (m), z1 = linear headway parameter
(s), z2 = quadratic headway parameter (s2), v (t)
= speed at time t (m/s).
Longitudinal Controller
Once the position of a vehicle immediately in
front of a following vehicle has been calculated,
the longitudinal controller moves the following
vehicle to its new position, using standard kinemat-
ics equations for vehicle speed and distance.
ai (t + τ) = α i ∂(xi −1 (t ) − xi (t ) − si (t ) ) + βi± ∂(vi −1 (t ) − vlation
i −1 (t ) − xi (t ) − si (t ) ) + βi ∂(vi −1 (t ) − vi (t ) )
±
Validating LORA-CB F with One, Two
and Three Hops
Wireless ad hoc networks basically employ multi-
hop communications, where packets are transmit-
ted from source to destination. Therefore, the basic
communication mechanism is peer-to-peer, with
packets retransmitted several times.
The first task of our study is to validate our
model at one, two, and three hops, comparing the
results of the test bed with the results of the model
we developed in OPNET. Finally, for more than
3 hops, we will validate our model comparing it
with the AODV, DSR and GPSR algorithms.
Table 1 compares test-bed and OPNET simu-
i ) ) results, which validates LORA_CBF in a
(t
small-scale network.

Validating LORA-CB F with More

With:
a(t + τ) = acceleration applied after delay time
(m/s2), x(t) = x-coordinate vehicle rear bumper at
time t (m), v(t) = speed at time t (m/s), i = index
subject vehicle (follower), i-1 = index subjects’
leader, α = distance error sensitivity (1/s), β+ =
speed difference sensitivity (for positive differ-
ence) (1/s2), β- = speed difference sensitivity (for
negative difference) (1/s2).
Table 1. Results validating LORA_CBF for one,
two and three hops
Than Three Hops
We have compared our model with the AODV,
DSR, and GPSR algorithms. The comparison is
reasonable because we have improved the data
reception mechanism by using an acknowledge-
ment packet in the AODV and DSR protocols.
When the timer for an acknowledgement data
packet expires, AODV and DSR start new Route
Requests (RREQ). Additionally, we have im-
proved GPSR with our short-term geographic
positioning predictive algorithm.

EED (ms) One Hop Two Hops Three Hops

S IMULATION ME TRICS
Test Bed (C-K
10.4 19.7 29.2
Toh)
In comparing the performance of the algorithms,
OPNET model 9.1 18.854 28.591
we chose to evaluate them according to the fol-
Throughput
One Hop Two Hops Three Hops lowing five parameters:
(Kbps)
Test Bed (C-K
Toh)
769.23 406.091 273.972 • Average end-to-end delay of data packets:
OPNET model 878.93 424.313 279.808 all of the possible delays caused by buffer-

129
 Inter-Vehicular Communications Using Wireless Ad Hoc Networks
ing during route discovery, queuing at the
interface queue, retransmission delays at the
MAC, and propagation and transfer times.
• Routing load: the number of routing packets
sent divided by the number of data packets
transmitted. The latter includes only the data
packets finally delivered at the destination
and not the ones that are dropped. The trans-
mission of each hop is counted once for both
routing and data packets. This provides an
estimation of network bandwidth consumed
by routing packets with respect to “useful”
data packets.
• Routing overhead: the total number of rout-
ing packets transmitted during the simula-
tion. For packets sent over multiple hops,
each packet transmission (each hop) counts
as one transmission.
• Overhead (packets): the number of rout-
ing packets generated divided by the total
number of data packets transmitted, plus
the total number of routing packets.
• Packet delivery ratio: the ratio of data packets
delivered to the number of data packets sent
Figure 3. Simulated scenario
 lanes
clockwise
 lanes counter
clockwise
Maximum transmission
range = 00 m
130
by the sender. Data packets, however, may
be dropped if the link is broken when the
data packet is ready to be transmitted.
The OPNET simulator was used to evaluate the
three routing protocols. The simulation models a
network of 250 mobile nodes traveling on a 6283m
circular road (Figure 3).
This configuration is reasonable for motor-
way traffic in the United Kingdom because the
low curvature rate of its roads permits vehicle
circulation at a more constant velocity. The IEEE
802.11b Distributed Coordination Function (DCF)
is used as the medium access control protocol. We
also developed a microscopic traffic simulation
model in OPNET to simulate vehicular mobility
on a motorway. A 300m transmission range was
chosen, which is consistent with current 802.11b
Wireless LAN standards and 5 dBi gain car-
mounted antennas. An experiment was carried
out to validate the transmission range between
two vehicles driving in opposite directions.
SIMULATION RESULTS
Figure 4.a shows routing overhead. In this simula-
tion, DSR performs better because it lacks a neigh-
bor sensing mechanism, and AODV increases
its routing overhead according to the distance
between nodes. LORA_CBF maintains its rout-
ing overhead at an almost constant level because
routing overhead is proportional to the frequency
of Hello messages, which is independent of the
maximum distance between communication part-
ners. AODV requires about 3 times the routing
overhead of DSR (also reported in Broch et. al.,
1998). Figure 4.b shows overhead, which is higher
for AODV. Generally, highly mobile environments
suffer from a greater frequency of broken links,
resulting in the retransmission of RERR messages.
In the case of AODV, overhead increases propor-
tionally to the number of Hello messages. Figure
4.c represents the routing load. AODV shows a
Inter-Vehicular Communications Using Wireless Ad Hoc Networks

Figure 4.a. Routing overhead Figure 4.b. Overhead

00
Routing Overhead (K packets)

6
99

Overhead (packets)
 98

9
6

96

9
6
9

 9
       
Hop Count Hop Count
LORA_CBF Motorway (1 Mbps - 1448 Bytes) LORA_CBF Motorway (11 Mbps - 1448 Bytes)
LORA_CBF Motorway (1 Mbps - 1448 Bytes) LORA_CBF Motorway (11 Mbps - 1448 Bytes)
AODV Motorway (1 Mbps - 1448 Bytes) AODV Motorway (11 Mbps - 1448 Bytes)
AODV Motorway (1 Mbps - 1448 Bytes) AODV Motorway (11 Mbps - 1448 Bytes)
DSR Motorway (1 Mbps - 1448 Bytes) DSR Motorway (11 Mbps - 1448 Bytes)
DSR Motorway (1 Mbps - 1448 Bytes) DSR Motorway (11 Mbps - 1448 Bytes)

Figure 4.c. Routing load Figure 4.d. End-to-End delay (EED)

00 60
000 0
900
e nd-to-e nd Delay (ms)
Routing Load (packets)

0
800
00
00
600 80

00 60
00
0
00
0
00
00 0
       

Hop Count Hop Count

LORA_CBF Motorway (1 Mbps - 1448 Bytes) LORA_CBF Motorway (11 Mbps - 1448 Bytes) LORA_CBF Motorway (1 Mbps - 1448 Bytes) LORA_CBF Motorway (11 Mbps - 1448 Bytes)
AODV Motorway (1 Mbps - 1448 Bytes) AODV Motorway (11 Mbps - 1448 Bytes) AODV Motorway (1 Mbps - 1448 Bytes) AODV Motorway (11 Mbps - 1448 Bytes)
DSR Motorway (1 Mbps - 1448 Bytes) DSR Motorway (11 Mbps - 1448 Bytes) DSR Motorway (1 Mbps - 1448 Bytes) DSR Motorway (11 Mbps - 1448 Bytes)

Figure 4.e. Delivery ratio

9
Delivery Ratio (packets)

8



6






   
Hop Count

LORA_CBF Motorway (1 Mbps - 1448 Bytes) LORA_CBF Motorway (11 Mbps - 1448 Bytes)
AODV Motorway (1 Mbps - 1448 Bytes) AODV Motorway (11 Mbps - 1448 Bytes)
DSR Motorway (1 Mbps - 1448 Bytes) DSR Motorway (11 Mbps - 1448 Bytes)

131
 Inter-Vehicular Communications Using Wireless Ad Hoc Networks
higher routing load than LORA_CBF and DSR.
The routing load also increases with distance and
depends on the amount of data delivered. End-to-
End delay (EED), presented in Figure 4.d, shows
that all of the algorithms have lower delays at a data
rate of 1 Mbps. In general, AODV has the greatest
delay because of its frequent retransmissions. DSR
performs the best because of its packet control
strategy. LORA_CBF has a slightly greater EED
when compared with DSR. Figure 4.e compares
the packet delivery ratio of all of the algorithms
considered in this study. LORA_CBF shows
good results at both data rates, and AODV has
a slightly worse packet delivery ratio than DSR.
Both AODV and DSR have their worst delivery
ratios at a data rate of 11 Mbps.
We observe that the AODV and DSR algo-
rithms suffer from sub-optimal routes and low
delivery ratios. On the other hand, algorithms
that employ GPS do not satisfy the requirements
of multi-hop vehicular ad hoc networks. In this
section, we will compare our Location-Based
Routing Algorithm with Cluster-Based Flooding
(LORA-CBF), with a very popular position-based
routing algorithm called GPSR, (Greedy Perim-
eter Stateless Routing) and demonstrate that at
an average speed of 42 m/s (~150 km/h), the lack
of a predictive algorithm in GPSR diminishes its
performance.
We implemented the GPSR algorithm on the
same circular dual carriageway scenario with the
same number of vehicles and a relative speed of
84 m/s (~300 km/h). Results show that without
a predictive algorithm, it is not possible to com-
municate a source-destination pair located further
than two hops. Our intention here is to show that
employing a predictive algorithm will improve
communication on a motorway.
Using GPSR, we applied the same short-term
geographic positioning predictive algorithm used
in LORA-CBF and the same physical and MAC
layer. In addition, the same metrics were previ-
ously employed to analyze the behavior of LORA-
132
CBF, AODV, and DSR to compare the relative
performance of LORA-CBF and GPSR.
Figure 5.a represents the delivery ratio of GPSR
and LORA-CBF. The short-term geographic
positioning predictive algorithm improved the
communication of GPSR by 90 percent. Both
algorithms perform similarly at data rates of 11
Mbps; however, GPSR has a slightly lower de-
livery ratio at data rates of 1 Mbps. In general,
these two algorithms perform similarly because
they both use the same forwarding mechanism
(greedy forwarding).
Each hop is equivalent to approximately 300m
and the delivery ratio is the percent of 100 packets
sent at 100-second intervals. Importantly, the
distance between the transmitter and the receiver
is measured in hop counts. End-to-End Delay is
shown in Figure 5.b. Because they employ the
same forwarding mechanism, GPSR and LORA-
CBF behave similarly in terms of EED. The
End-to-End Delay represents the average time
required to send one hundred packets at 100
second intervals. Also, the distance between the
transmitter and the receiver is measured in hop
counts.
Figure 5.c describes routing overhead. Here,
LORA-CBF has a slightly higher routing overhead
at a data rate of 11 Mbps compared with GPSR. On
the other hand, at a data rate of 1 Mbps, LORA-
CBF begins with a slightly higher routing overhead
at a distance of less than 5 hops. Both algorithms
have exactly the same routing overhead at a dis-
tance of 5 hops. At more than five hops, however,
LORA-CBF has a lower routing overhead than
GPSR. Routing overhead represents the total
number of routing packets transmitted during the
simulation. For packets sent over multiple hops,
each packet transmission (at each hop) represents
one transmission. Figure 5.d presents the route
discovery time. Both algorithms show similar
behavior at a data rate of 11 Mbps, however, at a
data rate of 1 Mbps, GPSR’s route discovery time
is greater. Importantly, GPSR suffers from more
Inter-Vehicular Communications Using Wireless Ad Hoc Networks

Figure 5.a. Delivery ratio Figure 5.b. End-to-End delay

00 0
Delivery Ratio (percent of packets)

99.8 0
99.6
00
99.

e e D (ms)
99. 80
99
60
98.8
0
98.6
98. 0
98.
0
98
   
   
Hop Count
Hop Count

LORA_CBF Motorway ( Mbps) LORA_CBF Motorway ( Mbps)

LORA_CBF Motorway ( Mbps) LORA_CBF Motorway ( Mbps)
GPSR Motorway ( Mbps) GPSR Motorway ( Mbps) GPSR Motorway ( Mbps) GPSR Motorway ( Mbps)

Figure 5.c. Routing overhead Figure 5.d. Route discovery time

90
00
80
Routing Overhead (number of

Route Discovery Time (ms)

00
0
6900 60
packets)

600 0
600 0

600 0
0
600
0
900
0
00    
   
Hop Count
Hop Count

LORA_CBF Motorway ( Mbps) LORA_CBF Motorway ( Mbps)

LORA_CBF Motorway ( Mbps) LORA_CBF Motorway ( Mbps)

GPSR Motorway ( Mbps) GPSR Motorway ( Mbps) GPSR Motorway ( Mbps) GPSR Motorway ( Mbps)

Figure 5.e. Overhead Figure 5.f. Routing load

00 
Overhead (percent of packets)

Routing Load (number of

99. 0

68
99
packets)

66
98.
6
98
6

9. 60

9 8
       

Hop Count Hop Count

LORA_CBF Motorway ( Mbps) LORA_CBF Motorway ( Mbps)

LORA_CBF Motorway ( Mbps) LORA_CBF Motorway ( Mbps)

GPSR Motorway ( Mbps) GPSR Motorway ( Mbps)

GPSR Motorway ( Mbps) GPSR Motorway ( Mbps)

133

packet collisions because the spatial separation
of its packets is more limited. On the other hand,
LORA-CBF reduces the route discovery time due
to its cluster-based flooding mechanism. Route
Discovery Time is the time required for the source
to send the first data packet.
Figure 5.e shows the overhead. Here, GPSR
has a slightly lower overhead compared to LORA-
CBF at data rates of 1 Mbps. At data rates of
11 Mbps, however, both algorithms perform
similarly. Overhead is the total number of routing
packets generated divided by the total number of
data packets transmitted, plus the total number of
routing packets. Overhead represents the percent
of 100 packets transmitted during 100 second
intervals. The routing load is shown in Figure
5.f. again, at a data rate of 11 Mbps, LORA-CBF
has a slightly higher routing load than GPSR,
however, at a data rate of 1 Mbps, LORA-CBF
has a slightly higher routing load when it initi-
ates data transmission. However, at five hops
(=1500m), both algorithms have the same routing
load. At greater distances, LORA-CBF exhibits
a lower routing load than GPSR. Again, due to
its spatial separation, GPSR packet collisions are
more frequent, resulting in an increased routing
load. Routing Load is measured in terms of the
number of routing packets transmitted divided by
the number of data packets transmitted.
Figure 6.a. End-to-End Delay for one hop
00
ABR -hop
LORA-CBF -hop
0
00
0
e e D (ms)
00
0
00
0
0 0
6 000
s ize of packets (bytes)
134
Inter-Vehicular Communications Using Wireless Ad Hoc Networks
IMPLEMENTATION OF THE
LORA-CB ALGORITHM ON A
TESTBED
We deployed LORA-CBF on a test bed using Linux
and equipped each node with an Enterasys wire-
less card, employing sockets to allow internodal
communication. Five laptops with ad hoc routing
capability were deployed in an outdoor environ-
ment representing a small-scale ad hoc network.
To validate LORA-CBF statically, we compared
LORA-CBF to the results of another wireless ad
hoc network test bed (Toh, 2002). In (Toh, 2002),
each node ran the Associativity-Based Routing
(ABR) protocol. The ABR and LORA-CBF al-
gorithms employ a periodic beaconing strategy
to inform neighbor nodes of their presence, using
both source-initiated on-demand ad hoc routing
protocols to discover routes. The main difference
between LORA_CBF and ABR is that the latter
selects its route based on its longevity. On the
other hand, LORA-CBF uses a predictive algo-
rithm to select the single best route based on the
geographic locations of neighbor nodes.
Figures 6.a and 6.b show 1, 2, and 3 hop results
for LORA-CBF and ABR. Results show similar
behavior for the different packet sizes selected
for the study. Each value on the graphs represents
results for 100 packets sent, and the margin of er-
Figure 6.b. End-to-End Delay for two and three
hops

ABR -hops
0 ABR -hops
LORA-CBF -hops
 LORA-CBF -hops
0
e e D (ms)

0

0

886 6 000 8
s ize of packets (bytes)
Inter-Vehicular Communications Using Wireless Ad Hoc Networks
ror is <15%. Discovery time for the first hop is 0
seconds due to the Hello mechanism. For 2 hops,
the discovery time is 1.6 seconds and for three
hops, discovery time increases to 1.74 seconds
because of the route discovery mechanism of the
two algorithms.
CONCLUSON AND FUTURE WORK
In the near future, automobile manufacturers may
use wireless ad hoc networks to improve traffic
flow and safety, in part, because it may be more
cost effective than continually undertaking mas-
sive construction projects, which are proving to
have only limited success. Consequently, future
developments in automobile manufacturing will
include new communication technologies that
offer more effective spacing and collision avoid-
ance systems, greater gas mileage (less brak-
ing), less pollution (cars are in movement), more
information and entertainment, etc. In order to
reduce communication costs and guarantee the
low delays required to exchange data between
cars, inter-vehicle communication (IVC) systems,
based on wireless ad-hoc networks, represent a
promising alternative for future road communica-
tion scenarios, as they permit vehicles to organize
themselves locally in ad hoc networks without
any pre-installed infrastructure.
LORA-CBF is an algorithm that can possibly be
used in future wireless ad hoc networks because
its reactive geographic routing algorithm employs
GPS in tandem with its predictive algorithm,
which is necessary for mobile networks to func-
tion optimally. Furthermore, LORA-CBF uses a
gateway selection mechanism to reduce conten-
tion in dense networks, which is predictable in
highly congested traffic conditions. Finally, the
hierarchical structure of LORA-CBF facilitates
its deployment as part of vehicular ad hoc net-
works because it requires minimal deployed
infrastructure.
In this chapter, we have discussed the mobility
involved in typical motorway traffic scenarios and
have provided simulated results of a very large
250 node network. We validate our simulation,
where possible, with measurements and analy-
sis. We also consider six lanes of moving traffic
(three in each direction) in all our simulations at
theoretical data rates.
We have considered two non-positional-based
routing algorithms (AODV and DSR) and two posi-
tional-based routing algorithms (LORA_CBF and
GPSR). Results show that mobility and network
size affects the performance of AODV and DSR
more significantly than LORA_CBF and GPSR.
Link failures are significantly more common in
highly mobile, dynamic networks. Link failures
trigger a new route discovery phase in all of the
algorithms; however, in AODV and DSR, this
occurs more frequently because of their routing
mechanisms. Thus, the frequency of route discov-
ery packets is directly proportional to the number
of route breaks. We observe that positional-based
routing protocols provide excellent performance
in terms of end-to-end delay and packet delivery
ratio. Their disadvantage, however, is that posi-
tion-based protocols must transmit additional
information, thus increasing network congestion.
Non-positional-based routing algorithms suffer
from sub-optimal routes and have a lower packet
delivery ratio because of dropped packets. In
addition, our Location Routing Algorithm with
Cluster-Based Flooding (LORA_CBF) is robust
in terms of delivery ratio, routing overhead, route
discovery time, and routing load.
Future work related to the development of LORA-
CBF will include integrating GPS, a predictive
algorithm, and dynamic geographical maps into
a sole architecture to be deployed in a test bed.
RE
Abdulhamid, H., Abdel-Raheem, E., & Tepe, K. E.
(2007). Channel estimation for 5.9 Ghz dedicated
135
 Inter-Vehicular Communications Using Wireless Ad Hoc Networks
shortrange communications receiver in wireless
access vehicular environments. IET Communica-
tions Journal, 1(6), 1274-1279.
Bantli, H., Ring, E., & Goff, E. (1997). Require-
ments for direct short range vehicle to roadside
communications. Proceedings of the IEEE Aero-
space Conference, 1, 41-61.
Bevas, R., & Bevas, V. (1997). Routing in ad hoc
networks using a spine. Proceedings in the Sixth
International Conference in Computer and Com-
munication Networks, (pp.34–39).
Boris & Arkady (1999). Link state routing protocol
with cluster based flooding for mobile ad-hoc com-
puter networks. Proceedings of the Workshop on
Computer Science and Information Technologies
CSIT’99, (pp.28–36).
Brasche, G., Rokitansky, C.-H., & Wietfeld, C.
(1994). Communication Architecture and Perfor-
mance Analysis of Protocols for RTT Infrastructure
Networks and Vehicle-Roadside Communications.
Proceedings of the IEEE 44th Vehicular Technology
Conference, 1, 384-390.
Broch, J., Maltz, D., Johnson, D., Hu, Y.-C., &
Jetcheva J. (1998). A performance comparison
of multi-hop wireless ad hoc networks routing
protocols. In Proceedings of the 4th ACM/IEEE
International Conference on Mobile Computing
and Networking (MOBICOM), (pp. 85-97).
Ching-Chuan, Hsiao-Kuang, W., & Mario, G.
(1997). Routing in clustered multihop, mobile
wireless networks with fading channel. IEEE
Singapore International Conference on Networks,
(pp. 197–211).
Cottingham, D. N., Wassell, I. J., & Harle, R. K.
(2007). Performance of IEEE 802.11a vehicular
context. 65th IEEE Vehicular Technology Confer-
ence, (pp. 854-858).
Cvetkovski, B., & Gavrilovska, L. (1998). A simu-
lation of a mobile Highway traffic. Proceedings
136
of the IEEE Vehicular Technology Conference,
2, 1429-1433.
Da Chen, Z., Kung, H.-T., & Vlah, D. (2001). Ad
Hoc Relay Wireless Networks over Moving Ve-
hicles on Highways. Proceedings in International
Conference on Mobile Computing and Networking,
(pp. 247-250).
Dickey, S. R., Huang, C., & Guan, X. (2007). Field
Measurements of Vehicle to Roadside Communica-
tion Performance. 66th IEEE Vehicular Technology
Conference, (pp. 2179-2183).
Dobias, F., & Grabow, W., (1994). Adaptive array
antenna for 5.8 Ghz vehicle to roadside communica-
tion. 44th IEEE Vehicular Technology Conference,
3, 1512-1516.
Festa, D. C., Longo, G., Mazzulla, G., & Muso-
lino, G. (2001). Experimental analysis of differ-
ent simulation models for motorway traffic flow.
Proceedings of the IEEE Intelligent Transportation
Systems Conference, (pp. 675-680).
Füßler, H., Mauve, M., Hartenstein, H., Käse-
mann, M., & Vollmer, D. (2003). MobiCom Poster:
Location-Based Routing for Vehicular Ad-Hoc
Networks. Proceedings in ACM SIGMOBILE
Mobile Computing and Communication Review,
7(1), 47-49.
Hoogendoorn, S. P., & Bovy, P. H.-L. (2001).
State-of-the-art of Vehicular Traffic Flow Mod-
elling. Special Issue on Road Traffic Modelling
and Control of the Journal of System and Control
Engineering, 215(4), 283-303.
Johnson, D., Maltz, D., & Hu, Y.–C. (2007). The
Dynamic Source Routing Protocol (DSR) for
Mobile Ad Hoc Networks for IPv4. http://www.
ietf.org/rfc/rfc4728.txt. Request for Comments
(Work in Progress).
Karp, B., & Kung, H.-T. (2000). Greedy Perim-
eter Stateless Routing for Wireless Networks.
Proceedings of the 6th ACM International Con-
Inter-Vehicular Communications Using Wireless Ad Hoc Networks

ference on Mobile Computing and Networking, the PC: New challenges for the operating system,
(pp. 243-254). (pp. 61-65).
Kosh, T., Schwingenschlögl, C., & Ai, L. (2002). Mussa, R., Upchurch, J. (2000). Simulator evalu-
Information Dissemination in Multihop Inter- ation of incident detection using VRC system.
vehicle Networks – Adapting the Ad-hoc On-de- Proceeding of the IEEE Intelligent Vehicles Sym-
mand Distance Vector Routing Protocol (AODV). posium, (pp. 442-450).
Proceedings in the 5th IEEE International Confer-
Perkins, C., Belding-Royer, E., & Das, S. (2003).
ence on Intelligent Transportation Systems, (pp.
Ad hoc On-Demand Distance Vector (AODV) Rout-
685-690).
ing. http://www.ietf.org/rfc/rfc3561.txt. Request for
Krishna, V., Chatterjee, & Pradhan (1997). A Comments (Work in Progress).
cluster-based approach for routing in dynamic
Raghupathy, Bevas, & Vaduvur (1998). Spine
networks. ACM SIGCOMM, Computer Commu-
routing in ad hoc networks. ACM/Baltzer Cluster
nication Review, (pp. 49–65).
Computing Journal, 1(2), 237–248.
Kwak, J. S., & Jae, H. L. (2004). Infrared trans-
Rokitansky, C.-H., & Wietfeld, C. (1993). Compari-
mission for intervehicle ranging and vehicle-to-
son of Adaptive Medium Access Control Schemes
roadside communications systems using spread
for Beacon-Vehicle Communications. Proceedings
spectrum technique. IEEE Transactions on Intel-
of the IEEE-IEE Vehicle Navigation and Informa-
ligent Transportation Systems, 5(1), 12-19.
tion Systems Conference, (pp. 295-299).
Li, J., Jannotti, J., De Couto, D., Karger, D., &
Rokitansky, C.-H., & Wietfeld, C. (1995). Methods
Morris, R. (2000). A Scalable Location Service
and Tools for Performance Evaluation and Valida-
for Geographic Ad Hoc Routing. Proceedings of
tion of Vehicle-Roadside Communication Proposed
the 6th ACM International Conference on Mobile
for Standardization. Proceedings in Vehicular
Computing and Networking, (pp. 120-130).
Technology Conference, 2, 964-970.
Lochert, C., Füßler, H., Hartenstein, H., Hermann,
Royer, E., & Toh, C. (1999). A Review of current
D., Tian, J., & Mauve, M. (2003). A Routing
routing protocols for ad hoc wireless networks.
Strategy for Vehicular Ad Hoc Networks in City
IEEE Personal Communications, (pp. 46-55).
Environments. Proceedings of the IEEE Intelligent
Vehicles Symposium, (pp. 156-161). Santos, R. A., Edwards, A., Edwards, R. M., &
Seed, N. L. (2005). Performance evaluation of
Matthaiou, M., Laurenson, D. I., Cheng-Xiang,
routing protocols in vehicular ad-hoc networks.
W. (2008). PHY 30-3 – Capacity Study of vehicle-
International Journal Ad Hoc and Ubiquitous
to-roadside MIMO channels with a line-of-sight
Computing, 1(1/2), 80-91.
component. IEEE Wireless Communications and
Networking Conference, (pp. 775-779). Toh, C.-K. (2002). Ad Hoc Mobile Wireless Net-
works: Protocols and Systems. Prentice-Hall
Minderhoud, M. M. (2002). Simulation model of
International.
motorways with next generation vehicles, techni-
cal specification. Traffic Safety Facts. (2004). NHTSA´s National
Center for Statistics Analysis.
Morris, R., Jannotti, J., Kaashock, F., Li, J., &
Decouto, D. (2000). A Scalable Ad Hoc Wireless Transportation Statistics Annual Report. (2006).
Network System. Proceedings of the 9th workshop U.S. Department of Transportation Bureau of
on ACM SIGOPS European Workshop: beyond Transportation Statistics.

137
 Inter-Vehicular Communications Using Wireless Ad Hoc Networks
Wietfeld, C., & Rokitansky, C.-H. (1994). Per-
formance of Vehicle-Roadside Communication
Systems supporting Multiple RTI-Applications.
Proceedings in Vehicle Navigation & Information
Systems Conference, (pp. 133-138).
Wietfeld, C., & Rokitansky, C.-H. (1995). Markov
Chain Analysis of Alternative Medium Access
Control Protocol for Vehicle-Roadside Commu-
nications. Proceedings in Vehicular Technology
Conference, 2, 958-963.
138
Zandbergen, A., & Van der Ree, A. (1992). Experi-
ments on vehicle to roadside microwave commu-
nication. IEEE MTT-S International Microwave
Symposium Digest, 2, 617-620.
Zou, X., Ramamurthy, B., & Magliveras, S. (2002).
Routing Techniques in Wireless Ad Hoc Networks
Classification and Comparison. Proceedings of
the Sixth World Multiconference on Systemics,
Cybernetics and Informatics, SCI.
 139

Chapter VIII
The Role of Communications
in Cyber-Physical Vehicle
Applications
Nicholas F. Maxemchuk
Columbia University, USA & IMDEA Networks, Spain

Patcharinee Tientrakool
Columbia University, USA

Theodore L. Willke
Columbia University, USA & Intel Corporation, USA

ABSRACT

Cyber-physical systems use sensing, communications, and computing to control the operation of physical
devices. Sensing and computing devices have been embedded in automobiles and in the transportation
infrastructure. Communications adds a new dimension to the capabilities of these systems. The embed-
ded computers and sensors in both vehicles and the infrastructure will be networked into cyber-physi-
cal systems that reduce accidents, improve fuel efficiency, increase the capacity of the transportation
infrastructure, and reduce commute times. The authors describe applications that improve the opera-
tion of automobiles, control traffic lights, and distribute the load on roadways. The requirements on the
communications protocols that implement the applications are determined and a new communications
paradigm, neighborcast, is described. Neighborcast communicates between nearby entities, and is par-
ticularly well suited to transportation applications.

Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.
 The Role of Communications in Cyber-Physical Vehicle Applications
INTRODUCTION
Cyber-physical systems use sensing, communica-
tions and computing to control the operation of
physical devices. Sensing and computing devices
have been embedded in automobiles to improve
their fuel efficiency and safety. For instance,
automobiles have several computers that are
embedded in engines to improve gas mileage and
detect faults; cruise control is used to maintain
a constant velocity; antilock brakes use sensing
and computing to prevent skids and stop cars as
quickly as possible; and computing is combined
with GPS systems to plan routes. Sensing and
computing are also used in traffic lights to reduce
travel times and fuel consumption by changing
signals only when traffic is waiting and by tim-
ing signals to permit traffic to flow in a dominant
direction unimpeded.
New applications of sensors are being intro-
duced in automobiles. Sensors detect obstacles
behind a car that is backing up and warn the
driver when there is a car in a blind spot or when
he is following too close to the car in front. The
front, back and side sensors are also used to as-
sist drivers when parallel parking. Sensors will
transform cruise control from a simple device that
maintains a constant speed to a more intelligent
system that controls the maximum speed, but
reduces the target speed as a function of nearby
automobiles (Heddebaut et al., 2005; Vahidi
& Eskandarian, 2003; Zhang et al., 1999). The
speed will be decreased when the preceding car
is traveling at less than the target speed to avoid
a rear end collision, or when cars in adjacent
lanes are traveling at much lower speeds to give
a driver sufficient time to respond to vehicles that
change lanes.
Communications adds a new dimension
to systems and applications. The embedded
computers and sensors in both vehicles and the
infrastructure will be networked together to im-
prove the operation of existing applications and to
make new classes of applications possible. These
140
cyber-physical systems will change the way we
use transportation.
Communications between nearby vehicles will
enable cooperative control paradigms that reduce
accidents more than computing and sensors alone.
For instance, a distributed anti-lock brake system
will not only allow a car to brake more quickly
by preventing skids, but will negotiate with sur-
rounding vehicles to avoid collisions while stop-
ping (Liu & Ozguner, 2003). In instances where
accidents cannot be avoided, the system will select
a stopping speed that minimizes the damage. Side
sensors can provide a warning when there are
cars in adjacent lanes, while communications can
negotiate with those cars to safely change lanes.
Vehicle control systems will increase the capac-
ity of highways by allowing cars on congested
roadways to safely travel faster and with less
space between vehicles. The communications
procedures that are used in automotive applica-
tions are described in the sections Centralized
Control and Distributed Control.
Communications between vehicles and the
infrastructure will improve the scheduling of
traffic signals and route planning. Instead of rely-
ing on sensors to detect cars that are waiting at a
light, communications will notify the computers
that control the lights before the vehicles arrive.
On congested roadways, traffic lights at adjacent
intersections will be coordinated to minimize the
stopping time for cars traveling in all directions.
The use of communications for controlling traffic
signals is described in the section, Interaction of
Automobiles with Traffic Signals. Route planning
computers will communicate with a traffic control
center to determine congested regions and may
also cooperate with other route planning com-
puters to avoid creating new congested regions,
as described in the section, Cooperative Route
Planning. Recently, sensors have been installed
in parking spots in San Francisco and commu-
nications with drivers reduces the time and fuel
consumed to search for parking spots.
The Role of Communications in Cyber-Physical Vehicle Applications
Communications can be implemented with
existing wireless technologies that are part of
the communications infrastructure, but will also
take advantage of evolving infrastructureless
networks. For example, vehicle ad hoc networks
(VANETs) can provide wireless connections
directly between vehicles. These networks are
ad hoc in the sense that they are self-organized
and self-managed by the participants. VANETs
provide inter-vehicle communications without
requiring an investment in a new communications
infrastructure, and will permit new applications to
be deployed more quickly and less expensively.
BACKGROUND
The field of cyber-physical systems (CPS) has
only recently emerged. The United States National
Science Foundation (NSF) and other bodies have
sponsored a number of workshops on the topic
since 2006 (NSF, 2006; NITRD, 2008). The NSF
has identified CPS as a key research area, in part
because it tackles the difficult problem of per-
forming distributed control within a distributed
computing environment (Wolf, 2007). There is
a keen focus on transportation systems, where
CPS may improve safety, efficiency and comfort.
However, the highly mobile and lossy nature of
vehicular networks further challenges the design,
verification, and performance of CPS.
Whether the system uses a distributed or cen-
tralized control design, the computing resources
are distributed and the system’s performance may
be largely a reflection of the network’s ability to
deliver data quickly, efficiently, and guarantee
other characteristics. Within VANETs, the wire-
less medium is inherently broadcast-friendly and
many of the applications involve transmitters that
are interested in sending the same message to more
than one receiver. This is true, for example, when
a vehicle would like to inform others nearby of
a road hazard. The message could be sent using
unicast but it would require the sender to know
the identity of each recipient and send a separate
transmission to each. In such a case, a broadcast
or multicast protocol that permits a message to
be reliably received simultaneously by a group
of vehicles qualified, for example, by location is
much more effective. A number of these protocols
are surveyed in (Willke et al., 2008) and (Sichitiu
et al., 2008). Many were specifically designed to
address the real-time, high reliability requirements
of inter-vehicle communications applications.
Vehicular applications have very diverse
communication requirements with respect to
latency, throughput, reliability and other metrics.
The applications may be generally classified into
three areas based on how they utilize information,
which is a significant factor in determining sensor
and communication performance requirements
(Willke et al., 2008):
1. Information services
a. General
b. Vehicle safety
2. Autonomous control
a. Front-back peer braking and intersec-
tion collision avoidance
b. Automated cruise control
3. Cooperative control systems (Centralized
and Decentralized)
a. Platooning and coordinated braking
b. Coordinated maneuvering
c. Traffic flow control
d. Route planning
Application examples are listed for each area.
Information services include mobile Internet and
other data dissemination that may be related to
safety but is not directly involved in vehicle con-
trol. Autonomous control includes applications for
which vehicles collect data from nearby peers to
individually optimize their own braking, accel-
eration, or steering. Cooperative control includes
applications that involve coordination and joint
planning amongst vehicles to optimize their col-
lective safety, efficiency, or comfort.
141
 The Role of Communications in Cyber-Physical Vehicle Applications
A number of consortia, companies, and gov-
ernments are working to improve transportation
by investigating applications, communication
protocols, and integrated vehicle systems. The
CarTALK 2000 project (Reichardt et al., 2002) is
focused on three applications for road safety: 1) a
driver warning system to distribute information
on road conditions, 2) adaptive cruise control and
automated braking, and 3) cooperative systems for
lane merges and collision avoidance in intersec-
tions. The Intelligent Vehicle Initiative (U.S. DOT,
2008) is studying how drivers should interact with
in-vehicle safety systems, performance criteria
for applications, and deployment issues, includ-
ing driver acceptance. The PReVENT project
(ERTICO-ITS Europe, 2008) has looked at the
fusion of GPS, infrared, radar and other sensors
with communications to improve the effectiveness
of road safety applications.
There are a number of initiatives focusing on
wireless technologies for VANETs. The IEEE is
standardizing a random access protocol for Wire-
less in Vehicular Environments (WAVE) (IEEE,
2006). WAVE is the foundation of the U.S. DOT’s
Dedicated Short Range Communication (DSRC)
project (IEEE, 2007), and the Vehicle Safety
Communications Consortium (VSCC, 2008) is
developing the requirements on DSRC when it is
used in safety applications. The European Com-
mission and car industry is developing a CDMA
protocol for multi-hop ad-hoc networks in the
CarTALK 2000 project (CarTALK, 2002). The
Car-2-Car Communication Consortium (C2C-CC,
2008) is looking at wireless protocols for intel-
ligent transportation systems. The Network on
Wheels program (BMBF, 2008) is investigating
the security issues for the protocols developed
in the C2C-CC, including, false data, jamming,
corruption, and denial-of-service attacks.
Other groups are integrating communications,
protocols and applications into complete Intel-
ligent Transportation Systems. One objective is
to provide vehicles with seamless and continuous
communication in a variety of environments,
142
such as highways and urban streets, and across
a variety of communications systems, including
vehicle-to-vehicle and vehicle-to-infrastructure
networks. The California Partners for Advanced
Transit and Highways proposed an Automated
Highway System (Horowitz & Varaiya, 2000) that
improves safety and capacity through automated
control of vehicles, traffic admission control, and
route planning. The ISO TC 204 Working Group
16 (2007) is developing CALM (Continuous
Communications for Vehicles). CALM’s archi-
tecture enables data collection, such as traffic
flow measurements, distribution of delay sensitive
information, such as emergency alerts, and access
to general Internet data from moving vehicles.
CALM integrates many sensor and communi-
cations technologies, including infrared, radar,
DSRC, 3G cellular, and others.
Vehicle warning and control systems place
unique demands on VANETs. Safety and motion
control applications typically have a real-time
requirement for information delivery between
vehicles. Some of the control mechanisms may fail
when delays exceed tens of milliseconds (Willke
et al., 2008), and the maximum delay must be
bounded with a very small loss probability (Li
& Wang, 2007). Because of the relatively high
message loss rates in wireless networks, rapid
recovery mechanisms are required to provide a
high probability that a message is delivered in
a short time. Many vehicle applications require
one-to-many or many-to-many broadcast, and
some applications require the additional ability to
determine when broadcast information is received
by all of the receivers. For example, a vehicle that
negotiates a lane change may require that all of
the vehicles that are cooperating receive the in-
structions and know that all of the other vehicles
have also received the instructions. If the vehicles
cannot confirm reception, it may be necessary to
abort the maneuver.
VANET architectures also face a number of
significant challenges not present in other types
of ad hoc networks (Li & Wang, 2007; Willke et
al., 2008):
The Role of Communications in Cyber-Physical Vehicle Applications
• As vehicles move with respect to one an-
other, the connectivity and routing change
more frequently than in slower moving
networks.
• On sparsely populated highways network
partitions occur, causing groups of vehicles
to lose connectivity.
• In many applications vehicles communi-
cate with one another based on position,
rather than identity, making it necessary
to integrate communications and location
services.
Protocol verification and conformance testing
are particularly important in vehicle control sys-
tems. In safety applications it is necessary to verify
that the protocol performs the required services
and that there are no unexpected or infrequently
occurring states that cause dangerous conditions.
The primary objective in vehicle control must be
to do no harm. We must make certain that the
systems never cause an accident or make an un-
avoidable accident more severe. In addition, many
of the applications require peer level cooperation
between different vehicles. It is expected that there
will be many providers of these systems and that
they will have to interact with one another. It is
particularly important that the implementations
are tested to make certain that they conform to
the verified protocols. Designing protocols that
can be verified and tested places constraints on
the protocols as noted in (Lee, 2008).
CETRALIZED CONTROL
Today drivers operate their vehicles autonomously.
They determine their speed based on traffic and
road conditions, traffic rules, and their own time
pressure. They judge the need for lane changes
and other maneuvers, navigate, and are respon-
sible for their own safety and the safety of those
with whom they share the road. Intelligent cruise
control and sensor warning systems are technolo-
gies that can improve automotive safety. These
mechanisms assist in the autonomous operation
of individual vehicles. Additional improvements
can be attained by enabling groups of vehicles to
share their sensor readings and coordinate their
actions. The coordination can be performed in a
centralized manner, obeying the instructions of a
leader, or in a distributed manner, by negotiating
with nearby vehicles.
In this section we consider centralized con-
trol mechanisms to coordinate vehicles. We
will start by describing an application called
platooning. The requirements of the application
dictate the characteristics of the communications
protocols. We will describe the Wireless Token
Ring Protocol, WTRP (Lee et al., 2001), that
was originally used in platooning applications
and a recent addition to the vehicle protocols,
the Mobile Reliable Broadcast Protocol, MRBP
(Willke & Maxemchuk, 2005). MRBP is specifi-
cally designed to operate in the high error rate
environment of wireless networks and to respond
to the frequently changing groups encountered in
vehicle coordination.
Platooning
Platooning occurs at present when a string of
trucks on a highway is organized in a convoy,
as depicted in Figure 1. The front vehicle is the
leader and sets the average speed for the group.
Each driver accelerates and applies his brakes
to closely follow the preceding vehicle. This im-
proves the aerodynamics of the group, leading to
increased fuel efficiency. Convoys also increase
the capacity of roadways by having vehicles travel
closer together at highway speeds.
Fuel efficiency and highway capacity increase
as inter-vehicle spacing is decreased. Sensors and
communications can be used to assist drivers and
safely decrease the spacing. Lasers, radar, and
sonar can be used to measure the distance between
vehicles and the vehicle velocities (Gehring &
Fritz, 1997). An onboard computer can use these
143
 The Role of Communications in Cyber-Physical Vehicle Applications
Figure 1. Lane of vehicles with platooning parameters
inputs to control acceleration and braking with
greater precision and responsiveness than human
drivers. These techniques are currently used to
implement anti-lock brakes.
Inter-vehicle communications can augment
sensor information by announcing changes in
the speed of the group before vehicles can sense
the change in the preceding vehicle. Vehicles
broadcast the sensor information to the rest of the
group so that vehicles can control their operation
in response to vehicles that they cannot directly
observe. For instance, a truck may apply its brakes
when a vehicle one, five, or ten vehicles ahead of it
starts to decelerate, rather than waiting for each of
the vehicles in between to detect braking. Drivers
typically respond to vehicles that are two or three
vehicles in front of themselves. Automated control
systems that are restricted to sensor information
from immediately adjacent vehicles use less in-
formation than humans, but make up for the lack
of information by responding more quickly. With
communications, automated control systems can
use more information than humans in addition to
responding more quickly.
The effect of centralized control on highway
capacity is demonstrated by considering two
control strategies for platoons. In the first strat-
egy, vehicles maintain a distance consisting of a
constant L0, that is the minimum distance between
vehicles at rest, plus a distance that is proportional
to velocity. This proportion is called the time
headway, hw. This strategy can be implemented
with or without sensors to assist drivers. The ef-
144
fect of hw selection on platoon control stability is
studied in (Swaroop, 1995). The second strategy,
which requires communication, is constant spac-
ing. In it, hw is set to zero, making L0 the constant
vehicle spacing. L0 may be set as low as a few
meters without causing vehicles to collide, as
even heavy vehicles may be controlled to within
two meters (Song, 2004).
The effect of the two vehicle control policies on
traffic capacity is shown in Figure 2. The capacities
are plotted against vehicle velocity, with 88 feet/sec
corresponding to a highway speed of 60 mph. All
capacities are reduced by 20% to account for lane
changes, merges and other disturbances. The time
headway policy with hw = 2 seconds represents a
safe manual following distance. The policy with
hw = 0.5 seconds represents an automated policy,
where sensors are used to reduce the safe distance.
The policy with hw = 0 represents a system that
uses communications to simultaneously control
all of the vehicles. The figure shows the increase
in capacity when we use technology to reduce
spacing and the further increase when we use
communications. In a centralized control system,
the increased capacity due to communications
only occurs within a platoon. The distributed
control procedures, described in the next section
permit the increased capacity to occur over the
entire congested region of a highway.
The leader in a convoy issues instructions for
all vehicles to change speed or brake simultane-
ously, controls maneuvers such as lane merges and
lane changes, and manages group membership.
The Role of Communications in Cyber-Physical Vehicle Applications
Figure 2. Traffic capacity for various vehicle headway policies
14000
12000
[vehicles/lane/hour]
[vehicles/lane/hour]
10000
8000
6000
Capacity
Capacity
4000
2000
0
0 20
The leader must ensure that the platoon operates
within the performance limits of the least capable
vehicle and can deny membership to vehicles that
may limit the performance of the convoy. The
communication protocol should provide reliable
message delivery between all of the vehicles and
should manage the group membership.
On a congested highway, groups of vehicles
will interact with one another. For example, the
leaders in two platoons in adjacent lanes may need
to coordinate a lane merger; or platoons may be
split or merged depending on the capabilities of
the vehicles. In a centralized system, the com-
munication between platoons may be organized
hierarchically, with a division leader coordinating
the operation of a group or nearby platoons and a
highway general controlling multiple divisions on
the same roadway. Communications follows the
hierarchical structure. A general communicates
with his division leaders and the division leaders
communicate with platoon leaders. Communica-
tion delays prevent the general from controlling the
C onstant Spac ing (10 feet)
Time Headway (0.5 sec )
Time Headway (2 sec )
40 60 80 100 120
velocity [feet/sec]
individual vehicles as well as the platoon leader. As
a result, the spacing between adjacent vehicles in
different platoons must be greater than the spacing
between adjacent vehicles in the same platoon. On
a congested highway, distributed control strate-
gies allow all vehicles to communicate with their
neighbors and reduce the communications delays
that occur on platoon boundaries.
WTRP
WTRP is a token passing protocol that gives
each source in a network the right to transmit in
order. In platooning applications, WTRP is used
to periodically update the state of each vehicle
each time it receives the token. The token pass-
ing mechanism guarantees that sources do not
transmit simultaneously and collide. Scheduled
protocols are well suited for periodic updates.
This type of token passing protocol was
initially used on wired loop topologies, where
the transmission order is dictated by the wiring,
145
 The Role of Communications in Cyber-Physical Vehicle Applications
but it has also been used on bus networks where
the order is specified by a token list. In wireless
networks, the probability of transmission error
is higher than in wired networks; and in vehicle
networks mobility causes the members of the
token group to change frequently.
In token networks, the token must be recovered
and restarted when it is lost. With high packet
loss, the token must be acknowledged when it is
passed and retransmitted when it isn’t acknowl-
edged to reduce the probability that the token is
lost. After a number of attempts to forward the
token, the token site must assume that the next
station has left the group and reconfigure the to-
ken group to exclude that station. The maximum
time for the token to circulate around the group
is proportional to the number of vehicles in the
convoy times the time to make the maximum
number of retries and a reconfiguration at each
of the stations. When a vehicle wants to transmit
an emergency message, it may have to wait up to
the maximum token cycle time.
An acknowledgement message guarantees that
the source message is received by the next node
in the token list but does not guarantee that it is
received by all of the nodes in the token group.
In a wireless network a node’s state information
may have to be transmitted several times before
all of the other nodes receive the message with
high probability.
When new vehicles join a platoon, or when a
vehicle leaves a platoon, the token list must be
modified. WTRP uses a separate protocol that
can add or delete one vehicle at a time.
WTRP is well suited for control strategies
that use periodic updates for all vehicles, rather
than procedures that require rapid, reliable dis-
semination of emergency messages. It is most
useful in small, infrequently changing convoys
in a transmission environment with a low mes-
sage loss rate.
146
MRB
MRBP uses a token to give receivers the right to
acknowledge transmitted messages rather than
giving the sources the right to transmit messages.
Source messages are broadcast to the entire group
using a contention resolution broadcast protocol,
such as the IEEE 802.11 protocol, as depicted in
Figure 3. MRBP is based on the Reliable Broadcast
Protocol, RBP (Chang & Maxemchuk, 1984), and
the Timed Reliable Multicast Protocol (Maxem-
chuk & Shur, 2001).
In MRBP the token is forwarded to each re-
ceiver in the list. The token site holds the token
for a fixed time and then transmits a token passing
message that acknowledges all of the source mes-
sages received while it held the token. The token
passing/acknowledgement message is broadcast
to all of the sources and receivers that take part
in the protocol.
MRBP uses an aggressive token passing
procedure. A token site forwards the acknowl-
edgement/token at its scheduled time, whether
or not the token site has received the preceding
tokens. When a scheduled acknowledgement isn’t
received, recovery attempts begin shortly after its
scheduled transmission and end after the maxi-
mum allowed recovery period, as in Figure 4.
A voting procedure is used to decide which
acknowledgements all of the receivers include in
their acknowledgment list. Each station votes on
the preceding acknowledgements when it passes
the token by listing its missing acknowledgements.
The vote on an acknowledgement message, Ai,
starts with the acknowledgement message that is
scheduled to be transmitted after Ai’s scheduled
transmission plus the maximum allowed time to
recover a missing message, as shown in Figure
4. The vote ends after the token circulates around
the group, returning to the site that transmitted
the first vote, plus the maximum time to recover
the acknowledgement from the last site to vote.
The Role of Communications in Cyber-Physical Vehicle Applications

Figure 3. Token passing mechanism in MRBP

Figure 4. Acknowledgement and message commitment timeline for MRBP

147
 The Role of Communications in Cyber-Physical Vehicle Applications
An acknowledgement is included in the list when
a majority of the stations in the group, at the
time that the vote starts, received or recovered
the message. A similar vote occurs for each of
the source messages that Ai acknowledges. The
vote on the acknowledged messages starts after
each receiver has an opportunity to first recover
the acknowledgement message and then recover
the acknowledged source messages.
Acknowledgements and acknowledged mes-
sages are included in the common list at each of
the receivers only when a majority of the receivers
have the message. The majority is determined by
counting “yes” votes, “no” votes, and abstentions.
Abstentions occur when voting receivers have left
the group. If the “yes” votes at a receiver exceed
the majority threshold, a message is included in
the list. If the “no” votes at the receiver exceeds
the majority threshold, the message is excluded
from the list. Neither a majority of “yes” or “no”
votes may be caused by abstentions or by votes
that the receiver has failed to recover. Forcing a
decision may cause this receiver to reach a differ-
ent decision than other receivers in the group. To
prevent having different data sets at the different
receivers, a receiver that reaches an anomalous
decision leaves the group.
The vote procedure is used to continuously
add and delete stations from the group. When
a station’s acknowledgement is not included in
the list, the station is removed from the token
group. A station joins the group by transmitting
a source message that must be acknowledged and
voted into the group. The joining station, as well
as each receiver in the group, tallies the vote on
the joining station. The distributed vote tallying
procedure guarantees that all receivers in the
group reach the same membership decisions, or
leave the group themselves, so that all stations in
the group have the same membership list.
MRB Characteristics
MRBP is an efficient acknowledgement protocol.
When there are no transmission losses each source
148
message is only acknowledged by one of the re-
ceivers, independent of the number of receivers
in the group. The voting protocol guarantees that
every receiver recovers the same set of source
messages.
In addition to guaranteeing that all of the re-
ceivers have all of the source messages, MRBP
places the messages in the same order at each of
the receivers. Acknowledgements are sequentially
numbered and each receiver recovers all of the
acknowledgements and the messages that they
acknowledge. The numbered acknowledgements
allow all of the receivers to place the source mes-
sages in the same sequence, which guarantees
that copies of the same control program running
on different vehicles reach the same conclusion.
The aggressive token passing protocol may cause
multiple receivers to acknowledge the same source
message, but the earliest acknowledgement takes
precedence when sequencing the messages.
MRBP rapidly distributes critical source mes-
sages with two levels of guarantees. In a relatively
short time, proportional to the maximum allowable
number of retransmission attempts to recover a
message, MRBP guarantees that all vehicles that
are part of the group have received the source
message, and in a longer time, proportional to
the number of vehicles in the token group, each
vehicle learns which other vehicles received the
message. The first level of guarantee is adequate
for vehicles to respond to an emergency message,
such as a rapid stop, and the second level of guar-
antee is needed for coordinated maneuvers, such
as a negotiated lane change.
The time for MRBP to deliver emergency
message broadcasts with a verified level of reli-
ability is much less than WTRP. WTRP does
not inherently provide acknowledgements for
transmissions, but it may be used in conjunction
with a higher layer that does (Lee, 2001). As-
sume that the group has M receivers and that a
message must be transmitted N times before the
probability that it is received and acknowledged
by all receivers reaches an acceptable level for the
The Role of Communications in Cyber-Physical Vehicle Applications
application. In this scenario, the source transmits
the message when it receives the token, waits for
each receiver to transmit an acknowledgement,
and, if needed, retransmits the message upon
regaining the token. WTRP requires (N+1)·M·∆T
to attain the acceptable level, where ∆T is the
average time to transfer the token. ∆T includes
the time to transmit the source message and other
control messages, including acknowledgements.
A similar nominal ∆T may used for both MRBP
and WTRP; but ∆T increases for WTRP if a token
is lost and ring reformation is required. In MRBP,
a source retransmits a message each ∆T if it does
not receive the acknowledgement. It takes at most
N·∆T for a token site to receive the source message
with the required reliability, at most N·∆T for each
receiver to recover the acknowledgement message,
and at most N·∆T for the receivers to recover the
acknowledged message, for a total of 3·N·∆T,
which is much less than (N+1)·M·∆T.
MRBP is the first reliable broadcast protocol
that operates with continuously changing groups.
It adds and deletes members to the group without
stopping the protocol. Earlier token passing proto-
cols are quasi-stationary. They operate on a fixed
group until the group changes, then reconfigure
the group and continue. MRBP is better suited for
vehicle applications where vehicles continuously
move with respect to one another.
DISTRIBUTED CONTROL
When highways are congested, platoons must
coordinate their actions, and the algorithms that
control the interaction between platoons have an
effect on safety and fuel economy that may rival
that of the protocols used within platoons. Central-
ized control procedures are not adequate to con-
trol all of the interacting vehicles on a congested
roadway. The number of vehicles is too large to
be processed quickly and the distances covered
increase the time to collect information.
In this section we discuss distributed tech-
niques in which vehicles collaborate on their
control without the direct involvement of a leader
or central authority. We first describe the evolu-
tion of an anti-lock braking application that uses
distributed techniques.
The distributed control procedures take into
account the actions of nearby vehicles and are
performed individually at each vehicle. Driv-
ers currently control their vehicles based on the
actions of several vehicles immediately ahead,
adjacent, and possibly behind them. Automated
systems improve on human performance by re-
sponding more quickly, and use communications
to take into account vehicles that are farther away
or which are obstructed from view by other traffic.
We define a vehicle’s neighborhood as the vehicles
it considers when performing an operation. We
also describe the Reliable Neighborcast Protocol
(RNP), which provides delivery and sequencing
guarantees in each neighborhood that are similar
to the guarantees provided by reliable broadcast
protocols within a broadcast group.
The neighborhoods of adjacent vehicles
overlap. On a congested highway, the actions
performed by one vehicle may indirectly affect
many others. The distributed control procedures
that we develop must not allow the actions taken
by a vehicle to propagate undamped. As an ex-
ample of how control procedures can malfunc-
tion we discuss “rolling traffic jams,” in which
vehicles that have not entered a highway are
affected by vehicles that have slowed down and
left the area.
On congested roadways and in densely
populated areas with many roadway entrances,
exits, and intersections, a vehicle’s neighborhood
changes frequently. We describe how RNP’s self-
organizing architecture can continuously modify
the neighborhoods on a highway.
In summary we compare the characteristics
of WTRP, MRBP, and RNP.
149
 The Role of Communications in Cyber-Physical Vehicle Applications
Distributed Anti-Lock Brakes
Distributed protocols will be used for a number
of assisted driving applications. Most applications
will initially provide warnings, then provide lo-
cal control functions, and eventually coordinate
maneuvers with adjacent vehicles.
For example, consider a distributed anti-lock
braking application. Currently sensors can de-
tect adjacent vehicles. These sensors are used to
provide a warning when a collision may occur.
Communications can improve the performance
of local sensors by sharing sensor readings be-
tween nearby vehicles to provide warnings about
vehicles that are farther away or obstructed from
view. For instance, communications can provide
a driver following a truck with a warning that a
car several ahead is braking hard. Communica-
tions can also provide warnings before sensors
can detect physical changes. Like virtual tail-
lights, vehicles nearby may receive a message
that a vehicle is braking before they can sense
its change in velocity.
In addition to providing a warning, a distrib-
uted anti-lock braking system can help control
a vehicle. When a vehicle senses or receives a
message that vehicle it is following is stopping,
the vehicle may automatically begin to brake.
The separation between vehicles can also be
transmitted to automatically control braking and
prevent tailgating.
Beyond local control, communications can be
used to negotiate a maximum safe rate of decel-
eration based on the braking distance of the least
capable vehicle in a string. In a negotiated stop, the
communications protocol should ensure that all of
the cooperating vehicles successfully receive the
message so that they react consistently. A braking
vehicle can be notified when the vehicles following
it have not received a message. Sequenced mes-
sage delivery is a useful protocol guarantee for
distributed control of vehicles. Software systems
that receive the same sequence of messages in each
vehicle are better able to predict the operation of
the other vehicles.
150
Drivers are reluctant to relinquish control of
their vehicles to automated systems. However,
the current generation of anti-lock brakes, which
uses sensors and computers to prevent vehicles
from skidding or braking unevenly, have gained
acceptance because drivers realize that the ability
of these systems to respond more quickly than
humans avoids accidents. A properly designed
distributed anti-lock braking system can also
gain acceptance by only providing functions that
reduce the likelihood and severity of accidents,
rather than assuming complete control of the
vehicle.
Rolling Traffic Jams
Many drivers have experienced rolling traffic
jams on congested highways. Traffic waves are
created when a vehicle on a congested roadway
changes speed suddenly. After a reaction delay,
the followers decrease their speed more quickly
to maintain the separation. Eventually the high-
way may come to a complete stop. Drivers that
are caught in these jams arrive at a point where
the traffic resumes its normal speed, without any
apparent cause for the delay. These waves actu-
ally move in the opposite direction of the traffic
flow so that the point where they are encountered
may not even be the point where the slow down
originally occurred. Rolling traffic jams can
persist for long periods of time and can even af-
fect vehicles that are not on the highway when
they first occurred. This makes it very difficult to
control the entire group of vehicles that encounter
these phenomena.
Sensors and automated control systems can
reduce rolling traffic jams by reducing driver
reaction times and over-reaction. This, in turn,
will reduce the amplification of the wave. Com-
munications can further reduce the size of the wave
by allowing vehicles to respond to the slow down
before it can be detected. This mode of operation
can attenuate, rather than amplify, the wave as it
moves away from the initial source of the distur-
The Role of Communications in Cyber-Physical Vehicle Applications
bance (Seiler et al., 2004). Distributed anti-lock
brakes can further reduce the phenomenon by
having vehicles decelerate less quickly than the
vehicles they are following. The safe distance be-
tween vehicles is a function of their speed. When
the vehicle in front is slowing down, the vehicle
following it can slow down less quickly and reduce
the spacing based on the new speed.
Communications and control technology can
also reduce rolling traffic jams by accelerating
a vehicle toward its target speed as soon as the
preceding vehicle starts to accelerate, instead of
waiting for the driver to respond to the change in
speed. Figure 5 depicts what happens after a lane
of traffic comes to a complete stop. The vehicles
are three meters in length and come to rest two
meters apart. The amount of time it takes for
vehicles in the lane to get beyond the point of the
disturbance after the first vehicle (i.e., Vehicle 1)
begins a 15 second acceleration back to 60 mph
is plotted for three cases. The main determinant
of waiting time is how long it takes each vehicle
to start accelerating after the vehicle in front of
it does. For example, using a typical human reac-
Figure 5. Waiting time in a traffic backup
140
Manual (t= 1 sec )
120 Automated (t= 0.2 sec )
Automated (Ideal)
100
Delay [sec]
80
60
40
20
0
0 10 20
tion time of one second, Vehicle 50 cannot begin
accelerating until the 49 vehicles before it do,
which takes 49 seconds. There is additional delay
then for Vehicle 50 to accelerate towards cruising
speed and reach the point of the disturbance. The
reaction time may be reduced using automated
cruise control, shown with a 0.2 second lag, or
further reduced by communications that permits
all vehicles to track their preceding vehicle’s ac-
celeration, thereby avoiding a cumulative delay.
Neighborcast
In order to implement distributed procedures,
each vehicle communicates with a set of vehicles
that are within a specified distance. The section
of the roadway within the specified distance
from a vehicle is a vehicle’s neighborhood, and
the vehicles within the neighborhood are the
vehicle’s neighbors, as shown in Figure 6. The
neighborhoods of adjacent vehicles may over-
lap. A protocol that implements communication
between a vehicle and its neighbors, rather than
within a specific set of vehicles, is called a neigh-
30 40 50 60 70 80 90 100
vehicle Position
151
 The Role of Communications in Cyber-Physical Vehicle Applications
Figure 6. Neighbors
borcast protocol (Maxemchuk et al., 2007). On
a congested highway, vehicles travel at similar,
but different, speeds and the set of neighbors that
interact continuously changes.
The size of a vehicle’s neighborhood depends
on the application, the road conditions, the speed
of the vehicles, the type of vehicles participat-
ing, the fraction of the vehicles that have similar
sensing and communications abilities, and the
communications procedures that are used. For
instance, consider a distributed anti-lock braking
application. On a dry road it is recommended that
manually controlled cars maintain a minimum
separation equal to the distance that they travel
in 2 seconds. At 60 mph, drivers should maintain
a separation of 180 feet. If the roads are icy or
wet the minimum spacing between cars should
increase; and the separation from a heavily loaded
truck to a car should be greater than between
cars, since the truck cannot stop as quickly. In
an automated braking system that uses sensors
to detect the distance to the preceding car, the
response time is faster than that of a driver, and
the safe separation may be reduced from 180 feet
to, say, 90 feet. In a system that uses communica-
tions to announce changes in speed, rather than
waiting for the change to be detected, the safe
distance may be further reduced by at least half
again. However, if, for example, the communi-
152
cations uses a token passing protocol that only
allows a vehicle to transmit once each second, the
distance a vehicle travels in one second should
be added to the separation between vehicles. In
addition, if some vehicles cannot participate in
the cooperative protocol, the average separation
will increase. The automated braking application
may use information from a neighborhood that
extends a certain distance in front of and behind
the vehicle, depending on the conditions.
RNP
Reliable neighborcast protocols, such as RNP
described in (Maxemchuk et al., 2007), provide
delivery guarantees to all neighbors. They are
similar to reliable broadcast protocols that pro-
vide delivery guarantees to all of the members
of a broadcast group. The difference between
the two is that each member of the neighborhood
has its own neighborhood, so that there isn’t a
single group.
RNP is an overlay on overlapping reliable
broadcast protocols, as shown in Figure 7. Each
vehicle participates in all of the reliable broadcast
groups that cover its location. The overlap guar-
antees that each vehicle can communicate with
every other vehicle in its neighborhood in at least
one of the broadcast groups covering its location.
The Role of Communications in Cyber-Physical Vehicle Applications
Therefore, vehicles can communicate with their
neighbors without forwarding messages through
multiple broadcast groups. RNP combines the
messages from the various broadcast groups, and
extends the guarantees provided by the underlying
reliable broadcast groups to the neighborhood.
When RNP is implemented on top of the MRBP
protocol, it provides guaranteed, sequenced and
confirmed message delivery to every vehicle in
the neighborhood. Sequenced message delivery
guarantees that any two vehicles that receive the
same two messages from vehicles in their neigh-
borhood will use those messages in the same order.
RNP filters out messages from other vehicles that
are in the broadcast groups but not in a vehicle’s
neighborhood and eliminates duplicate messages
when the same message is received in multiple
broadcast groups.
There are many ways to architect the under-
lying broadcast groups. At one extreme, a single
broadcast group may cover the entire highway:
Every vehicle is in the same broadcast group,
and is therefore in the same group as all of its
neighbors. With this architecture, each source
message is transmitted in a single broadcast group.
However, every message transmitted by every
vehicle on the highway must be received at every
other vehicle, even though most messages are not
from neighbors and will be discarded by RNP.
Figure 7. Overlapping broadcast groups
This mode of operation cannot be supported on
long, heavily populated highways by the limited
bandwidth of current wireless communications
networks. At the other extreme, a broadcast group
equal in size to the neighborhood may be created
about each vehicle. With this architecture, a vehicle
belongs to one broadcast group for each neighbor
in addition to one for itself. Each message from
a vehicle must be acknowledged in each of these
groups.
The efficiency of overlay architectures is de-
fined as the fraction of unique received messages
from neighbors over the total number of received
messages in (Maxemchuk et al., 2007). In the
most efficient architecture that has been found
for the linear highway applications, the size of the
broadcast group is approximately equal to the size
of a neighborhood, each broadcast group overlaps
the adjacent broadcast groups by half the size of
the neighborhood, and each vehicle belongs to at
most two underlying broadcast groups, so that
each source message is transmitted in at most two
broadcast groups. By definition, the efficiency of
using a single broadcast group for a highway will
approach zero, the efficiency of using one group
per vehicle is approximately 1/(k+1), where k is
the average number of neighbors, and the most
efficient architecture has an efficiency of ½.
153
 The Role of Communications in Cyber-Physical Vehicle Applications

In a practical system the overlap should be Self-Organizing Groups

larger than minimum overlap so that a moving
vehicle can join the underlying broadcast group of
other vehicles before entering their neighborhood.
In this way, all of the vehicles in a neighborhood
can communicate safely, even as vehicles change
neighborhoods and broadcast groups. The addi-
tional overlap must be large enough that the time
it takes a vehicle to cross the additional overlap
is greater than the time it takes the underlying
broadcast protocol to add the vehicle to the
broadcast group. Thus, its size is a function of
the vehicle speed and the underlying broadcast
protocol. In the case of MRBP, the time it takes
to add a node equals the time it takes the token
to cycle through all of the nodes in the group plus
several multiples of the maximum time allowed
to request and retransmit lost messages.
The underlying broadcast groups can be defined
to be stationary, be fixed at regular intervals on
highways, or move with the average speed of
vehicles in each broadcast group. The overlap is
larger if the broadcast group is fixed. For instance,
if it takes 10 seconds for the MRBP protocol to
add a new vehicle to a stationary broadcast group
and the vehicles move at 50 mph, the size of the
additional overlap must exceed 730 feet. However,
if the broadcast group moves at 50 mph, and all of
the vehicles travel between 45 and 55 mph, then
the overlap is only 73 feet. Since the time to add
a vehicle to an MRBP group is proportional to the
number of vehicles in the group, moving groups
that require less overlap, and therefore cover fewer
vehicles, require less time to add new ones.

Table 1. Protocol comparison

Characteristics
Target receivers
Medium access scheme
Reliability
Time until the target receivers
have successfully received a
source message
Overhead (typical control mes-
sages)
Suitable vehicular applications
WTRP
all vehicles in the token ring
collision-free message transmis-
sion
highly probable broadcast by
using repetitions
(N+1)·M·∆T
token (excludes acknowledge-
ment process)
communication within small,
infrequently changing groups
of vehicles on uncongested
roadways and in low message
loss environments. Also well
suited for transmitting periodic
updates.
MRBP
all vehicles in the broadcast
group
CSMA for source message
transmission and scheduled
token for acknowledgement
transmission
guarantee of message delivery
to all vehicles in the broadcast
group (commit process)
3·N·∆T (excludes commit
agreement policy)
token including votes and
message acknowledgements;
retransmission requests
communication within small,
frequently changing groups of
vehicles on congested roadways
and in high message loss
environments. Also suited for
transmitting infrequently occur-
ring emergency messages.
RNP (implemented with MRBP)
all vehicles in the neighborhood
CSMA for source message
transmission and scheduled
token for acknowledgement
transmission
guarantee of message delivery
to all vehicles in the neighbor-
hood (commit process from
MRBP and overlapping groups)
3·N·∆T (excludes commit
agreement policy)
adds vehicle and group infor-
mation to MRBP token; some
additional messages for adjust-
ing group size
communication with frequently
changing nearby vehicles on
congested roadways and in high
message loss environments.
Example applications are dis-
tributed anti-lock braking and
lane change negotiation

154
The Role of Communications in Cyber-Physical Vehicle Applications
Moving the underlying groups at the average
speed of the vehicles in the group requires a self-
organizing protocol, as described in (Maxemchuk
et al., 2007). In this protocol, when a vehicle
passes the token it also passes its location. If the
vehicle is at the edge of the group, the edge moves
with the vehicle (to the extent possible) while
maintaining the target overlap between adjacent
groups. When a group exceeds a maximum size,
it is split into two smaller groups with the proper
overlap between them. When non-adjacent groups
overlap, vehicles are covered by more than two
broadcast groups and two or more groups may
be combined into a single group.
A comparison of the characteristics of WTRP,
MRBP, and RNP implemented with MRBP is
shown in Table 1.
INTERACTION OF AUTOMOBILE
WITH TRAFFIC SIGNALS
Communications between automobiles and traffic
signals can (1) reduce the time spent waiting at
intersections, thereby reducing fuel consump-
tion, (2) decrease the probability of collisions by
informing cars when vehicles on a cross street
will share an intersection and (3) give precedence
to emergency vehicles entering an intersection.
The technology may progress from the use of
sensors to control a single traffic light and timers
to maintain the flow of traffic on an avenue, to
the use of communications and sensing to first
provide better control of individual lights and
then to coordinate nearby lights. Eventually com-
munications and control will be extended to the
automobiles approaching an intersection. Vehicles
will be organized into platoons and their arrival
at intersections will be scheduled to minimize
waiting times. In this section we will describe the
use of communications to control individual lights
and to coordinate nearby traffic signals.
Controlling Individual Traffic Lights
At present, sensors are used at intersections to de-
tect the presence of waiting cars so that the traffic
lights do not change unnecessarily. Communica-
tions can improve on local sensors by having the
automobiles, or distant sensors, notify the traffic
signals before vehicles arrive at the intersection.
A traffic signal may change in anticipation of
approaching vehicles or may delay a change in
the signal to allow vehicles that are approaching
to proceed without stopping.
Consider a lightly used roadway in which all
of the cars travel at 30 mph, accelerate uniformly
from 0 to 30 mph or decelerate uniformly from 30
to 0 mph in 5 seconds, and arrive at the intersec-
tion according to a Poisson process. On a lightly
used roadway, there is little interaction between
successive automobiles on the same roadway.
Define the intersection to extend from the point
where a car must start decelerating to stop at
the traffic light, to the point where it reaches 30
mph after stopping at the traffic signal. A car that
stops at the light spends 10 seconds plus the time
waiting at the light to traverse the intersection. A
car that does not stop traverses the intersection
in 5 seconds.
Consider an intersection where the major
roadway has four times as much traffic as the
minor roadway, and the traffic light has a fixed
timing. The timer allows traffic on the major road
to proceed four times as long as on the minor
road. If the light on the minor road is green or
amber for 10 seconds, then the light on the major
road will be green or amber for 40 seconds. The
light is amber for 5 seconds to allow vehicles that
enter the intersection to stop before the light on
the intersecting road turns green. To simplify the
calculation, assume that cars that approach the
intersection while the light is red do not slow down
if the light will turn green before they arrive at
the traffic signal. On the major road (4/5)ths of the
cars take 5 seconds to get through the intersec-
tion, because they do not stop, while (1/5) th of the
155
 The Role of Communications in Cyber-Physical Vehicle Applications
cars take an average of 15 seconds: 10 seconds
to come to a stop and accelerate, and an average
of 5 seconds waiting at a red light. On the minor
roadway, (1/5)th of the cars don’t stop and take
5 seconds to get through the intersection, while
(4/5)ths of the cars take 30 seconds, 10 seconds to
stop and an average of 20 seconds waiting. Since
(4/5)ths of the cars are on the major road and (1/5)th
are on the minor road, the average time to traverse
the intersection is 10.6 seconds.
Now consider a very lightly loaded road, with
the same traffic ratio, where the traffic light on
the major road is green unless a sensor detects
a car waiting on the minor road. When a car
on the minor road arrives, it stops, waits for 5
seconds while the traffic light on the major road
turns from amber to red, and then accelerates.
At very low traffic levels, most of the cars on
the major road arrive when the signal is green
and take 5 seconds to traverse the intersection,
while most of the cars on the minor road arrive
when the signal is red and take 15 seconds to get
through the intersection. The average time in
the intersection is 7 seconds. There is almost a
65% reduction in the time spent slowing down,
accelerating and stopped at the intersection. As
the traffic levels increase, cars on the major road
arrive at red signals and cars on the minor road
arrive when the signal is not red.
In a system that uses communications, the
traffic on the minor roadway will notify the
traffic light when the vehicle is 5 seconds away:
The light on the major road will turn amber, and
the light on the minor road will turn green as
the vehicle arrives at the intersection. On a very
lightly used road, all of the vehicles will travel
through the intersection in 5 seconds, and the time
spent decelerating, stopping and accelerating is
completely eliminated.
The 4-Way Stop Problem
On a lightly used road, cars arrive at and leave
the intersection independently, but as the traffic
156
becomes heavier, cars arrive at the intersection
in groups and cars stopped at the light must wait
for the cars in front of them to leave. The opera-
tion of the signals must consider the interaction
between the cars on the same roadway. The objec-
tive is control the right-of-way so as to minimize
the waiting time for cars entering from all four
directions. This scheduling problem is called the
4-way stop problem (O’Gorman, 2002).
In addition to controlling the right-of-way,
we can control the arrival times of cars at an
intersection. Cars are formed into platoons that
pass through the intersection. The platoon leaders
communicate with one another to agree on when,
and in what order, they will use an intersection
(Li & Wang, 2006). The scheduling objective
can change from minimizing the time in the
intersection to never stopping at an intersection.
In addition, when multiple vehicles are entering
an intersection and are at risk of collision, com-
munications can be used to force the cars to brake
to prevent the collision, or at least to reduce the
severity of an accident by forcing the collision to
occur at a lower speed.
Coordination of Traffic Lights
Controlling individual traffic signals is beneficial
in low traffic conditions or in rural areas with
a small number of traffic signals. In congested
cities, changing the schedule of a traffic signal at
an intersection also affects the traffic at nearby
intersections. The operation of signals must be
coordinated. At present, signals are scheduled to
allow traffic in the dominant traffic direction to
flow unimpeded. For instance, on Riverside drive
in the Manhattan borough of New York City, the
signals are coordinated so that traffic traveling
downtown at 30 mph during the morning rush
hour never stops for a red light, and traffic travel-
ing uptown during the evening rush hour has a
similar advantage.
The Role of Communications in Cyber-Physical Vehicle Applications
This strategy works well sometimes. Unfor-
tunately, it is not always possible to maintain 30
mph: Buses pick up and discharge passengers and
cars enter and leave parking spots, which reduces
the average speed. When there are travel disrup-
tions, the sequenced traffic lights may not provide
any advantage. Communications can be used to
determine the actual vehicle arrival times at traffic
signals, rather than using expected travel speeds to
estimate the arrival times, and adjust the schedule
of the lights to accommodate the traffic.
In addition, there are busy cross streets that
prevent this simple scheduling strategy from
working properly. For instance, on Riverside
Drive there are frequent traffic jams at 165th Street
during the evening rush hour. This street is par-
ticularly busy because a hospital is located at the
intersection and it is also a major merge for traffic
destined for the George Washington Bridge and
Yankee Stadium. A well attended evening game
at the stadium causes a several mile long traffic
jam around this intersection. The schedule of the
traffic light at the intersection can be improved
by the sensing and communications techniques
that have been described because the traffic jam
does not spread equally in all directions. By
detecting the traffic, the timing of the signal can
be adjusted to drain the traffic experiencing the
longest delays.
However, independently controlling this light
is not adequate. The traffic from this intersection
affects all nearby intersections. In addition to im-
proving the flow of traffic at this intersection, we
should use communications to improve the flow
of traffic at adjacent intersections. For instance,
when traffic cannot proceed on 165th Street, traffic
on the nearby streets that cross 165th Street should
be given precedence as long as the vehicles are
not turning into the congested street. This would
allow some traffic to leave the area and reduce the
congestion. By using the traffic signals to control
turns, we can create an orderly queue of vehicles
that wraps around several blocks before entering
the congested intersection.
In general, in any congested, developed area,
the operation of a traffic signal will affect nearby
signals. The group of signals that affect one an-
other can cover the entire city. It is not reasonable
to coordinate all of the traffic signals in a city
based on the instantaneous traffic. Instead, we
must develop distributed processing techniques
that control a signal based on the traffic in the
area and communications with nearby signals.
The control of traffic signals is similar to the
control of vehicles in the neighborcast protocol.
Each signal should cooperate with signals within
a certain distance. As the distance increases, the
need for cooperation decreases. However, the
signals that a signal cooperates with will have
their own neighbors with whom they cooperate.
Since the neighborhoods of adjacent traffic sig-
nals overlap, the effects of a change in a traffic
light can propagate over an area that is much
larger than its own neighborhood. Recent work
on coordinating traffic signals uses the signals to
organize cars into platoons and uses the arrival of
platoons to communicate between adjacent lights
(Gershenson, 2005).
COOPEATIVE ROUTE PLANNING
Global positioning systems (GPS) with mapping
software are currently used to plan routes in au-
tomobiles. Most of these systems perform all of
the calculations locally and do not require com-
munications. A number of services are becoming
available that use publicly available reports on road
construction and accidents to avoid roadways that
are likely to become congested. These services
require communications between the mapping
software in the automobile and a repository that
maintains and distributes information collected
from the public sources.
Communications can be used to develop data
on congestion by having automobiles report their
speed on the roadways that they traverse. Most
GPS units calculate speed as well as location. The
157
 The Role of Communications in Cyber-Physical Vehicle Applications
data can be used by mapping services to calculate
routes that follow the shortest distance, take the
least time, or use a metric that is a combination
of these and other factors to calculate the best
route.
The types of calculations that mapping ser-
vices perform will depend on the success of the
service. Initially, when a small number of drivers
use the service, the routes can be determined us-
ing conventional shortest path algorithms, with
each link on the path weighted with the selected
metric. There are many shortest path algorithms
and the calculations can be performed locally if
the mapping service distributes the metrics for
the nearby roadways.
As the services become more successful, the
routing software will have to take into account
the change in traffic that is caused by distribut-
ing congestion information. For instance, as-
sume that all of the automobiles have access to
the same mapping service, there are two equal
distance highways between points A and B, and
the service reports more congestion and longer
delays on one of the highways. If all of the auto-
mobiles change their route and take the highway
with less reported congestion, that highway will
become more congested. At the next report, all
of the traffic will be encouraged to move back to
the first highway, and the highway system will
become poorly used, with all of the traffic first
on one highway and then on the other.
This type of instability occurred in the ARPA-
net when reported delays were used to select paths
using a distributed version of the Bellman-Ford
algorithm. An example of this problem is presented
in (Bertsekas & Gallager, 1992, p. 374). One way
to prevent this type of instability is by sending a
portion of the traffic on each of the paths and only
moving a small fraction of the traffic from one
path to the other during a time interval, rather than
moving all of the traffic. The procedure is formal-
ized for general networks in the optimal routing
protocol described in (Bertsekas & Gallager,
158
1992, p. 433). There are centralized procedures
that quickly converge to optimal solutions and
distributed procedures based on the first deriva-
tives of the delay. The distributed calculations
can scale to arbitrarily large networks and can
be used for our system of roadways. Proportional
routing can be implemented by assigning prob-
abilities to the alternate paths, then using a random
procedure to select one of the paths according to
the probabilities.
The routing problem on highways is more
complicated than the routing problem in com-
munications networks. In a communications
network all of the traffic uses the same data and
obeys the routing decisions. In the automotive
application not all of the automobiles may have
access to the data, and we can expect competition
between several mapping services that use differ-
ent path selection algorithms. Mapping services
can recommend routes to the drivers, but cannot
force a driver to follow the recommendation. If
the software recommends that a driver follow a
route that will save a few minutes but the driver
is more comfortable driving the other route, the
driver may elect to follow the longer route. Finally,
the delay as a function of the link utilization in
networks is approximated by a simple mathemati-
cal form. In most instances, the delay is calculated
using Kleinrock’s independence approximation
(Kleinrock, 1964). The first derivative of the delay
with respect to the utilization in networks can be
calculated. The relationship between traffic and
delay on a roadway is more complicated.
In any case, the same proportional routing
procedures that can be used in networks can
be used on highways. The mapping service can
recommend that a fraction of its customers follow
alternate routes and modify the proportions by
small amounts to determine if the change improves
the operation of the roadways. If increasing the
fraction of the traffic that is directed to a road-
way improves the operation, another increase is
suggested. Otherwise, the fraction is decreased.
The Role of Communications in Cyber-Physical Vehicle Applications
CONCLUSON
The price of computing and communications has
decreased to the point where several computers
and wireless devices are present in many consumer
products. Sensor technology, although not as
advanced, is starting to appear in more products.
Increasingly these technologies will be used to
control physical devices in new ways, in what are
being called cyber-physical systems. Nowhere
are cyber-physical systems more important than
in our transportation networks. These applica-
tions will address the grand challenges to reduce
highway fatalities, decrease energy consumption,
and extend the lifetime and usefulness of the
transportation infrastructure.
In this chapter we have shown how these
objectives are being addressed for automobiles.
The techniques are equally useful to increase the
safety and capacity of airports, and to increase
the capacity and reduce the commute times for
buses and trains.
REFERENCES
Bertsekas, D., & Gallager, R. (1992). Data net-
works. Englewood Cliffs, NJ: Prentice Hall.
C2C-CC. Retrieved June 8, 2008, from: http://
www.car-to-car.org
CarTALK 2000. Communication Architecture.
Retrieved June 8, 2008, from: http://www.car-
talk2000.net
Chang, J.-M., & Maxemchuk, N. F. (1984). Reli-
able Broadcast Protocols. ACM Transactions on
Computer Systems, 2, 251-273.
ERTICO-ITS Europe. PReVENT Project. Re-
trieved June 8, 2008, from: http://www.prevent-
ip.org
BMBF. Network on Wheels. Retrieved June 8,
2008, from: http://www.network-on-wheels.de
Gehring, O., & Fritz, H. (1997). Practical results
of a longitudinal control concept for truck pla-
tooning with vehicle to vehicle communication.
IEEE Conference on Intelligent Transportation
Systems (pp. 117-122).
Gershenson, C. (2005). Self-Organizing Traffic
Lights. Complex Systems, 16(1), 29-53.
Halle, S., Laumonier, J., & Chaib-Draa, B. (2004).
A decentralized approach to collaborative driving
coordination. Proceedings of the 7th International
IEEE Conference on Intelligent Transportation
Systems, Washington, DC (pp. 453-458).
Heddebaut, M., Rioult, J., Ghys, J. P., Gransart,
Ch., & Ambellouis, S. (2005). Broadband vehicle-
to-vehicle communication using an extended
autonomous cruise control sensor. Measurement
Science and Technology, 16, 1363-1373.
Horowitz, R., & Varaiya, P. (2000). Control design
of an automated highway system, Proceedings of
the IEEE, 88(7), 913-925.
ISO TC 204 Working Group 16. (2007). CALM
Continuous Communications for Vehicles. Re-
trieved June 8, 2008 from: http://www.calm.hu
IEEE. DSRC Home. Retrieved June 8, 2008 from:
http://grouper.ieee.org/groups/scc32/dsrc/index.
html
IEEE. (2006). IEEE Trial-Use Standard for
Wireless Access in Vehicular Environments
(WAVE) - Multi-channel Operation. IEEE Std
1609.4-2006.
Kleinrock, L. (1964). Communication nets: Sto-
chastic message flow and delay. New York, New
York: McGraw-Hill.
Lee, D., Attias, R., Puri, A., Sengupta, R., Tripakis,
S., & Varaiya, P. (2001). A wireless token ring
protocol for intelligent transportation systems.
Proceedings of IEEE Intelligent Transportation
Systems Conference (pp. 1152-1157).
159
 The Role of Communications in Cyber-Physical Vehicle Applications
Lee, E. A. (2008). Cyber-Physical Systems: Design
Challenges. 11th IEEE International Symposium on
Object Oriented Real-Time Distributed Comput-
ing (pp. 363–369).
Li, F., & Wang, Y. (2007). Routing in vehicular ad
hoc networks: A survey. IEEE Vehicular Technol-
ogy Magazine, 2(2), 12-22.
Li, L., & Wang, F.-Y. (2006). Cooperative Driv-
ing at Blind Crossings Using Intervehicle Com-
munication. IEEE Transactions on Vehicular
Technology, 55, 1712-1724.
Liu, Y., & Ozguner, U. (2003). Effect of inter-
vehicle communication on rear-end collision
avoidance. Intelligent Vehicles Symposium (pp.
168-173).
Maxemchuk, N. F., & Shur, D. (2001). An Internet
multicast system for the stock market. ACM Trans-
actions on Computer Systems, 19, 384-412.
Maxemchuk, N. F., Tientrakool, P., & Willke, T. L.
(2007). Reliable Neighborcast. IEEE Transactions
on Vehicular Technology, 56, 3278-3288.
NITRD. National Workshop on High-Confidence
Automotive Cyber-Physical Systems (2008). Re-
trieved September 12, 2008, from http://varma.
ece.cmu.edu/Auto-CPS/
National Science Foundation (2006). NSF Work-
shop on Cyber-Physical Systems. Retrieved
September 12, 2008, from http://varma.ece.cmu.
edu/cps/
O’Gorman, E. (2002). Using group communica-
tion to support inter-vehicle coordination. Un-
published master’s thesis, University of Dublin,
Ireland.
Reichardt, D., Miglietta, M., Moretti, L., Morsink,
P., & Schulz, W. (2002). CarTALK 2000: Safe and
comfortable driving based upon inter-vehicle-
communication. Proceedings of the IEEE Intel-
ligent Vehicles Symposium (pp. 545-550).
160
Seiler, P., Pant, A., & Hedrick, K. (2004). Distur-
bance propagation in vehicle strings. IEEE Trans-
actions on Automatic Control, 49, 1835–1842.
Sichitiu, M. L., & Kihl, M. (2008). Inter-Vehicle
Communication Systems: A Survey. IEEE Com-
munications Surveys & Tutorials, 10(2), 88-105.
Song, B., & and Hedrick, J. K. (2004). Design
and experimental implementation of longitudinal
control for automated transit buses. Proceedings
of the American Control Conference (pp. 2751-
2756).
Swaroop, D. (1995). String stability of intercon-
nected systems: An application to platooning
in automated highway systems. Unpublished
doctoral dissertation, University of California,
Berkeley, CA.
Tokuda, K., Akiyama, M., & Fujii, H. (2000).
DOLPHIN for inter-vehicle communications sys-
tem. Proceedings of the IEEE Intelligent Vehicles
Symposium (pp. 504-509).
U.S. DOT. Intelligent Vehicle Initiative. Retrieved
July 4, 2008, from http://www.its.dot.gov/ivi/ivi.
htm
Vahidi, A., & Eskandarian, A. (2003). Research
advances in intelligent collision avoidance and
adaptive cruise control. IEEE Transactions on In-
telligent Transportation Systems, 4(3), 143-153.
VSCC. VSC Project Overview. Retrieved June
8, 2008, from http://www-nrd.nhtsa.dot.gov/pdf/
nrd-12/CAMP3/pages/VSCC.htm
Willke, T. L., & Maxemchuk, N. F. (2005). Co-
ordinated interaction using reliable broadcast in
mobile wireless networks. Proceedings of the 4th
International IFIP-TC6 Networking Conference,
NETWORKING 2005, Waterloo, Canada (pp.
1168-1179).
Willke, T. L., Tientrakool, P., & Maxemchuk, N. F.
(2008). A Survey of Inter-Vehicle Communication
The Role of Communications in Cyber-Physical Vehicle Applications

Protocols and Their Applications. IEEE Commu-
nications Surveys and Tutorials. To appear.
Wolf, W. (2007). The Good News and the Bad
News. Computer, 40(11), 104-105.
Zhang, Y., Kosmatopoulos, E. B., Ioannou, P. A., &
Chen, C. C. (1999). Autonomous intelligent cruise
control using front and back information for tight
vehicle following maneuvers. IEEE Transactions
on Vehicular Technology, 48, 319-328.

161
162

Chapter IX
Integrating TrafILc Flow
Features to Characterize the
Interference in Vehicular
Ad Hoc Networks
Lili Du
Purdue University, USA

Satish Ukkusuri
Rensselaer Polytechnic Institute, USA

Shivkumar Kalyanaraman
Rensselaer Polytechnic Institute, USA

ABSTtract

Vehicular Ad Hoc Networks (VANETs) are composed of vehicles equipped with advanced wireless commu-
nication devices. As a paradigm of decentralized advanced traveler information systems (ATIS), VANETs
have obtained interests of researchers in both communication and transportation fields. The research
in this chapter investigates several fundamental issues, such as the connectivity, the reachability, the
interference, and the capacity, with respect to information propagation in VANETs. The authors’ work is
distinguished with previous efforts, since they incorporate the characteristics of traffic into these issues in
the communication layer of VANETs; this mainly addresses the issue of the interference. Previous efforts
to solve this problem only consider static network topologies. However, high node mobility and dynamic
traffic features make the interference problem in VANETs quite different. To investigate this problem,
this chapter first demonstrates the interference features in VANETs incorporating realistic traffic flow
features based on a validated simulation model. Then, analytical expressions are developed to evalu-
ate the interference under different traffic flow conditions. These analytical expressions are validated
within the simulation framework. The results show that the analytical expressions perform very well to
capture the interference in VANETs. The results from this work can facilitate the development of better
algorithms for maximizing throughput in the VANETs.

Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.
Integrating Traf.c Flow Features to Characterize the Interference in V ehicular Ad Hoc Networks
Introduction
With increased urbanization, there are increasing
concerns about congestion and severity of surface
transportation incidents. It is no longer possible to
address the issue of congestion management by
adding new transportation infrastructure due to
the significant costs involved and the impacts on
land use. Transportation agencies in the United
States and in other parts of the world are explor-
ing innovative technology oriented methodolo-
gies which can alleviate congestion and safety
problems. Recent years, Intelligent Transportation
Systems (ITS) incorporated advanced informa-
tion technologies have been claimed to perform
efficiently. Among such systems, ATIS is consid-
ered to be a promising technology for improving
traffic conditions by helping travelers use existing
transportation facilities efficiently, and interpret
the real-time traffic information correctly (Dia
& Purchase, 1999; Jin & Recker, 2006; Lo &
Szeto, 2004; Mouskos et al., 1996; Srinivasan &
Mahmassani, 2003; Wang, 2007).
Some systems are already in place, such as
Copilot in the U.S. and TomTom in Europe. All
of these systems rely on a centralized system,
such as the Traffic Management Center (TMC)
to gather traffic information from probe ve-
hicles or roadside sensors, generate route choice
messages and disseminate route suggestions to
drivers. However, the structure and mechanisms
of these existing centralized approaches suffer
from significant disadvantages, such as heavy
infrastructure requirements, large computational
needs, and high initial investment cost and so on.
Hence, the current centralized ATIS is not fully
satisfactory.
At the same time, the rapid advances in wire-
less communication technology make the tasks
of collecting, interpreting, and disseminating
information among vehicles feasible. ATIS
based on the inter-vehicular communication is
referred as to decentralized ATIS. Compared
with the current centralized ATIS, it has several
advantages: (i) they are infrastructure-light in
that they do not rely on roadside sensors and traf-
fic management centers. Instead, they exchange
and collect traffic information by inter-vehicular
communication; (ii) the decentralized system is
robust in emergency and disaster related situations
because it is self-organized and independent of
fixed-infrastructures. (iii) Information exchange
in the decentralized system can be used for other
applications and file sharing such as 511.
Our study investigates vehicular ad hoc
networks, in which vehicles serve as data col-
lectors and anonymously transmit traffic and
road condition information from every major
road within the transportation network. Such
information will provide transportation agen-
cies with the information needed to implement
active strategies to relieve traffic congestion and
improve safety. Several fundamental problems
such as the connectivity, the reachability, the
capacity, and the interference in VANETs are
explored in the related study before. These is-
sues demonstrate the information propagation in
VANETs from different views. Connectivity and
reachability address the information propagation
opportunities between any node pairs in VANETs
at any time snapshot and in a short time interval
respectively. The capacity defined in this study
demonstrates the information dissemination
capability in VANETs, which is limited by the
interference between concurrent transmissions.
Here, we briefly introduce the first three topics and
then emphasize on the interference in VANETs,
which is the main contribution of the proposed
work in this chapter.
The rest of this chapter is organized as fol-
lows: section 2 introduces VANETs. Section 3
presents our research scope. Section 4 reviews
the previous work about the interference in wire-
less network. Section 5 provides the background
and assumptions for the study of the interference
here. Section 6 describes our methodology and
main results. Section 7 summarizes the results
of this study.
163
 Integrating Traffic Flow Features to Characterize the Interference in Vehicular Ad Hoc Networks
VEHhicular AD HOC NETtwork
Vehicular Ad Hoc Networks (VANETs), which
are composed of vehicles equipped with advanced
wireless communication instruments, such as
Global Position Systems (GPS) receivers and Per-
sonal Digital Assistant (PDA) based cell phones, is
a promising paradigm of the decentralized ATIS.
This infrastructure-light and decentralized system
is expected to provide tremendous benefits for
alleviating traffic congestion, improving traffic
safety, and further enhancing many other aspects
of our life.
Nowadays, worldwide academia, industry, and
governments are investing significant amounts of
time and resources on studying, deploying, and
testing the performance of VANETs. Several
academic activities for VANETs were initiated
such as ACM International Workshop on Ve-
hicular Ad Hoc Networks with MOBICOM and
Vehicle-to-Vehicle Communications Workshop
(V2VCOM) with MobiQuitous. Many on-going
national/international projects included consor-
tia such as the Vehicle Safety Consortium (US),
Car-2-Car Communication Consortium (Europe)
and Advanced Safety Vehicle Program (Japan);
standardization efforts like the Dedicated Short
Range Communications (DSRC) and IEEE
802.11p (WAVE), and field trials like the large-
scale Vehicle Infrastructure Integration Program
(VII) in the US.
These initial works found that VANETs pos-
sess unique characteristics as compared to general
mobile ad hoc networks (MANETs). For instance,
high vehicular speed and unexpected driver ac-
celeration or deceleration makes the topology
of VANETs change much faster. The number
of vehicles in VANETs cannot be scheduled or
controlled in real-time because drivers will enter
or exit at any time. These features further bring
forth challenges in VANETs as below:
1. The high vehicular mobility makes VANETs
subject to frequent fragmentation. Conse-
164
quently, information dissemination becomes
even difficult in VANETs as many informa-
tion routing paths become disconnected
before they are utilized. Therefore, building
a stable connected VANET is a difficult but
critical task.
2. The varying traffic flow characteristics in
the transportation network could possibly
lead each vehicle to experiencing different
VANETs in a single trip. Hence, an adaptive
and distributed topology control strategy
should be developed for individual vehicles
so that the inter-vehicle communications of
VANETs perform well.
3. The poor quality of data associated with
link/edge latencies and possibility of varia-
tion between the data obtained from different
vehicles make developing a traffic routing
algorithm even harder. This online traffic
routing algorithm is necessary to build an
advanced framework for the decentralized
ATIS (Gao & Chabini, 2006; Jaillet & Wag-
ner, 2006; Shavitt & Sha, 2005; Waller &
Ziliaskopoulos, 2002).
4. The possibility of malicious messages and
the accessibility of any traveler’s origin
destination information can pose a serious
security and privacy problem. New ap-
proaches for communication security have
to be designed to guarantee the reliable and
secure services in VANETs.
5. All nodes sharing the same medium access
channel leads to congestion in very dense
networks. Moreover, unstable link connec-
tions only provide limited bandwidth. Both
of these unfavorable conditions in VANETs
result in low information transmission ef-
ficiency. Consequently, advanced medium
access protocols are needed to improve the
level of information transmission efficiency
in VANETs.
6. Neither the simulation tools in transporta-
tion, such as PARAMICS, CORSIM and
VISSIM, nor the simulation platforms in
Integrating Traffic Flow Features to Characterize the Interference in Vehicular Ad Hoc Networks
communication networking such as ns2,
OPNET and Qualnet can test and evalu-
ate network protocols in VANETs. A new
simulation platform, which integrates
communication techniques and realistic
transportation simulation, is desirable.
These application potential and research
challenges aroused many interests of scholars
in both the transportation area and the wireless
communication area. Previous simulation works
(Artimy et al., 2005a, 2004, 2005b, 2006) present
the following points: VANETs are usually mod-
eled as one-dimensional networks; a dynamic
transmission range is expected to provide better
connectivity; the connectivity of a VANET on
the road with one lane is more sensitive to vehicle
velocity than a VANET on the multi-lane road-
way and so on. Note that the interaction between
vehicular mobility and information propagation
is usually studied by simulation methods, but
analytical work is less present (Jin & Recker, 2006;
Wang, 2007); most of the models for VANETs
are built on stable traffic flow, but the unstable
feature of traffic flow is not considered. Encour-
aged by the above points, we conduct analytical
study on the information propagation in VANETs
incorporating the traffic flow features, which will
be discussed next.
Our STUDY SCOPE
In an effort to address some of the limitations
of the previous work, this research investigates
several properties of VANETs in consideration of
the traffic flow characteristics. In particular, we
address the geometric connectivity, reachability,
capacity, and interference respectively.
The first topic regards the analytical expres-
sions to characterize the connectivity of VANETs
on freeway segments, which represents the prob-
ability that any two vehicles are connected at a
given time instant (Ni & Chandler, 1994). Spe-
cifically, we considered a VANET as a nominal
system with uncertain disturbance. The nominal
system was represented by a free traffic flow, in
which space headway was assumed to obey an
exponential distribution. The unexpected driver
behavior such as acceleration, deceleration, and
lane changing were modeled as uncertain dis-
turbance, which was further characterized by a
robustness factor in our analytical model. Our
regression results showed that the robustness
factor is a function of the traffic flow parameters
including average traffic speed, average space
headway and its variance. The simulation vali-
dation demonstrated that our analytical expres-
sion can evaluate the geometric connectivity of
VANETs more accurately than previous efforts in
literature. The readers are referred to the complete
analytical characterization of connectivity which
is discussed in (Ukkusuri & Du, 2008).
Due to the observation that the relative
movement between individual vehicles can cre-
ate opportunistic connections between vehicles
during a short time interval, connectivity, which
demonstrates the opportunities that the infor-
mation can go through the whole network at a
given time instant, is sometimes not sufficient
to fully understand the information propagation
performance. We dealt with this problem in the
second part of our research. We first defined an
“information flow network” and then introduced
“reachability” to characterize information propa-
gation performance in a short time period. An
information flow network is a time expanded
graph composed of asynchronous communication
links (based on geometric distance) and nodes
(vehicles). The reachability is the probability
that every two vehicles in the information flow
network are connected in a given time interval.
To capture various driver behaviors, we separated
the drivers into three clusters which were aggres-
sive, defensive, and slow drivers respectively.
Correspondingly, we approximated the relative
movement between individual vehicles by the
relative movement between different driver clus-
165
 Integrating Traffic Flow Features to Characterize the Interference in Vehicular Ad Hoc Networks
ters. Based on this approximation, we developed
analytical formula to evaluate the reachability
during a short time period. Our results showed that
the relative movement between vehicles enables
individual vehicles to communicate with more
neighbors and therefore improves the opportu-
nity that the traffic information is transmitted
forward. Simulations were setup to validate our
assumptions and analytical results. The reader
is suggested to reach our completed work by Du
and Ukkusuri (2008b).
While the first two topics focused on informa-
tion propagation opportunities, the third topic
considered the information dissemination among
vehicles. Specifically, we studied the maximum
concurrent transmissions in VANETs, which is
referred to as the capacity of VANETs in this
study. We first explored the capacity of VANETs
using an integer programming (IP) formulation.
Since the IP model is computationally hard to
solve, we further developed a statistical model
to characterize the capacity of VANETs in terms
of some significant parameters in the traffic flow
networks as well as the communication networks.
To improve the prediction accuracy of the statis-
tical model, the central composite experiment
design method was applied. The interested reader
can also reach the completed work by Du and
Ukkusuri (2008a).
The next problem that we are interested in is
the interference in VANETs considering the rela-
tive movement between individual vehicles. This
problem will be discussed deeply in the following
paragraphs in this chapter.
IntERrfinANETS
Informally speaking, interference can be modeled
to a limited extent under the following assumption:
a transmission from node u to node v is success-
ful only if there is no other node w which has a
connection to node v and transmitting to node v
simultaneously (Burkhart et al., 2004; Rajara-
166
man, 2002). Along the line of the above model,
this study defines the interference as the number
of reachable vehicles around each individual ve-
hicle (two vehicles are in the transmission range
to each other).
To understand interference clearly, we need to
introduce two terms, hidden and exposed nodes,
which are shown in Figure 1. Hidden nodes are
two nodes that share the set of nodes within the
transmission ranges of both, but they are not in the
transmission range of one another. For example,
in Figure 1 left, node A and node C are not in
the transmission range of each other, but node B
is in the intersection of node A transmission and
node C transmission. Therefore, node A and node
C cannot transmit information to node B at the
same time, otherwise collisions will happen on
B. Exposed nodes are two nodes which are within
the transmission range of one another, thus, they
cannot transmit information at the same time, and
otherwise information collisions happen to both of
them. As the example shown in Figure 1 right, if
node A and node C transmit information to node
B and node D respectively at the same time, then
the collisions will not happen on B and D, but node
A and node C are within the transmission range
of each other, they will interfere with each other.
Hence, only one of node A and node C can be ac-
tive at a given time (Blum et al., 2004). Based on
the description above, it is clear that interference
is closely related to the node distribution and the
transmission range.
Two direct reasons encourage us to investigate
the interference of VANETs. First, interference is
closely related to the trade-off between through-
put11 and connectivity. (They together influence
the performance of VANETs significantly). On the
one hand, in order to reduce the interference (i.e.
increase throughput), the small transmission range
is preferred. On the other hand, there is surely a
limit of decreasing the transmission range, while
maintaining network connectivity. This trade-
off makes the interference very important to the
topology controlling in VANETs.
Integrating Traffic Flow Features to Characterize the Interference in Vehicular Ad Hoc Networks
Figure 1. Hidden nodes and exposed nodes
A B C D
Second, vehicular mobility, unexpected driver
behavior, and variable traffic flow condition make
the interference that individual vehicle experi-
ences in VANETs quite different with nodes
in static networks. Specifically, the topology of
VANETs changes quite rapidly from time to time,
hence the number of neighbors which interferes
the transmission of one vehicle is not constant.
Moreover, the topology change depends on the
macroscopic traffic condition, such as congestion
level, as well as the microscopic traffic behavior,
such as driver’s acceleration, deceleration and
lane changing behavior. Hence, measuring the
dynamic interference of VANETs in view of
vehicular mobility as well as traffic flow features
is a challenging and important work. To the best
of our knowledge, very little work has been done
on this topic.
In view of above points, the following study
is dedicated to addressing the interference of
VANETs on a freeway segment from the trans-
portation point of views. Our main contributions
are in two folds: (1) based on simulation data, we
explore the interference that individual vehicles
experience under different traffic flow condi-
tions; (2) incorporating both microscopic and
macroscopic traffic features, we develop closed
form stochastic expressions to characterize the
interference in VANETs.
RELlatd WORK
Extensive work has been done for the interference
in static wireless network. Most of the previous
work addressed the interference issues implicitly
A C B
by constructing topologies satisfying the features
of sparseness or low node degree, such as (Hou
& Li, 2006; Li et al., 2003; Ramanathan & Ro-
sales-Hain, 2000). However, recent work revealed
that the previous implicit notion of interference
is not sufficient to reduce the actual interference
(Burkhart et al., 2004). The following review the
efforts on the explicit interference reduction in
wireless network.
Heide et al. (2002) introduced an explicit
definition of interference based on the current
network traffic. The weakness of this definition
is the requirement of a priori network traffic in-
formation, which is usually not available. With
the assumption that each node can adapt their
transmission power, they explore the trade-off
between the congestion, power consumption, and
dilation in a wireless network.
Burkhart et al. (2004) proposed another explicit
definition: the interference of a link (u, v) is the
number of nodes covered by two disks centered
at node u and node v with transmission range r.
Formally, the definition is shown below:
con(uv) = {w | w is covered by D(u | uv) or D(v | uv)}
(1)
where, cov(uv) denotes the set of nodes that can
be affected by node u and node v when they
communicate to each other with exactly the
minimum power needed to reach one another.
|uv| is the distance between node u and node v.
D(u; r) denotes the disk centered at node u with
transmission range r. This interference definition
is also called interference based on the coverage
model. Using this definition, the authors disprove
167
 Integrating Traffic Flow Features to Characterize the Interference in Vehicular Ad Hoc Networks
the widely advocated assumption that sparse
topologies imply low interference. Furthermore,
they propose a centralized algorithm to compute
an interference-minimal connectivity preserving
topology. Nejad and Li (2005) adopt this definition
and further develop another algorithm to construct
the network topology so that the maximum inter-
ference is minimized.
Inspired by Fussen et al. (2004), Rickenbach et
al. (2005) questioned the definition in (Burkhart et
al., 2004) from two aspects: (i) It is sender-centric,
i.e., the interference is considered to be an issue at
senders instead of at receivers. Hence, this defini-
tion hardly reflects the real-world interference.
(ii) It is of more technical nature and not robust
enough to withstand the addition or removal of an
individual node in the network. In contrast to this
sender-centric interference definition, the authors
present a receiver-centric interference model: the
interference of a node u demonstrates the number
of nodes covering node u with their disks induced
by their transmission range reachable to node u.
Furthermore, an algorithm with approximation
ratio 4√∆(∆ is the maximum node degree) to find
the optimal connectivity-preserving topology is
proposed.
Since the coverage area of omni-directional
antenna does not have a clear-cut boundary,
Moscibroda and Wattenhofer (2005) studies the
average interference problem, where the nodes are
grouped into active nodes (A) and passive nodes
(P). The interference of a node p ≠  P is the number
of nodes a ≠  A whose transmission ranges cover
p. Correspondingly, the average interference of a
topology is the sum over all interference divided by
the number of passive nodes. A greedy algorithm
is proposed to compute an O(logn) approxima-
tion to the connectivity-preserving interference
problem, where n is the number of nodes in the
network.
So far, the studies for the interference reduc-
tion in wireless networks are based on the static
network topology. Due to the high vehicular
mobility, the topology of VANETs changes fre-
168
quently. This leads to the results and algorithms
of previous works possibly not applicable any
more, but there are few efforts on investigating
the dynamic interference in VANETs. Therefore,
exploring new interference models for VANETs
is necessary.
Sy Mod
We begin our study with some assumptions. We
consider VANETs with n vehicles equipped with
wireless communication instruments moving on
a freeway segment. Vehicles are homogenously
distributed with traffic density λ. The commu-
nication radio is with the directional antenna,
and referring the traffic density, its transmission
range r can be adjusted adaptively so that certain
number neighbors can be covered (Blough et al.,
2006; Xue & Kumar, 2004). Let ni denote the
vehicles i, then the distance between vehicles
ni and nj is denoted by xij. When xij < r, we call
node j is the neighbor of node i, vice versa. For
the consecutive vehicles, xij is also referred to as
space headway.
Medium access control. Due to the time vary-
ing network topology and the lack of centralized
control in VANETs, we assume medium access
control applies the spirit of a random access
scheme. Once the distance between two vehicles
is less than their transmission range, the com-
munication between them is triggered, but only
successes if there is no interference. Additionally,
we suppose a node may not send and receive mes-
sages at the same time but can transmit to more
than one other node at the same time. Specifi-
cally, a transmission is successful if both of the
following conditions are satisfied: (1) xij ≤ ri, (2)
any other node nk, such that xik ≤ rk or xjk ≤ rk is
not transmitting (Jain et al., 2003).
Preliminary interference calculation. We as-
sume that each vehicle has the same numbers of
reachable-neighbors2 (Blough et al., 2006; Xue &
Kumar, 2004), and Kc = λr. With these assump-
Integrating Traffic Flow Features to Characterize the Interference in Vehicular Ad Hoc Networks
tions, we immediately find that under the static
situation, the expected interference is equal to
3Kc. From Figure 2, we observe that there are Kc -
neighbors in the forward and backward direction
which will interfere the transmission of vehicle A.
Additionally, vehicle A itself will averagely cover
Kc, therefore, there is one more Kc -neighbors will
interfere the transmission of vehicle A. However,
due to the high mobility of vehicles in VANETs,
the network topology will change very quickly.
Consequently, there is a high probability that
other vehicles will move in or out the transmis-
sion range of the vehicles. This results in more
than Kc neighbors to interfere the transmission
of an individual vehicle in a short time interval.
This critical characteristic of the interference in
VANETs has not been considered in the past ef-
forts. The goal of this study is to further explore
the interference of one successful transmission
in a short time period, T, incorporating traffic
flow conditions and relative movements between
vehicles.
Expct INTERFERENCEnc
undr DYNAMIC Traffic
Condition
We now present our framework for modeling the
interference in VANETs. We first do a simulation-
based study, and then we present our method to
measure the interference incorporating traffic flow
characteristics. Both the macroscopic vehicle dis-
tribution and the microscopic relative movement
between vehicles are considered in developing
Figure 2. The example of Kc = 3
Kc=3
9 7 5 3 A 1
  
  
Kc
the models. Finally, we validate our stochastic
models of interference in VANETs using a well
calibrated simulation model.
A Smulation-Based Study
To investigate how the influence of relative
movement and the traffic flow congestion level
impact the interference of VANETs, we first per-
form a simulation-based study based on an open
source dynamic traffic simulation framework-
MITSIMLab (Ben-Akiva et al., 2002), which
is developed by MIT Intelligent Transportation
System Program. We setup a one-way freeway
segment, which is about 10000 ft long and has
three lanes. Vehicles have different speed at
each time instant but their speed limit is 65mph.
Each simulation is conducted from 8:00am to
8:25am under constant traffic demand. The first
15 minutes is redeemed as warm-up time and
data is collected from 8:15am to 8:25am. In order
to obtain traffic flow under different congestion
level, simulations with demand rates changing
from 50vph (vehicle per hour) (very light traffic
condition) to 8000vph (heavy congestion traffic
condition) are conducted. Simulation outputs
provide us individual vehicle position, speed and
traffic density, λ per second.
In order to demonstrate the effect of relative ve-
hicular movement on the interference of VANETs,
we study the interference of individual vehicles
in a short time interval, say 30s. In addition, our
preliminary interference analysis shows that the
number of reachable neighbors for individual
vehicles is the key influencing variable to measure
2 4 6 8 10 12
Kc Kc
169
 Integrating Traffic Flow Features to Characterize the Interference in Vehicular Ad Hoc Networks

the interference (equal to 3Kc). Therefore, the fol-
lowing work focuses on exploring the reachable
neighbors of individual vehicles.
We name the average number of reachable
neighbors for an individual vehicle under a static
VANET topology as Static Vehicular Degree
(SVD), but the average number of reachable
neighbors for an individual vehicle in a time
interval is referred to as Time-period Vehicular
Degree (TVD). Without loss of generality, assum-
ing SVD=3, and vehicular transmission range,
r = SVD/λ at each time interval, we check SVD
during the interval T (=30s)3.
The results reported in Figure 3 show that
when we increase the traffic demand step by
step, vehicles become more and more crowded.
Consequently, the reachable neighbors of indi-
vidual vehicles in 30s increase. This means that
the probability of the interference will impact the
transmission of individual vehicles. Moreover,
carefully studying the curves in Figure 3, we
obtain more insights:
1. Traffic Demand <500vph: Figure 4 shows
traffic is sparse; Figure 3 shows SVD is
smaller than SVD. This is reasonable since
there are not enough neighbors around in-
dividual vehicles in a short time interval
2. 500vph < Traffic Demand < 3700vph: Fig-
ure 4 shows traffic is still in the free flow
regime; Figure 3 shows TVD is close to 4,
which is slightly higher than SVD. This
phenomenon shows the relative movement
between vehicles increases the opportunity
that vehicles are in the transmission range
of each other. Since the traffic is still light,
SVD is only a little higher than the SVD.
3. 4000vph < Traffic Demand < 6200vph:
Figure 4 shows traffic is light congested;
Figure 3 shows TVD reaches to the highest
value. This results from higher traffic density
and relative vehicular movement together.
4. Traffic Demand > 6600vph: Figure 4 shows
traffic becomes very congested; Figure 3
shows that TVD degrades. The reason is that
when the traffic becomes heavy congestion,
the relative movement between vehicles is
limited. Consequently, fewer vehicles will
show up around an individual vehicle close
enough to interfere its transmission.
The above observations clearly demonstrate the
interaction between the interference of VANETs
and the features of traffic flow, such as the mi-

Figure 3. The number of reachable neighbors of individual vehicles in a time interval T = 30s under
different traffic flow

8
7
6
Simulation Degree

5
4 DVD
3 SVD
2
1
0
0

0
50

00

00

00

00

00

00

00
10

90
17

19

37

46

52

60

68

80

De ma nd

170
Integrating Traffic Flow Features to Characterize the Interference in Vehicular Ad Hoc Networks
croscopic vehicular relative movement and the
macroscopic traffic flow state. In the next section,
we further explore the relationship between the
interference of VANETs and traffic flow features
by analytical methods.
Investigating Time-Period
Vehicular Degree Analytically
Our analytical work first investigates the TVD
of VANETs incorporating the microscopic traf-
fic flow features, such as the relative movement
between vehicles. The VANETs of interests are
further modeled by the method used in (Du &
Ukkusuri, 2008b). For completeness, we briefly in-
troduce the main idea by the following points:
1. The VANETs are composed of the vehicles
running on the road during the time inter-
val, T. Vehicular velocity is assumed to be
constant during the short time interval.
2. Based on the velocity of vehicles, we group
the drivers into three categories: aggressive
drivers (A), defensive drivers (D), and slow
drivers (S). α, β, and γ are used to denote their
proportions on the road, α + β + γ = 1.
3. Define pij as the probability that vehicle i
and j, j > i are approaching to each other,
and p(δij ≥ 0) = pij , then, we find there are
three possible cases for node i and j, ( j > i)
approaching each other. The corresponding
probabilities are described below:
• Vehicle i is an aggressive driver and ve-
hicle j is a defensive driver, pij = αβ;
• Vehicle i is a defensive driver and
vehicle j is a slow driver, pij = βγ;
• Vehicle i is an aggresive driver and
vehicle j is a slow driver, pij = αγ.
4. The relative velocity of any node i and j ap-
proaching each other, δij is assumed to be
represented by the relative speed between
adjacent lanes with a white noise:
δij = δL + εij, εij ~ N(0,σ2), (2)
where, σ2 is equal to the variance of the rela-
tive speed between any two vehicles which
are approaching. Note that εij may not be a
normal distribution. The accurate distribu-
tion of εij can be obtained from microscopic
traffic data in a transportation system. This
work is beyond the scope of this study.
5. We numbered the vehicles from entry to exit
at time zero, and then drew all communica-
tion link happened in the time interval on
the topology of VANETs at time zero. This
time expanded graph is referred to as Infor-
mation Flow Networks (IFN). Based on the
IFN, we develop the analytical expressions
to measure the reachable neighbors for in-
dividual vehicles in a given time interval.
To facilitate the further discussion we need
the following notations:
• xtij: The distance between vehicle i and
vehicle j at time t.
• p t ij: The probability of the event that
xtij < r, 0 < t ≤ T.
• sij: A random variable which denotes the
event that xij ≤ r at time zero.
1, x ij0 ≤ r|i < j;
sij = 
0, o.w.
• wij: A random variable which denotes the
event that xij ≤ r after a time period t, if xij > r
at time zero. when
1, x ijt ≤ r,and x ij0 ≥ r|i < j,0 < t ≤ T;
w ij = 
0, o.w.
• zij: A random variable which denotes the
event that a node has chance to exchange
message with its forward nodes during the
whole time interval (See Box 1).
171
 Integrating Traffic Flow Features to Characterize the Interference in Vehicular Ad Hoc Networks

Box 1.

1, node i exchanges messages with node j in the time interval [0, T ], i < j;
zij = 
0, o.w.

Here, the SVD of individual vehicles represents ≈  n 

 n   n ÷
the average number of vehicles in the transmis- E …ƒ z ij ⁄ = p(i ≠ A) ∆E …ƒ sij ⁄ + E …ƒ w ij ⁄ ◊
∆ ◊
sion rang of an individual vehicle during the time  j=i +1    j=i +1   j=i +1  
period. Therefore, it is composed of two compo- ≈  n   n ÷
nents. The first one is the neighbors which exist + p(i ≠ D) ∆E …ƒ sij ⁄ + E …ƒ w ij ⁄ ◊
∆ j=i +1 ◊
at time zero and the other part is the neighbors     j=i +1  
which occur in (0, T]. Correspondingly, the fol- ≈  n   n ÷
+ p(i ≠ S) ∆E …ƒ sij ⁄ + E …ƒ w ij ⁄ ◊
lowing equations hold: ∆ j=i +1 ◊
    j=i +1  
(4)
z ij = sij + w ij
n n n
Note that sij is only related to the traffic den-
ƒ
j= i +1
z ij = ƒ
j= i +1
sij + ƒ
j= i +1
w ij
sity and the vehicle’s transmission range, and
 
n  n   n  the mobility of vehicle has no effect on it. Thus,
E …ƒ zij ⁄ = E …ƒ sij ⁄ + E …ƒ w ij ⁄ (3) we have
 j=i +1   j=i +1   j=i +1 
 n 
Since vehicle i can be aggressive (A), defen- E …ƒ sij ⁄ = λr (5)
sive (D), or slow driver(S), we can express the  j=i +1 
expected number of neighbors of an individual However, the value of wij is closely related to
vehicle as below: the relative movement between vehicles, so the
value of E(wij)can be derived as below:

Figure 4. Traffic state

Tr avel Ti me
1200
1150
1100
Average travel time

1050
1000
950
900
850
10000
50
400
750
1100
1450
1800
2150
2500
2850
3200
3550
3900
5000
6400

Demand

172
 Integrating Traffic Flow Features to Characterize the Interference in Vehicular Ad Hoc Networks

E ƒ

n
j = i +1
wij  = ƒ

n
j = i +1
E ( wi j ) = ƒ
n
j = i +1
(p
∂1 + (1 − pijT ) ∂0 )n= ƒ j =i +1 p (r < xij0 < T δij− )
T
ij
n

E …ƒ zij ⁄ = 3λr
 j =i +1 
n
E ( wi j ) = ƒ
n
( p T
∂1 + (1 − p T
) ∂0 ) = ƒ j =i +1
n
p ( r < x 0
< T δ − 
)
j = i +1 j = i +1 ij

ij ij ij
(6)
n
{
+3 αλ ƒ j =i +1 p (r < xij0 < r + 2T (δ L + εij ))

When vehicle i is driven by an aggressive,
defensive or slow driver in time interval T, the
probability that it is approaching the forward
vehicles is different. Only considering the relative
movement between different driver groups and
referring to result of δij and pij at the beginning ofbetween vehicles. Additionally, it shows that no
this section, E(wij) is calculated below:
• Node i is an aggressive driver.
+β(α + γ )ƒ
n
}
p (r < xij0 < r + T (δ L + εij )) .
j = i +1
(11)
Equation (11) measures the interference of
VANETs with a view to the relative movement
matter under which kinds of traffic flow condi-
tions, the relative movement between adjacent
lanes will increase the interference of VANETs.
In the next step, we calculate the distribution of

0
( ) (
0 vehicles on the road under different traffic flow
E ( wij ) = βp r < xij < r + T (δ L + εij ) + γp r < xij < r + 2T (δ L + εij ) ;
conditions. This will allow us to obtain better
)
( 0
) ( 0
p r < xij < r + T (δ L + εij ) + γp r < xij < r + 2T (δ L + εij ) ; (7) )
insights into the expected interference in view
of macroscopic traffic flow states.
• Node i is a defensive driver.
E xpected Interference Under
E ( wij ) = γp (r < xij0 < r + 2T (δ L + εij )) Different Traffic Conditions
(8)
Our analytical expression in (11) has shown
• Node i is a slow driver. that the interference of VANETs depends on
space headway distributions under different
E(wij) = 0. (9) traffic congestion levels. We therefore specify
the macroscopic traffic feature (space headway
Inserting Equation (5), (7), (8), and (9) to distribution) and integrate it into the study of the
Equation (4), the expected neighbors of node i is VANET interference.
equal to: ( see Box 2). The immediate difficulty is that even though
Therefore, during a short time period, the inter- traffic time headway has been well studied un-
ference of a successful transmission (EI (T)) is der different traffic congestion levels, there are
no well-developed stochastic models for space
headway distribution. Usually, corresponding to

Box 2.

E ƒ z  = αγ ƒ
n n
p (r < xij0 < r + 2T (δ L + εij ))
 
j = i +1 ij j = i +1
(10)
+ β(α + γ )ƒ
n
j = i +1
p (r < xij0 < r + T (δ L + εij )) + λr.

173
 Integrating Traffic Flow Features to Characterize the Interference in Vehicular Ad Hoc Networks

free flow, light congestion and heavy congestion SchÄonhof et al., 2006). Similarly, we use ex-
traffic condition, time headway is classified into ponential distribution to present the distribution
three states: random headway state, intermedi- of the space headway under free flow. The cor-
ate headway state, and constant headway state. responding interference in a short time period
Exponential distribution, Erlang distribution, can be written as shown in Box 3.
and Normal distribution have been proved to be Under the intermediate headway state, the
reasonable stochastic models for the three head- distribution of the time headway is approximated
way states (May, 1990). To apply these well-built by Erlang distribution (May, 1990). Consequently,
time headway models to space headway, we make the space headway distribution can be modeled
more assumptions: (i) the space head- ways be- as Erlang:
tween consecutive vehicles are independent and
b −1 x
identically distributed; (ii) the vehicular speed is 1 x −
f ( x) = e θ (14)
supposed to be uniform (Tsugawa & Kato, 2003) θΓ(b) … ⁄
θ
with mean. Thus, the relationship between space
where,
headway (Hs) and time headway (Ht) is
• x is the space headway.
Hs = Htδ, (12)
H s2
• b is the shape parameter. b =
where δ, is the average speed of vehicles. If δ is a σ 2s
constant, the space headway will have the same 1
• is the scale parameter.
distribution with the time headway. With above θ
assumptions we explore the interference of one
The mean of the space headway is bθ = Hs,
successful transmission under different space
H σ2
headway states. so the estimated scale parameter θ = s = s
Under random headway state (free flow con- 1 b Hs
and traffic density λ = , Furthermore, xij,
dition), vehicles can be thought of as traveling bθ
independently. Except for the minimum headway j = i + 1, …, n obeys Erlang distribution, so let
specification, the time headway is considered as p(xij ≤ r) = Er(r,|j = i|b,) The interference can be
written as shown in Box 4.
random headway. The random time headway is
Under the constant headway state, the distribu-
usually described by exponential distribution
tion of the time headway is described by normal
in analytical models (May, 1990; Saito, 2006;
distribution (May, 1990). We additionally use

Box 3.

 n 
E …ƒ zij ⁄ = 3K c
 j =i +1 
 ≈ ≈ 1÷ 1 ÷÷
 
( ) λ
≈
+3 αγ ƒ j =i +1 ∆E ∆ r + 2T (δ L + εij ) ,| j − i |, ◊ − E ∆r ,| j − i |, ◊◊
n

 λ 
(13)

≈ ≈ 1÷ 1 ÷÷
+β (α + γ )ƒ
n
j = i +1 ( ) ≈
∆E ∆ r + T (δ L + εij ) ,| j − i |, ◊ − E ∆r ,| j − i |, ◊◊ .
  λ  λ 

174
Integrating Traffic Flow Features to Characterize the Interference in Vehicular Ad Hoc Networks

Box 4.

≈ n ÷
E ∆ ƒ zij ◊ = 3K c
 j =i +1 

{
+ αγ ƒ (Er ((r + 2T (δ + ε ),| j − i | b, θ))− Er (r,| j − i | b, θ)) (15)
n
j = i +1 L ij

+β(α + γ )ƒ (Er ((r + T (δ + ε )),| j − i | b, θ)− Er (r,| j − i | b, θ))}.

n
j = i +1 L ij

Box 5.

 n 
E …ƒ zij ⁄ = 3K c
 j =i +1 

{
+3 αγ ƒ
n
(N ((r + 2T (δ + ε )),| j − i | H ,| j − i | σ )− N (r,| j − i | H ,| j − i | σ )) (16)
j = i +1 L ij s
2
s s
2
s

+β(α + γ )ƒ (N ((r + T (δ + ε )),| j − i | H ,| j − i | σ )− N (r,| j − i | H ,| j − i | σ ))}.

n
j = i +1 L ij s
2
s s
2
s

normal distribution to describe the space headway
with mean Hs and variance σs2. Moreover, for xij,
j > i + 1, the distance between nonadjacent ve-
hicles also obeys normal distribution with mean
|j – i|Hs and variance |j – i|2σ2s. Let p(xij < r) =
N(r, |j – i|Hs , |j – i|2σ2s ), then the interference of
a successful transmission can be measured by the
equation shown in Box 5.
In Equations (13), (16), and (15), the composi-
tion of drivers and relative movement between
adjacent lanes roughly present the relative move-
ment between vehicles. Different space headway
distribution characterizes the dynamic traffic flow
state. Hence, we conclude Equations (13), (16), and
(15) together analytically measure the interference
of VANETs on account of both macroscopic and
microscopic traffic flow features. To validate
our analytical results, we perform the following
experiments based on simulation.
Validating the Expected Degree
Under Different Traffic Flow
Conditions
We have examined the expected vehicular degree
under different traffic flow states by simulations,
where Kc = 3, T = 30s, and demand changes from
50vph to 8000vph. Based on our VANET models
and well-build time headway distribution mod-
els, we further developed the stochastic models
to approximate the interference under different
traffic headway states. Now, based on the simula-
tion models we validate our stochastic models in
Equation (13), Equation (15), and Equation (16).
We consider the drivers as aggressive drivers, if
their velocity is fifteen percent higher than the
average velocity of all vehicles. Correspondingly,
a driver is a slow driver if his/her velocity is 15%
lower than the average speed. All other drivers are

175
 Integrating Traffic Flow Features to Characterize the Interference in Vehicular Ad Hoc Networks

Figure 5. The relative difference between adegree and sdegree under free flow (abs(.): absolute value.
adegree: the expected vehicular degree calculated by our analytical expression; sdgree: the expected
vehicular degree measured by simulation)
0. 25

0. 2
sDegree)/sDegree
abs(aDegree-

0. 15
T=30
0. 1

0. 05

0
100 200 800 900 1000
De ma nd

Figure 6. The relative difference between adegree and sdegree under light congestion flow
0. 16
0. 14
0. 12
sDegree)/sDegree
abs(aDegree-

0. 1
0. 08 T=30
0. 06
0. 04
0. 02
0
00

00

00

00

00

00

00

00

00

00
42

44

46

48

50

52

54

56

58

60

demand

Figure 7. The relative difference between adegree and sdegree under heavy congestion flow

0. 16
0. 14
0. 12
sDegree)/sDegree
abs(aDegree-

0. 1
0. 08 T=30
0. 06
0. 04
0. 02
0
6200 6400 6600 6800 7000 7500 8000
demand

176
Integrating Traffic Flow Features to Characterize the Interference in Vehicular Ad Hoc Networks
treated as defensive drivers. The relative move-
ment between individual vehicles is substituted by
the average relative speed between driver groups.
The validation results are reported in Figure 5,
Figure 6, and Figure 7 respectively. When time
headway is under random state (free flow), the dif-
ference between simulation counts and analytical
calculation on TVD is around 15%. While time
headway under intermediate state (light conges-
tion), or constant state (heavy congestion), the
difference is about 10%. Clearly, our stochastic
models perform reasonably good to evaluate the
TVD in the time intervals.
Our simulation experiments also demonstrate
that the accuracy of our models can be improved
with cost of adding more numerical computations
to cluster drivers more accurately. For example, in
our simulations, if we consider drivers as aggres-
sive drivers by its velocity 13% higher than the
average value, the performance of the analytical
interference evaluation might become better.
ConcluSION
There are two interweaved layers within VANETs,
namely, the communication layer and the trans-
portation application layer. Most of the previous
research about the communication layer treated
VANETs as a special class of Mobile Ad Hoc
Networks (MANETs) and ignored the traffic flow
features. To overcome this deficiency, our research
incorporated both macroscopic and microscopic
traffic flow characteristics into the study of several
fundamental issues in the communication layer
of VANETs such as the connectivity, the reach-
ability, and the capacity of VANETs.
Along the line of our completed work, the study
presented here investigates another fundamental
issue in VANETs: interference. Due to the high
vehicular mobility and the dynamic traffic flow
features, the interference in VANETs presents
distinguishing features. To characterize the inter-
ference in VANETs, we first set up the simulation
models to explore the interference under different
traffic flow conditions. The results demonstrate:
When each vehicle adjusts its transmission range
to reach the same number of neighbors on aver-
age according to the traffic density, the interfer-
ence of individual transmission in a short time
interval, such as T = 30s, is much higher than the
corresponding interference under static vehicle
distribution; Relative movement between vehicles
and traffic congestion level impact the interfer-
ence of VANETs significantly; Under free and
light congestion traffic condition, the interference
increases with demand. However, under heavy
congestion, the interference will decrease the
traffic as demand increases.
Furthermore, under our assumptions, we
develop the stochastic models to estimate the
expected interference of VANETs. Our closed
form expressions approximate the interference in
VANETs taking accounting of both the macro-
scopic and the microscopic traffic flow characteris-
tics. The validation results based on the simulation
data demonstrate that our analytical expressions
perform well under various traffic conditions. In
summary, our research efforts try to investigate
the information propagation in VANETs bridg-
ing the features in both the communication layer
and the transportation layer and therefore help to
build more efficient systems.
REFERENCES
Artimy, M., Robertson, W., & Phillips, W. (2005b).
Assignment of dynamic transmission range based
on estimation of vehicle density. In VANET’05,
Cologne, Germany, Sep., 2.
Artimy, M., Robertson, W., & Phillips, W. (2006).
Minimum transmission range in Vehicular Ad
Hoc Networks over uninterrupted highways. In
The 9th International IEEE Conference on Intel-
ligent Transportation Systems, Toronto, Canada,
Sep., 17-20.
177
 Integrating Traffic Flow Features to Characterize the Interference in Vehicular Ad Hoc Networks
Ben-Akiva, M., Davol, A., Toledo, T., Koutso-
poulos, H. N., Burghout, W., Andréasson, I.,
Johansson, T., & Lundin, C. (2002). Calibration
and evaluation of mitsimlab in stockholm. In
Procedings of 81st Transportation Research Board
Meeting, Washington, DC, USA, Jan. 13-17.
Blough, D. M., Leoncini, M., Resta, G., & Santi,
P. (2006). The k-neighbors approach to interfer-
ence bounded and symmetric topology control in
ad hoc networks. IEEE Transactions on Mobile
Computing, 5(9), 1267-1282.
Blum, J., Eskandarian, A., & Hoffman, L. (2004).
Challenges of intervehicle ad hoc networks. IEEE
Transactions on Intelligent Transprotation Sys-
tems, 5(4), 347-351.
Burkhart, M., Rickenbach, P. von, Wattenhofer,
R., & Zollinger, A. (2004). Does topology control
reduce interference. In MobiHoc’04, Roppongi,
Japan, May 24-26.
Dia, H., & Purchase, H. (1999). Modeling the
impacts of advanced traveler information systems
using intelligent agents. Road and Transport
Research, 8.
Du, L., & Ukkusuri, S. (2008a). Optimization
models to characterize the broadcast capacity
of vehicular ad hoc networks. Transportation
Resaerch Part C.
Du, L., & Ukkusuri, S. (2008b). The relative
mobility of vehicles improves the performance
of information flow in vehicle ad hoc networks.
Networks and Spatial Economics.
Du, L., Ukkusuri, S., & Kalyanaraman, S. (2007).
Geometric connectivity of vehicular ad hoc net-
works: Analytical characterization. In The Fourth
ACM Workshop on Vehicular Ad Hoc Networks
(VANET), Montreal, Canada, Sep. 10.
Fussen, M., Wattenhofer, R., & Zollinger, A.
(2004). Sensor Networks: interference reduction
and possible application. Diploma Thesis, ETH
Zurich, Dept. of Computer Science.
178
Gao, S., & Chabini, I. (2006). Optimal routing
policy problems in stochastic time-dependent
networks. Transportation Research Part B, 40.
Heide, F. M. A. D., Schindelhauer, C., Volbert,
K., & Gruenewald, M. (2002). Energy, conges-
tion and dilation in radio networks. Proceedings
of the 14th Annual ACM Symposium on Parallel
Algorithms and Architectures (SPAA), Winnipeg,
Manitoba, Canada, Aug. 11-13.
Hou, T., & Li, V. (2006). Transmission range
control in multihop packet radio network. IEEE
transactions on wireless communications, 5(1),
38-44.
Jaillet, P., & Wagner, M. (2006). Online rout-
ing problems: Value of advanced information
as improved competitive ratios. Transportation
Science, 40(2).
Jain, K., Padhye, J., Padmanabhan, V., & Qiu, L.
(2003). Impact of interference on multi-hop wire-
less network performance. In MobiCom 2003, San
Diego, California, USA, Sep. 14-19.
Jin, W., & Recker, W. (2006). Instantaneous in-
formation propagation in a traffic stream through
inter-vehicle communication. Transportation
Research Part B, 40(3), 230-250.
Li, N., Hou, C., & Sha, L. (2003). Design and
analysis of an mst-based topology control algo-
rithm. In Proc. of the 22nd Annual Joint Confer-
ence of the IEEE Computer and Communications
Societies (INFOCOM), San Franciso, CA, USA,
Mar. 30-Apr. 3.
Lo, H., & Szeto, W. (2004). Modeling advanced
traveler information services: static versus dy-
namic paradigms. Transportation Research Part
B: Methodological, 38(6), 495-515.
May, A. D. (1990). Traffic flow fundamentals.
Englewood Cliffs, NJ: Prentice-Hall, Inc., A
Division of Simon and Schuster.
Integrating Traffic Flow Features to Characterize the Interference in Vehicular Ad Hoc Networks
Moscibroda, T., & Wattenhofer, R. (2005).
Minimizing interference in ad hoc and sensor
networks. In DIALMPOMC’05, Cologne, Ger-
many, Sep., 2.
Mouskos, K. C., Greenfeld, J., & Pignataro, L. J.
(1996). Toward a multi-modal advanced traveler
information system. NJIT Research, 4.
Nejad, K., & Li, X. (2005). Low- interference
topology control for wireless ad-hoc networks.
Ad Hoc Sensor Wireless Networks, 1, 41-64.
Ni, J., & Chandler, S. (1994). Connectivity prop-
erties of a random radio network. IEE Proc.
Commun., 141(4), 289-296.
Rajaraman, R. (2002). Topology control and rout-
ing in ad hoc network: A survey. ACM SIGACT
News, 33(2), 60-73.
Ramanathan, R., & Rosales-Hain, R. (2000).
Topology control of multihop wireless networks
using transmit power adjustment. In Proc. of
the 19th Annual Joint Conference of the IEEE
Computer and Communications Societies (IN-
FOCOM), Tel-Aviv, Israel, Mar. 26-30.
Rickenbach, P., Schmid, S., Wattenhofer, R.,
& Zollinger, A. (2005). A robust interference
model for wireless ad hoc networks. In Proc. 5th
IEEE International Workshop on Algorithms for
Wireless, Mobile, Ad-Hoc and Sensor Networks
(WMAN), Denver, Colorado, USA, April 4-8.
Saito, H. (2006). Performance analysis of com-
bined vehicular communication. IEICE Transac-
tions on Communications, E89B(5),. 1486-1494.
SchÄonhof, M., Kesting, A., Treiber, M., &
Helbing, D. (2006). Coupled vehicle and infor-
mation flows: Message transport on a dynamic
vehicle network. Physica A-Statistical Mechanics
and ITS Applications, 363(1), 73-81.
Shavitt, Y., & Sha, A. (2005). Optimal routing in
gossip networks. IEEE Transactions on Vehicular
Technology, 54(4).
Srinivasan, K. K., & Mahmassani, H. S. (2003).
Analyzing heterogeneity and unobserved struc-
tural effects in route-switching behavior under
ATIS: a dynamic kernel logit formulation.
Transportation Research Part B: Methodologi-
cal, 37(9), 793-814.
Tsugawa, S., & Kato, S. (2003). Evaluation of
incident information transmission on highways
over inter-vehicle communications. In Intelligent
Vehicles Symposium, 2003. Proceedings. IEEE,
Columbus, Ohio, USA, Jun. 9-11.
Ukkusuri, S., & Du, L. (2008). Geometric con-
nectivity of vehicular ad hoc networks: Analytical
characterization. Transportation Research Part
C, 16(5), 615-634.
Waller, S., & Ziliaskopoulos, A. (2002). On the
online shortest path problem with limited arc cost
dependencies. Networks, 40(4).
Wang, X. (2007). Modeling the process of infor-
mation relay through inter-vehicle communica-
tion. Transportation Research Part B, 41(6),
684-700.
Xue, F., & Kumar, P. (2004). The number of
neighbors needed for connectivity of wireless
network. Wireless Networks, 10, 169-181.
Endnot
1
Throughput of an ad hoc network is usually
defined as the maximum bits per second that
is achievable by all source-destination pairs
of nodes.
2
The vehicles which are in the transmission
range of an individual vehicle.
3
To check TVD by simulation, the reader is
referred to the algorithm described in Du et
al. (2008b).
179
180

Chapter X
Proactive Traffic Merging
Strategies for Sensor-Enabled
Cars
Ziyuan Wang
University of Melbourne, Australia

Lars Kulik
University of Melbourne, Australia

Kotagiri Ramamohanarao
University of Melbourne, Australia

ABSTtract

Congestion is a major challenge in today’s road traffic. The primary cause is bottlenecks such as ramps
leading onto highways, or lane blockage due to obstacles. In these situations, the road capacity reduces
because several traffic streams merge to fewer streams. Another important factor is the non-coordinated
driving behavior resulting from the lack of information or the intention to minimize the travel time of a
single car. This chapter surveys traffic control strategies for optimizing traffic flow on highways, with
a focus on more adaptive and flexible strategies facilitated by current advancements in sensor-enabled
cars and vehicular ad hoc networks (VANETs). The authors investigate proactive merging strategies
assuming that sensor-enabled cars can detect the distance to neighboring cars and communicate their
velocity and acceleration among each other. Proactive merging strategies can significantly improve
traffic flow by increasing it up to 100% and reduce the overall travel delay by 30%.

Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.
Proactive Traffic Merging Strategies for Sensor-Enabled Cars
Introduction
Road traffic congestion is a major challenge
nowadays. The integration of sensing, communi-
cating and local computing within cars facilitates
new opportunities to optimize road capacity, and
thereby reducing traffic congestion. Currently, a
variety of automotive sensors are available to col-
lect data related to a vehicle and its surroundings.
Communication techniques, such as Dedicated
Short Range Communications (DSRC) support
the information exchange between cars. Recent
research primarily focuses on: safety by reducing
traffic accidents based on, for example, collision
warnings (ElBatt et al., 2006); traffic flow con-
trol to reduce traffic congestion and to increase
road efficiency by adapting traffic lights and
speed limits (Papageorgiou & Kotsialos, 2002);
and automation to reduce a driver’s burden, e.g.,
using Adaptive Cruise Controls (ACC) (Kesting
et al., 2008).
One important cause of congestion is bottle-
necks at merging areas where a ramp leads onto
the highway. There is a considerable amount of
research to address merging strategies, in particu-
lar approaches from queuing theory or statistics.
This chapter explores the benefits of applying
simple traffic control rules at the merging sections
assuming all cars are sensor-enabled.
The goal of merging strategies is to mitigate
traffic congestion by optimizing the utilization
of roads, ensuring road safety. We believe that
a promising approach is achieved by sensor-
enabled cars using well-informed decisions and
cooperating driving. Traditional traffic control
approaches such as traffic signal timing, ramp
metering, and real-time information can benefit
from the sensor-enabled cars and Vehicular Ad
hoc NETworks (VANETs).
The rest of the chapter is structured as fol-
lows: we first review foundations of strategies to
control traffic flow, factors that affect traffic flow,
desirable and undesirable characteristics of traffic
flow. Next we overview traffic control strategies
and examine merging strategies. Then we present
the criteria to evaluate merging strategies. Finally,
we outline future trends and challenges for traffic
control research and conclude the chapter.
Background
In this section, we first review current research
on traffic models. These models provide the
foundation to explore the effects of implementing
new traffic control strategies. We argue that the
mobility models commonly used in the study of
communication ranges and information dissemi-
nations in VANETs are not sufficient for investi-
gating traffic control strategies. Then, we review
automotive sensors that can be implemented in
the sensor-enabled cars. Those sensors collect
traffic information as the input for the traffic
model applied in traffic control strategies. Finally,
we highlight some basic vehicular traffic theory
concepts, with a focus on the traffic congestion
on highways to provide a better understanding of
traffic flow and design the flow control strategy.
Vehicular Traffic Models
Traffic models and simulations provide an under-
standing of traffic phenomena to manage traffic
in such a way that traffic flow is optimized and
congestion is alleviated. From a computational
perspective, various tools and techniques such
as car-following theories and cellular automata
have been applied to traffic modeling.
By far the most widely used mobility model is
the Random WayPoint model (RWP). This model
was first used by Johnson and Maltz (1996), and
it has become the de facto standard in mobile
computing research, including VANETs.
RWP has several limitations as a model for
VANETs. First, it describes the movement pattern
of independent nodes, however, the movement of
vehicles is not independent as vehicles interact
with other vehicles in their vicinity. Second,
181
 Proactive Traffic Merging Strategies for Sensor-Enabled Cars
RWP assumes that node movement occurs in an
open field, whereas the movement of vehicles is
constrained by the roads. Third, the node chooses
a random velocity from the uniformly distributed
velocity range in RWP. In contrast, a vehicle’s
velocity is constrained by many factors, such as
the road conditions, traffic lights, the number
and type of intersections, and the existence of
obstacles. Therefore, mobility models based on
real street maps have been proposed for VANETs.
In Saha and Johnson’s model (SJM) (Saha &
Johnson, 2004), street map data is converted to
(x, y) coordinates. The road segments are repre-
sented as the edges in an undirected graph and the
intersections are represented by the nodes in an
undirected graph. The STRAW approach (Choff-
nes & Bustamante, 2005) includes intersection
management, such as the coordination of traffic
lights and stop signs. Saha and Johnson point out
that RWP is a good approximation of car move-
ment in many cases. While it is often sufficient
for evaluating the communication performance
of a VANET, it is not a good approximation to
evaluate a flow control strategy for a real road.
Although Saha and Johnson (2004) applied real
street maps for simulating vehicle movement,
some assumptions restrict its application to flow
control strategies, which rely on realistic vehicular
traffic characteristics. In this regard, models with
varying velocity are highly needed. As a result,
vehicular traffic models have attracted extensive
research.
Traffic models can be classified according to
two criteria. First, according to the use of space,
time and state variables, traffic models can be
categorized as continuous or discrete. Earlier
theoretical approaches, in particular, fluid dy-
namic theory, kinetic theory, and car-following
theories are based on continuous models. Cellu-
lar Automata (CA) models are discrete. Second,
depending to the level of detail of the simulation,
there are macroscopic and microscopic models.
Macroscopic models are based on equations to
182
describe the collective characteristics of traffic,
such as average speed and traffic density. Re-
searchers often build such models by adapting
physical models similar to the behavior of fluid
and gases. Recognizing that unlike molecules,
cars move under the control of people, drivers’
behaviours have been taken into account. Macro-
scopic models are sufficient to predict the general
patterns of cumulative number of vehicles. In
general, they are suitable to describe large-scale
traffic networks such as an entire traffic of a city.
However, they fail to represent the interactions
among vehicles. These models cannot describe
variations in velocities, as velocities vary in
overtaking situations or within platoons. Thus,
macroscopic models are not adequate to evaluate
the performance of local merging algorithms.
In contrast, microscopic models overcome
these limitations by describing the individual
behavior of vehicles. Microscopic approaches
include car-following models (Bando et al., 1995)
and CA models (Chowdhury et al., 2000). We
detail microscopic models from the longitudinal
and the lane-changing aspect in the following
two sections.
Longitudinal Models
Longitudinal models describe the acceleration and
deceleration of vehicles. In car-following models,
the assumption is that the behavior of a driver
depends only on its predecessor. Several models
have been developed such as the optimal velocity
model (Bando et al., 1995), the intelligent driver
model (Treiber et al., 2000), and the human driver
model (Treiber et al., 2006), to name a few.
In the Intelligent Driver Model (IDM), a vehicle
tends to approach the maximum permitted velocity
and maintains at least the safety distance to the
front vehicle. The safety distance depends on the
following car’s velocity and velocity difference
to its front car. The acceleration and deceleration
depend on its own velocity, safety distance and
Proactive Traffic Merging Strategies for Sensor-Enabled Cars
actual distance to the front car. It has the follow-
ing advantages (Treiber & Helbing, 2002): the
model is a realistic description of both the indi-
vidual driving behavior and collective patterns of
the traffic flow, such as stop-and-go waves; and
many aspects of traffic control strategies can be
simulated by representing different driving styles,
which are easy to implement.
Nevertheless, the computational cost of such
detailed models is very high and it is hardly pos-
sible to apply them to very large networks. Since
cellular automata are by design ideal for large-
scale computer simulations, they can be efficiently
used to describe traffic flow. In the CA models
(Chowdhury et al., 2000), the position, speed, ac-
celeration as well as time are treated as discrete
variables. In this approach, a lane is represented
by one-dimensional cells. Each cell can be either
empty or occupied by at most one vehicle at a given
instant of time. At each discrete time step, the state
of the system is updated following a well defined
prescription. The computational efficiency of the
discrete CA models is the main advantage of this
approach over the car-following approach. Nagel
and Schreckenberg (1992) introduce a very simple
CA model that provides a microscopic description
of the vehicular motion using a set of update rules.
The rules include 1) acceleration, 2) deceleration,
3) randomization, and 4) vehicle movement.
The first two rules describe an optimal driving
strategy; the driver accelerates if the vehicle has
not reached the maximum speed vmax and brakes
to avoid accidents. Such a cellular automaton is
deterministic and the stationary state depends
only on the initial conditions. However, drivers
do not always exhibit this optimal behavior. They
vary their driving behavior without any obvious
reasons. Therefore, it is necessary to introduce
the parameter “slowdown probability” p, which
is essential for a realistic description of traffic
flow. It mimics the complex interactions between
the vehicles and is also responsible for the spon-
taneous formation of jams. Although it is one of
the simplest traffic flow models, it is nonetheless
capable of reproducing important properties of real
traffic flow, like the density-flow relation and the
spatio-temporal evolution of traffic jams.
Lane-Changing Models
Besides longitudinal models, realistic models of
lane changing behaviors are critical for investi-
gating traffic patterns at merging sections. Based
on car-following models, several lane-changing
models are proposed. Ahmed et al. (1996) identify
three types of lane changing operations, namely
discretionary, mandatory and forced lane chang-
ing. A discretionary lane change is desired when
a driver wants to improve his driving condition by
changing a lane (for example, to gain speed). A
mandatory lane change occurs when a driver has
to change a lane in order to maintain his path to
the destination (for example, changing lane from
a closed lane or changing a lane to take an exit).
The execution of discretionary and mandatory
lane changing is based on the critical gap ac-
ceptance behavior, i.e., drivers seek a minimum
acceptable gap, called critical gap. Traditional
gap acceptance models fail to capture a driver’s
lane changing behavior in a congested network,
because there is no space for acceptable gaps.
In such situations, drivers force other drivers to
create an acceptable gap. Merging operations in a
heavily congested freeway is one of the examples
of such behavior.
Kesting et al. (2007) propose to minimize
the overall braking induced by a lane change
to derive lane-changing rules for discretionary
and mandatory lane changes for a wide class of
car-following models. Both the utility of a given
lane and the risk associated with lane changes
are determined in terms of longitudinal ac-
celerations calculated with microscopic traffic
models. The minimum safety distance prevents
critical lane changes and collisions, the incentive
criterion takes into account the advantages and
disadvantages of other drivers associated with
a lane change via the “politeness factor”. This
183
 Proactive Traffic Merging Strategies for Sensor-Enabled Cars
Table 1. Traffic models overview
Movement
Unconstrained
Scale
Local
(Microscopic)
Global RWP (Johnson & Maltz,
(Macroscopic) 1996)
parameter allows one to vary the motivation for
lane changing from a purely selfish to a more
cooperative driving behavior.
Based on the Nagel-Schreckenberg cellular
automata model, Pedersen and Ruhoff (2002)
describe a model to include ramps. The idea is to
place “shadow cars” on a highway next to cars on
ramps, which enables drivers to take ramp cars
into account.
The mobility model is a bridge between the
communication and the traffic control algorithms.
It should reflect vehicle movement as well as the
impact of communication on the traffic flow.
Table 1 summarizes the discussed models.
We categorize them into four classes. Other than
classifying as microscopic and macroscopic we
further identify unconstrained and road map
movement. All the models come from physical
research belong to road map movement group.
We distinguish the restrictions on movement in
order to keep accordance with the mobility models
applied in VANET research.
Snsors
Traffic dynamics measured by sensors are input
variables to the car-following models underly-
ing traffic control strategies. For example, IDM
requires the following input variables: velocity,
distance to the front vehicle, and velocity differ-
ence. We survey major automotive sensors and
techniques. We have found that the assumed
sensor-enabled cars are realistic for current tech-
184
Road Map
CA (Nagel & Schrecken-
 berg, 1992), IDM (Treiber
et al., 2000)
SJM (Saha & Johnson,
2004), STRAW (Choffnes
& Bustamante, 2005)
nology. Among different types of sensors, such
as position, velocity, or acceleration sensors, we
identify what is the most critical type of sensor.
We also investigate how different levels of sensors’
accuracy influence their applications.
Automotive sensors have been developed
since the early 1970s. Recent advances in Micro-
Electro-Mechanical Systems (MEMS) technology
and wireless communications have enabled the
development of low-cost, multifunctional sensor
nodes that are small in size and with short-range
wireless communication capabilities (Akyildiz et
al., 2002). These tiny sensor nodes, which consist
of sensing, data processing, and communicating
components, can be installed to modern cars to
facilitate transport applications. This new type of
sensor-enabled cars is able to sense information
about its own position and local traffic conditions,
process the information, and communicate the in-
formation to other vehicles in its neighborhood.
There are two kinds of automotive sensing:
out-vehicle environment sensing and vehicle-
state sensing, summarized in (Li et al., 2005).
Out-vehicle environment sensing collects in-
formation about the driving environment. For
example, this includes sensors for extracting lane
boundaries, especially when not clearly marked
or in bad weather conditions; detecting nearby
vehicles and estimating their position, speed, and
acceleration; and detecting the unexpected traffic
participants (such as pedestrians) and obstacles.
Vehicle-state sensing focuses on measuring a ve-
hicles movement and monitoring its actuators. For
Proactive Traffic Merging Strategies for Sensor-Enabled Cars
example, researchers have studied how to detect
a vehicle’s position, velocity, and acceleration; an
engine’s pressure and temperature; and a tire’s
pressure, temperature, and friction coefficients.
The accuracy of the information obtained by
sensors impacts parameter settings in the traffic
model. For example, if the position or distance
information is provided with an accuracy of 20
m, the safety distance calculated by the merg-
ing algorithm would be of little value. Besides
distance and velocity, there are several factors
that impact the minimum safety distance, such
as a tire’s pressure, temperature, and the friction
coefficient for braking. Currently, most work on
traffic control neglect these factors and focus
on position and velocity. To envision a complete
model that integrates these factors will lead to
significant improvement in road safety and ef-
ficiency of control strategies.
Position
Most VANETs applications consider the avail-
ability of real-time updated position information.
They differ, however, on the position accuracy
required in order to be able to function properly.
For instance, some applications can work with
position information with errors in the range of
10–20 m, while other applications, especially
critical safety applications, require more accurate
and reliable positioning systems with sub-meter
precision.
In cooperative driving applications, vehicles
in a VANET exchange messages among them to
drive and share the available space on the road
cooperatively. In these applications, the vehicles
can assume partial control over driving. In order
to deal with safety, position information reliability
and accuracy are crucial. Accurate positioning
ensures localization with a meter or sub-meter
precision in order to estimate accurately the
distances between vehicles, while a reliable lo-
calization will ensure that updated information
will always be available.
A number of localization techniques have been
proposed for computing the position of mobile
nodes. Most localization techniques can be ap-
plied easily to VANETs. GPS receivers have a
position error in the range of 1–10 m (Hightower
& Barriello, 2001). Unfortunately, GPS receivers
are not the best solution in some cases since they
cannot work in the parking or in dense urban areas
where there is no direct visibility to satellites. As
a result, GPS information is likely to be combined
with other localization techniques such as dead
reckoning, cellular localization, and image/video
localization, to cite a few.
Velocity and Acceleration
Conventional methods measure the speed of a
vehicle by wheel speed sensors. The drawbacks
of wheel speed sensors are inaccuracy and low
resolution. Radar speed measurement systems are
commonly used due to their reasonable cost and
acceptable accuracy. A GPS based velocity sensor
(Bevly, 2004) has been recently developed that
can be used instead of radar sensors and wheel
speed sensors.
The acceleration and deceleration is measured
by accelerometers and gyroscopes. Currently, ac-
celeration sensors are used in applications such
as safety systems, vehicle stability systems and
electronic suspension. Recent trends of automo-
tive acceleration sensors are to employ MEMS
technology. Their advantages over traditional
sensors are low cost, robustness, self diagnostics,
and multiplex network connectivity (Fleming,
2001).
The IDM traffic model relies on the availabil-
ity of the following sensed information: velocity,
acceleration, forward distance and backward
distance. To increase the reliability of sensed
data, communication is used to check the sensed
data, thus each car also has a transceiver to com-
municate with cars in its vicinity. We use sensor
fusion (Nakamura, Loureiro, & Frery, 2007) to
increase the accuracy of information. In Table 2,
185
 Proactive Traffic Merging Strategies for Sensor-Enabled Cars
Table 2. Automotive sensors
Sensor
Position Laser Range Finder (Bonanno et al., 2003)
GPS (Hightower & Barriello, 2001)
Velocity Speckle Velocimeter (Bonanno et al., 2003)
Velocimeter (Vlacic et al., 2001)
Dickey-John radar (DICKEY-John, n.d.)
Acceleration PASCO CI-6558 (CI-6558, n.d.)
we summarize three types of automotive sensors,
and report their range and accuracy.
Vehicular Traffic Theory
Road efficiency is measured by the traffic flow.
The number of cars or velocity is not sufficient to
describe efficiency separately. The density-flow
relation forms the basis of the so-called funda-
mental diagram. The fundamental diagram is of
major importance for the design of road systems
and traffic management systems. It can be used
to predict the capability of a road, or its behaviour
when applying ramp metering or speed limits. The
origin of the generic shape of the fundamental
diagrams (see Figure 1) can be understood as
follows. When the density D is sufficiently small,
the average speed V is almost independent of D
because the vehicles are too far apart to interact.
However, for larger values of D, V decreases more
rapidly with increasing D because the forward
movement of the vehicles is strongly hindered
by others on account of the reduction in the av-
Figure 1. Fundamental diagram of traffic flow
186
Range Accuracy
50 m 5 cm
24 satellites worldwide 1–10 m
 0.1%
0.4–400 km/h 1%
0.53–96.6 km/h 1–3%
-5 to +5 g 0.01 g
erage separation between them. Because of this
variation of V with D, the flow Q = DV exhibits a
maximum at D = Dc. Then, an increased D results
into an actual decrease of the flow, and finally in
a completely congested phase where no car can
move and the flow is zero.
Classical theories based on the fundamen-
tal diagram of traffic flow consider only two
phases--free flow and congestion. In contrast,
Kerner (2004) proposes three-phase traffic theory
considering freely flowing cars, full-scale traffic
jam, and synchronized traffic flow. If a transition
from free flow to congested flow takes place it
sometimes yields a sudden drop in speed but the
flow remains nearly constant. The new phase in
the jammed region is called synchronized flow.
However, its nature as well as its reasons are still
under discussion.
Kesting et al. (2008) identify five traffic situa-
tions: free traffic, congested traffic, upstream jam
front, downstream jam front, and bottleneck sec-
tions. They associate those traffic situations with
a specific set of IDM parameters. Therefore, their
ACC strategy adapts the driving characteristics
(that is, the parameters of the ACC controller) to
the traffic conditions.
A better understanding of the cause of con-
gestion assists in developing strategies to reduce
congestion. A 2005 US study (Federal Highway
Administration, 2005) states that there are seven
major causes of congestion. They are bottlenecks
(40%), traffic incidents (25%), bad weather (15%),
work zones (10%), poor signal timing (5%), and
special events/other (5%). Chen et al. (2001) believe
Proactive Traffic Merging Strategies for Sensor-Enabled Cars
that congestion does not occur because demand
exceeds capacity rather because of the inefficient
operation of highways.
Flow Control Strat
In this section, we discuss different techniques
to improve traffic flow control, in particular, we
focus on the context of freeways. The main aim
of the flow control is to improve the efficiency of
road usage. Optimization of freeways is based on a
homogenization of vehicle traffic (Helbing, 2001).
Stop-and-go traffic, jams, and congested traffic
are associated with reduced efficiency due to a
capacity drop, often triggered by perturbations
in the traffic flow. For example, if a car pushes
into a small gap to change lane, the impact may
appear small locally but is globally significant as
the following cars may need to slow down sharply.
This is known as “slinky-type effect” (Bose &
Ioannou, 2003), which can lead to traffic conges-
tions and car accidents. Therefore, perturbations
should be suppressed by control strategies such
as variable speed limits, dynamic route guidance,
driver assistance systems, and ramp metering.
Centralized and Decentralized
Systems
In theory, traffic control systems can be catego-
rized into centralized and decentralized systems.
In practice, there are no pure centralized systems.
The present research mainly adopts a hybrid sys-
tem architecture, i.e., a combination of a central-
ized and decentralized approach (Varaiya, 1993;
Anda et al., 2005). Varaiya (1993) discusses pure
centralized and decentralized designs. A central-
ized approach has a tremendous computation
and communication cost. On the other hand, a
decentralized approach requires local computa-
tion, which is expensive. In his approach, cars are
grouped into tightly spaced train-like platoons and
the first car of the platoon is controlled centrally
whereas within platoons cars are decentralized.
Estrin et al. (1999) propose that in sensor net-
works localized algorithms, where sensors only
interact with other sensors in a restricted vicinity,
collectively achieve a desired global objective.
Localized algorithms have three advantages. First,
the communication scales well as the number of
vehicles grows. Second, the algorithms are robust
to network partitions and node failures. Third, un-
like the centralized systems, localized algorithms
provide much greater local details.
Lygeros et al. (1996) introduce a methodol-
ogy for designing hybrid controllers for large
scale, multiagent systems. The system is initially
designed fully decentralized where each vehicle
has only access to local information obtained
by sensors. If the level of performance is unsat-
isfactory, some centralization is introduced by
communication of information between vehicles.
Moreover, they suggest what the most important
pieces of information are. The effect of this more
global knowledge is to reduce the disturbance
generated by other vehicles. The optimum strat-
egy for each vehicle does not coincide with the
global optimum. Ideally, to achieve the global
optimum a centralized control scheme computes
the global optimum and commands the vehicles
accordingly. A centralized controller may be
undesirable, however, as the design process will
probably be very complicated, it is likely to be
computationally intensive, may be less reliable and
the information that needs to be exchanged may be
too large for the available communication capabili-
ties. In their opinion, a completely decentralized
solution is too inefficient and a completely cen-
tralized solution is prohibitively complex and/or
expensive. Therefore, a compromise will feature
semi-autonomous vehicle operation. In (Lygeros
et al., 1998), a method is proposed to deal with
multiple control requirements. To achieve the goal
of improving the throughput of the highway system
while maintaining the safety, their method is to
prioritize the multiple requirements, i.e., safety
is more important than throughput.
187
 Proactive Traffic Merging Strategies for Sensor-Enabled Cars
Researchers have agreed that the traffic control
problem on a global network level is practically
unsolvable by traditional optimization techniques
(see, e.g., Lammer et al., 2006; Lammer et al.,
2006) because of the exponential complexity
of the underlying algorithms. Hence, a number
of decentralized optimal strategies whose ac-
tions are coordinated heuristically by a superior
control layer have been proposed (Gartner et al.,
2001). Lammer et al. (2006) present a self-orga-
nizing, decentralized control method for global
coordination of traffic signal control. They map
the problem to phase-oscillator models. By syn-
chronizing these oscillators, the desired global
coordination is achieved. The concept applies
to networks where time sharing mechanisms
between conflicting flows in nodes are required
and where a coordination of these local switches
on a network level can improve the performance.
Gershenson (2004) proposes self-organizing
methods using simple rules to coordinate traffic
lights to improve traffic flow.
Automation Control
Automated Highway Systems (AHS) have been
proposed based on automated vehicles. Fully
automated vehicle control (Varaiya, 1993) is able
to increase the capacity dramatically. Varaiya
proposes a four-layer control architecture to
handle the control problems. Starting at the top,
the layers are called the network, link, planning,
and regulation layers. The network layer assigns
a route to each vehicle entering the system. The
link layer assigns a section to each vehicle and a
target speed for each section in order to increase
traffic flow. The planning layer coordinates the
movement of neighboring vehicles. The regula-
tion layer executes feedback control laws to finish
certain tasks.
Different traffic automation strategies are clas-
sified according to the level of communication:
188
• Autonomous: Automation is concerned
within individual vehicles and no commu-
nication between vehicles;
• Cooperative: Vehicles cooperate by com-
municating selected information; and
• Platooning: Vehicles form tightly spaced
groups by communicating detailed informa-
tion at a high frequency.
Merging control algorithms for automated
vehicles focus on calculating the required speed
and time to arrive at the merging point (Kato &
Tsugawa, 2001; Lu et al., 2004), as well as the
minimum safety distance (Kanaris et al., 2001).
Hall and Tsao (1997) examine the impact of
merging strategies on the capacity of AHS. They
present three merging strategies integrated with
ramp metering for cooperative and/or platooning
vehicles:
• Release-to-gap: When the ramp vehicle
reaches the merge point, the intended gap
also reaches the merge point so that the ramp
vehicle and the gap are properly aligned at
the merge point;
• Preplatooning:Ramp vehicles into platoons
before merging; and
• Release-to-tag: When the ramp vehicle or
platoon reaches the merge point, the tail of
the mainline platoon reaches the merge point
so that the ramp vehicle or platoon tag along
the mainline platoon.
Their simulation results show that the release-
to-tag strategy is the most efficient merging
strategy, with 20% reduction from the theoretical
maximum capacity. Other strategies lead to as
much as 70% reduction. They point out that the
main cause of the capacity reduction is the first-
in last-out phenomenon, where an available gap
for merging may be taken by a following vehicle
rather than the vehicle that arrived first. However,
their merging strategies and simulations assume
Proactive Traffic Merging Strategies for Sensor-Enabled Cars
that automated vehicles must stop at a check-in
facility and then proceed to the merging area.
In contrast, Lu et al. (2004) show that ramp
vehicles merging into the middle of a platoon
provides higher efficiency in high-density traf-
fic conditions. They propose merging control
algorithm with the concept of virtual platooning,
which forms a platoon of the ramp vehicle and
main road vehicles before the ramp vehicle arrives
at the merging point. This idea emphasizes four
factors that must be considered simultaneously:
vehicle relative positions, timing, speed and ac-
celeration.
At present no approach has proven itself supe-
rior to all its competitors in every respect. Since
there are many different criteria to define perfor-
mance, and so many parameters that determine
performance, one approach may be the winner
in several respects but inferior in other respects.
Although the concept of AHS improves safety,
road utility, and driver comfort, the requirement of
major changes to the existing highway infrastruc-
ture such as dedicated lanes, and a high percentage
of automated vehicles, makes AHS unlikely to
come to reality in the foreseeable future.
Driver Assistance Systems
The partial automation ACC systems are already
available on the market. ACC systems only control
longitudinal driving with the aim of improving
safety and driver comfort. Merging, overtaking,
or lane-changing still need the intervention of
drivers.
In the literature, there is research on how the
ACC systems impact traffic flow, especially in
relation to congestion. There is an agreement that
ACC systems can achieve a smoother traffic flow,
which leads to an increase of the road capacity
and reduction in congestion (Davis, 2004; Kesting
et al., 2008; Liang & Peng, 2000). Kesting et al.
(2008) show that traffic congestion is eliminated,
if there are 25% ACC vehicles in the simulation
based on the IDM model. Similarly, Davis (2004)
reports that with 20% ACC vehicles no jams are
formed in single-lane simulations. Furthermore,
Davis identifies a potential problem with merging.
ACC vehicles maintain desired headways and
do not “give way” when a vehicle enters from a
ramp. The ability to merge may be hindered by
the lack of minimum safety gaps. In another work,
Davis (2007) focuses on the merging problem
and proposes cooperative merging strategy to
increase throughput and reduce travel times. In
Davis’ strategy, an ACC equipped car adjusts its
position according to the front car on the other
road to create a safety gap without slowing down
sharply. However, it only considers the nearest
front car and thus the benefit is limited.
Ramp Metering
Ramp metering has been successfully imple-
mented since the mid-1960s as an efficient freeway
management strategy. All ramp metering strate-
gies calculate suitable ramp flow that is allowed
to the freeway. The ramp flow is converted to a
green-phase duration of the traffic light.
Ramp metering strategies are typically clas-
sified as follows:
• Fixed-time strategies are based on constant
historical demands and static models to
derive optimized settings. One example of
fixed-time strategies is Ideal Ramp Metering
(IMP) (Chen et al., 2001). To find out when a
freeway section is most efficient, i.e., when
there is maximum flow on the highway
section, Chen et al. analyze Los Angeles
highways data from Performance Measure-
ment System (PeMS) database (Chen et al.,
2001). They show that the maximum flow
occurs near 60 mph (≈ 100 km/h), which is
also the velocity under free flow conditions.
They assume that the pattern of demand is
unchanged. Therefore, to maintain the ve-
locity of 60 mph, IMP stops the ramp flow
whenever the occupancy exceeds the critical
189
 Proactive Traffic Merging Strategies for Sensor-Enabled Cars
occupancy (at which the freeway flow is
maximal). The main drawback of fixed-time
strategies is the absence of real-time data.
Fixed-time strategies are oversimplified, be-
cause demands are not constant and further
disturbances such as accidents may occur.
• Reactive strategies are based on real-time
measurements of the freeway flow up-stream
of the ramp and the freeway occupancy
downstream of the ramp. They aim to keep
the freeway traffic conditions close to pre-
specified set (desired) values. Reactive
strategies can be performed independently
for each ramp or extended to several ramps.
The ALINEA strategy outperforms other
strategies in field trials (Papageorgiou et
al., 2003) ALINEA can typically achieve a
20% decrease of the total travel time. The
limitations of reactive strategies are: 1) ap-
propriate set values are hard to get; and 2)
the scope of their control is local.
• Nonlinear optimal strategies are based on
macroscopic dynamic models to coordinate
ramp metering in freeway networks. The
total time spent by the vehicles in a free-
way network can be reduced by up to 50%
in simulations (Papageorgiou et al., 2003).
This type of strategies has not been imple-
mented in the field. According to (Chu et al.,
2004) two area-wide coordinated algorithms
BOTTLENECK and ZONE can outperform
the local reactive algorithm ALINEA.
In summary, ramp metering is an efficient
control approach. On the other hand, it presents
several drawbacks:
• Ramp metering is a sensitive technique. If
ramp metering is not accurate enough, then
it can lead to traffic congestion by weak me-
tering that overloads the mainstream flow,
or prevent fully utilize of the mainstream
capacity by overly strong metering.
190
• Ramp metering may increase the waiting
time for vehicles on the ramp, create ramp
queues and transfer the congestion problem
to the ramp and local street networks.
• Ramp metering including those reactive
strategies still have the limitation of adap-
tive to the highly variable and unpredictable
traffic conditions.
ProactiVE MERrging
Algorithm
This section presents our research into proactive
traffic merging algorithms. We show how sen-
sor-enabled cars can assist in improving merging
algorithms. We compare proactive merging algo-
rithms against a conventional merging strategy:
priority-based merging.
In a priority-based strategy, the cars from one
stream always have the right of way, for example,
cars from the ramp must give way to the cars on
the main road. The ramp car only merges when
the safety gap between two cars on the main
road is large enough. Meanwhile, the cars on the
main road completely ignore the ramp cars and
do not create gaps for the ramp car to merge.
The shortcoming of this strategy is the lack of
coordination between the two traffic streams,
which results in unnecessary delays for one traffic
stream. We explore proactive merging algorithms
with the aim of using the current road facilities
more efficiently, including the improvement of
overall traffic flow, decrease of the travel time
and economic fuel consumption.
Single Lane Scenario
We investigate three proactive merging algorithms
for a main road consisting of a single lane. All
algorithms assume that there is a decision start
point S before the merging start point O (see Figure
2). S indicates the point where a car decides where
and when to merge between O and E based on
different types of spatial information. The con-
Proactive Traffic Merging Strategies for Sensor-Enabled Cars

cept of sliding decision point makes a proactive Table 3. Traffic merging algorithms overview
merging algorithm more adaptive. The position
Knowledge Right of Way Assumption
of decision point can be changed according to dif-
ferent traffic conditions. For frequently changing The car
Distance- Velocity does
Position closest to the
traffic conditions it is better to choose a decision based
merging point
not vary much

point closer to the actual merging point because The car first Acceleration
Velocity- Position
it is difficult to predict changes. based Velocity
arrives at the does not vary
All algorithms are initiated by the ramp car merging point much

once it arrives at the decision start point S. Proac-

tive algorithms essentially determine which car
has the right of way by using either the shortest
distance to O or the shortest arrival time at O. The
first strategy is a distance-based merging algo-
rithm (referred to as D) that measures the distance
from the main road car to the merging point O
and compares it with the distance between S and
O (denoted as SO). If the main road car’s distance
is greater than SO, then the ramp car merges first
(because the ramp car’s distance is less than SO).
Otherwise, we recursively compare the distance δ
of the subsequent main road car and the distance
δ’ of the ramp car to O. We continue until δ >
δ’ in order to find the slot for the ramp car. The
second strategy is a velocity-based algorithm
(referred to as V) that also takes the velocity into
account. It calculates the time for a car to arrive
at the merging point and compares the time for
cars from the two traffic streams. The car with
shortest time merges first. The third strategy is
platoon-velocity-based merging algorithm (re-
ferred to as PV). It is a more refined version of
the velocity-based merging algorithm V. Instead
of comparing the ramp car against one main road
car as in V, PV compares the ramp car with a
group of main road cars on the segment SO and
assigns a slot between two main road cars for the
ramp car. The slot is chosen based on the arrival
time calculated at S.
In Table 3, we summarize distance-based
and velocity-based merging algorithms.

Figure 2. Notations of proactive merging algorithms

Figure 3. Illustration of lane switch in CPV



191
 Proactive Traffic Merging Strategies for Sensor-Enabled Cars
Each algorithm performs better than others un-
der certain traffic condition. A refined adaptive
algorithm switches to the strategy that is best fit
for the current traffic situation. More details of
single lane merging algorithms can be found in
(Wang et al., 2007).
Multiple Lane Scenario
For main roads with multiple lanes cars are able
to change lanes to overtake other cars or to make
space for new arriving ramp cars. Our experiments
will show that PV is the most effective proactive
merging algorithm for main road with a single
lane. Thus, we extend it to roads with multiple
lanes. In addition, we develop a refined version
of PV, which we call Cascading PV (referred to
as CPV).
CPV assumes two lanes but can be general-
ized to roads with more lanes. Cars on the outer
lane and the inner lane of the main road can
change lanes prior to the merging area to create
a larger gap for the merging ramp cars as shown
in Figure 3. CPV aims to maximize the minimum
safety distance on both lanes which might result
in a lane switch between two cars.
Multi-lane merging is an open research area
with contradictory reports. The simulation results
we show in the Discussion section demonstrate
that appropriate lane changing on the main traf-
fic stream before the merging area calculated by
CPV results in better performance than PV. Hidas
(2002; 2005) proposes a lane changing model
for merging scenarios. This model distinguishes
forced and cooperative lane change from free
lane change, which helps modelling the merging
behaviours under congested traffic conditions.
Sarvi and Kuwahara (2007) model the ramp car
behaviour and the main road back car behaviour.
They report that lane changing on the main road
before and/or after merging area can cause capac-
ity reduction. There may be a number of reasons
why lane changing can lead to different results for
192
different merging algorithms. These approaches
have different underlying microscopic traffic
models and use different criteria to determine
whether or not a car should change its lane. In
addition, all approaches use different simulations.
Most studies (Hidas, 2002; Hidas, 2005; Sarvi
& Kuwahara, 2007) simulate an open boundary
road section, whereas in our simulation the main
road is designed to be a loop. The loop provides
a feedback to show the impact of merging algo-
rithms. In summary, future research is required
to explain the different findings.
EVAaluation
This section shows how to assess merging al-
gorithms by identifying a range of performance
criteria, such as delay, flow, and fuel efficiency.
It will also discuss how and why sensor-enabled
cars can improve merging algorithms.
Performance Metrics
Merging algorithms can be evaluated from two
perspectives: safety and efficiency. Although all
merging algorithms have to ensure road safety
our major research aim is to optimize the traffic
throughput. In addition, we also consider other
factors than efficiency such as fuel consumption.
In this section, we introduce criteria for traffic
control algorithms and show how to measure
the performance. We identify which criteria de-
pend on each other and illustrate how optimized
throughput implies higher speeds of cars coupled
with tighter gaps, while safety requirements imply
lower speed and larger gaps.
Safety
To evaluate the safety characteristics of a merging
strategy we consider two types of events.
Proactive Traffic Merging Strategies for Sensor-Enabled Cars
• Collision: There is more than one car at
the same position of the road at the same
time.
• Near-collision: The gap between two suc-
cessive cars is smaller than the minimum
safety gap but if the front car does not brake
a collision does not occur.
We distinguish those two events and count
the number of collisions and the number of near-
collisions relative to the total number of cars in
a time frame.
Efficiency
A good merging strategy utilizes the road re-
sources more efficiently. We use the following
criteria to evaluate efficiency.
• Flow: The average number of cars passing
a certain position over a period of time. We
evaluate the flow as the product of density
and velocity over the entire main road. The
largest flow would occur if cars are bumper
to bumper (maximum density) and move at
the speed limit (maximum velocity). Clearly,
this is not achievable in reality for safety
reasons. If the density is maximal then cars
are not moving, yielding a minimum traffic
flow of zero. Therefore, the maximum of
traffic flow occurs at a certain density with
a corresponding velocity.
• Capacity: The maximum number of cars
passing a certain position over a period of
time, i.e., the maximum traffic flow. Traffic
congestion occurs when the demand exceeds
the capacity, thus achieving higher capac-
ity can effectively avoid traffic congestion.
However, the real capacity is typically lower
than the theoretical maximum capacity be-
cause it is not only determined by the road
configuration (e.g., the number of lanes, the
lane width) but also by disturbances in the
traffic flow, such as lane changing or merg-
ing. We expect that a merging strategy has
a significant impact on the real capacity of
a road. Therefore, the observed reduction of
the real capacity from the theoretical capac-
ity is an important criterion for merging
strategies.
• Stability: The ability to adapt to a new traf-
fic situation. It impacts the capacity and the
total number of merged cars. We expect that
a stable merging strategy achieves higher
capacity by accommodating more merged
cars. A merging car impacts the stability of
two traffic streams because in one traffic
stream a new car is added and in the other
stream this car is removed. As a result, the
cars of both streams adjust their speeds and
gaps to achieve approximate equilibrium
conditions. As the density of one stream
increases the gaps between the cars of this
stream decrease until no farther cars can
merge without violating the safety distance.
The gaps between the cars distribute more
evenly in the stable traffic flow than the
dynamic traffic flow. Thus, some larger gaps
exist in the dynamically changing traffic
flow and can accommodate more cars to
merge.
• Delay: The total required time for a strategy
to fill the target lane with a desired number
of merging cars. It describes how quickly
the system absorbs merging cars. In both
merging scenarios, the car (on the ramp
or on the blocked lane) locates a gap from
the adjacent lane and adjusts its speed to fit
in the gap in order to merge. A bad merg-
ing strategy may cause more delay for the
merging stream because the car has to wait
if it cannot adjust its speed appropriately to
merge.
• Fuel consumption: Varies as much as 45%
(Victoria, n.d.) between different drivers
using identical cars. Multiple studies show
that accelerating slower will dramatically
improve fuel efficiency: the author in (Reed,
193
 Proactive Traffic Merging Strategies for Sensor-Enabled Cars

2005) found fuel savings up to 37% with algorithm in a variety of simulation settings.
an average of 31% simply by accelerating Our work shows the following three parameters
a car slowly. Similar findings (Parker & are most important to measure the efficiency of
Stedman, 1994) show slower acceleration a merging algorithm: the ramp length, the initial
saving between 14% and 21% of fuel use over density of the main road, and the rate of incom-
aggressive acceleration. Different merging ing ramp cars. In the simulations, we initially
strategies have a significant impact on the study uniformly distributed traffic flow and then
acceleration. we study arrival ramp traffic following Poisson
distribution.
Simulation
Discussion
We developed a Java-based controlled simulation
environment to test various traffic merging strate- In the proactive merging strategies, the actual
gies. It is based on an existing microscopic traffic merging point is dissociated from the decision
model IDM (Treiber et al., 2000), which we pre- point where to merge. This decoupling results in
sented in the Longitudinal Models section. IDM a significant reduction of traffic merging delay at
specifies the following parameters: maximum the expense of a lower overall average velocity
velocity, safe time headway, maximum accel- on the main road. A lower velocity is expected
eration, comfortable deceleration, and minimum because higher traffic load on a road requires a
distance. Certain values guarantee collision-free smaller minimum safety distance, which is only
driving. We compare the performance of proac- achievable by a lower velocity. Figure 4 shows the
tive merging algorithms against a priority-based performance of different merging algorithms. The

Figure 4. The performance of merging algorithms with a short ramp of 100 m and high ramp flow rate

R PV CPV R PV CPV
400
Total Number of Cars

30
Delay (min)

300
20

200
10

100
0 1 5 10 15 20 25
110 150 190 230 270 310
Time (min)
Total Number of Cars

(a) (b)
R PV CPV
30
R PV CPV Max
velocity (m/s)

65
Flow (cars/min)

55
25
45
35
25
20
15 1 5 10 15 20 25
1 5 10 15 20 25 Time (min)
Time (min)

(c) (d)

194
Proactive Traffic Merging Strategies for Sensor-Enabled Cars
dashed line in Figure 4.c illustrates the maximal
flow in equilibrium condition. R refers to the
priority-based merging algorithm. The strategy
R performs less well because once a high load
is achieved R simply breaks, i.e., no farther cars
can merge. CPV is a well performing strategy.
It approaches the theoretical maximum flow.
As expected, CPV has a smaller overall average
velocity because it is the only way to achieve a
higher traffic load.
In general, the advantages of proactive merg-
ing algorithms over the priority-based algorithm
are the following:
• The delay in proactive merging algorithms
decreases by a third compared to the priority-
based algorithm before the loop is saturated.
Cars on the ramp can merge into the main
road more quickly.
• The capacity of the main road increases by
up to 67%.
• The traffic flow in proactive merging al-
gorithms is higher. Since the traffic flow is
defined by the product of the average velocity
and the density, there are more cars on the
main road in proactive merging algorithms
than the priority-based algorithm due to the
smaller delay. At first, the larger number of
cars overcomes the slightly smaller average
velocity; then as more cars merge in, the main
road starts getting saturated, the average ve-
locity drops to ensure the safety distance.
A disadvantage is the slightly lower average
velocity in proactive merging algorithms. This is
a side-effect of a larger number of cars merging
into the main road. The distance between cars
gets smaller, which leads to a lower velocity
because the velocity depends on the minimum
safety distance between cars. However, this
lower overall velocity is simply a byproduct of
the higher traffic flow.
We must manage complex tradeoffs among
several factors, in particular the delay, flow, and
velocity. The traffic flow increases at the begin-
ning as the ramp cars merge, peaks before the
main road is saturated, and then decreases as the
average velocity decreases. In order to maintain
a high traffic flow as well as to accommodate
a large number of ramp cars, we need to find a
balance between the velocity and the number of
cars that merge into the main road.
Overall, CPV leads to remarkable results in
terms of delay and flow. It assigns a slot for a ramp
car among a group of main road cars in the seg-
ment between the decision point and the merging
point, and creates a larger gap for the ramp car by
switching lanes of the cars on the main road prior
to the merging area. In summary, CPV optimizes
the delay and flow at the expense of the average
velocity on the main road. On the other hand, the
lower average velocity enables CPV to achieve a
significant increase for the total number of cars
on the main road.
Futurrnd
There are two research directions that we believe
will draw future attention:
• Robustness. Proactive merging strategies
rely on two assumptions. First, the spatial in-
formation monitored by sensor-enabled cars
is accurate and precise. Although automotive
sensors are typically required to have a total
error less than 3% over their entire measured
range (Fleming, 2001), Hernandez (2005)
also reports that for car speed lower than 5
km/h sensors give skewed and misleading
information. Consequently, future research
will need to focus on the impact of imper-
fect information. Second, we envision that
navigation systems not only provide route
guidance but also provide recommenda-
tions of how to drive cooperatively with the
vehicles in the vicinity. Therefore, a merg-
ing algorithm must accommodate various
195
 Proactive Traffic Merging Strategies for Sensor-Enabled Cars
human factors, such as reaction times, and
be prepared to handle drivers who do not
follow the system’s recommendations.
• Scalability. Currently, proactive merging
algorithms are localized and provide only
recommendations for a single intersection.
Highly scalable algorithms are required to
optimize traffic flows at a network level that
involves a large number of intersections as
well as multiple on-ramps and exit-ramps
at freeway networks. Scalability can be
achieved through locality in both time
and space. Sensor-enabled cars will detect
information in their spatial vicinity only
use it for decisions in the near future as
long-term estimations occur high compu-
tational complexity. As a result, we expect
that control strategies will be distributed
consisting of local decisions according to the
neighborhood conditions, instead of having
a single centralized system managing the
entire traffic network. We expect that such
a strategy will scale well. Novel models with
a sufficiently fine level of granularity will
attract further research.
ConcluSION
This chapter provides an overview of new ideas
and approaches in the area of traffic flow control
with the assistance of VANETs. We report on the
state-of-the-art in this area in terms of vehicular
traffic theories, traffic models, merging strategies,
and performance measurement methodologies.
In this chapter, we have highlighted proactive
merging strategies and the potential benefits of
applying sensor-enabled cars. We have shown in
our research the significant improvement in traf-
fic flow and decrease in travel time mainly result
from the decoupling of the merging point and
the decision point, and multilane optimizations,
196
such as pre-lane-changing. We have identified
potential research directions in the robustness
and scalability of merging algorithms. Assisted
by advanced sensing and communication tech-
nologies, traffic control strategies and merging
algorithms will lead to more efficient use of the
current road networks and ultimately help to al-
leviate traffic congestions.
REFERENCES
Federal Highway Administration. (2005). Traffic
congestion and reliability: Trends and advanced
strategies for congestion mitigation. (http://ops.
fhwa.dot.gov/congestion report/, last accessed:
April 2008)
Ahmed, K., Moshe, E., Koutsopoulos, H., &
Mishalani, R. (1996). Models of freeway lane
changing and gap acceptance behavior. In Proc.
ISTTT ’96 (pp. 501–515).
Akyildiz, I. F., Su, W., Sankarasubramaniam, Y.,
& Cayirci, E. (2002). Wireless sensor networks: a
survey. Computer Networks, 38 (4), 393–422.
Anda, J., LeBrun, J., Chuah, C.-N., Ghosal, D.,
& Zhang, H. M. (2005). VGrid: Vehicular ad hoc
networking and computing grid for intelligent
traffic control. In Proc. IEEE VTC 2005-spring,
5, 2905–2909.
Bando, M., Hasebe, K., Nakayama, A., Shibata,
A., & Sugiyama, Y. (1995). Phys. Rev. E, 51(2),
1035–1042.
Bevly, D. M. (2004). Global positioning system
(GPS): A low-cost velocity sensor for correcting
inertial sensor errors on ground vehicles. J. Dyn.
Syst. Meas. Contr., 126(2), 255–264.
Bonanno, G., Fantoni, R., Fichera, A., Fornetti, G.,
Moriconi, C., Poggi, C., et al. (2003). The sensing
subsystem of RAS.
Proactive Traffic Merging Strategies for Sensor-Enabled Cars
Bose, A., & Ioannou, P. A. (2003). Analysis of
traffic flow with mixed manual and semi-auto-
mated vehicles. IEEE Trans. Intell. Transport.
Syst., 4(4), 173–188.
Chen, C., Jia, Z., & Varaiya, P. (2001). Causes
and cures of highway congestion. IEEE Control
Syst. Mag., 21(6), 26–32.
Chen, C., Petty, K., Skabardonis, A., Varaiya, P.,
& Jia, Z. (2001). Freeway performance measure-
ment system: Mining loop detector data. Transp.
Res. Rec., 1748, 96–102.
Choffnes, D. R., & Bustamante, F. E. (2005). An
integrated mobility and traffic model for vehicu-
lar wireless networks. In Proc. VANET ’05 (pp.
69–78). New York, NY, USA: ACM.
Chowdhury, D., Santen, L., & Schadschneider,
A. (2000). Statistical physics of vehicular traf-
fic and some related systems. Physics Reports,
329, 199.
Chu, L., Liu, H. X., Recker, W., & Zhang, H.
M. (2004). Performance evaluation of adaptive
rampmetering algorithms using microscopic traf-
fic simulation model. J. Transp. Engrg., 130(3),
330–338.
CI-6558, P. (n.d.). Acceleration sensor. (http://
hug.phys.huji.ac.il/PHYS HUG/MAABADA/
MABADA A/Pasco/ci/ci-6558/012-06361b/012-
06361b.pdf, last accessed: April 2008).
Davis, L. C. (2004). Effect of adaptive cruise
control systems on traffic flow. Physical Review
E, 69(066110).
Davis, L. C. (2007). Effect of adaptive cruise
control systems on mixed traffic flow near an
on-ramp. Physica A, 379(274–290).
DICKEY-John Products. (n.d.). Radar II. (http://
www.dickey-john.com/products/public-works/
ground-speed-sensors/radar-ii, last accessed:
April 2008)
ElBatt, T., Goel, S. K., Holland, G., Krishnan, H.,
& Parikh, J. (2006). Safety: Cooperative collision
warning using dedicated short range wireless
communications. In Proc. VANET ’06 (pp. 30–39).
New York, NY, USA: ACM.
Estrin, D., Govindan, R., Heidemann, J., & Ku-
mar, S. (1999). Next century challenges: scalable
coordination in sensor networks. In MobiCom ’99
(pp. 263–270). New York, NY, USA: ACM.
Fleming, W. J. (2001). Overview of automotive
sensors. IEEE Sens. J., 1(4), 296–308.
Gartner, N. H., Pooran, F. J., & Andrews, C. M.
(2001). Implementation of the OPAC adaptive
control strategy in a traffic signal network. In
Proc. IEEE ITSC ’01 (pp. 195–200).
Gershenson, C. (2004). Self-organizing traffic
lights. Complex Systems, 16, 29.
Hall, R. W., & Tsao, H.-S. J. (1997). Capacity of
automated highway systems: merging efficiency.
In Proc. ACC ’97, 3, 2046–2050.
Helbing, D. (2001). Traffic and related self-driven
many-particle systems. Rev. Mod. Phys., 73(4),
1067–1141.
Hernandez, W. (2005). Robust multivariable
estimation of the relevant information coming
from a wheel speed sensor and an accelerometer
embedded in a car under performance tests. Sen-
sors, 5, 488–508.
Hidas, P. (2002). Modelling lane changing and
merging in microscopic traffic simulation. Trans-
port. Res. Part C, 10 (5-6), 351–371.
Hidas, P. (2005). Modelling vehicle interactions in
microscopic simulation of merging and weaving.
Transport. Res. Part C, 13(1), 37–62.
Hightower, J., & Barriello, G. (2001). Location
systems for ubiquitous computing. IEEE Com-
puter, 34(8), 57–66.
197
 Proactive Traffic Merging Strategies for Sensor-Enabled Cars

Johnson, D. B., & Maltz, D. A. (1996). Dynamic theory and optimal control. In Proc. IEEE CDC
source routing in ad hoc wireless networks. In ’96 (pp. 1190–1195).
Imielinski & Korth (Eds.), Mobile computing,
Lygeros, J., Godbole, D. N., & Sastry, S. (1998).
353. Kluwer Academic Publishers.
Verified hybrid controller for automated vehicles.
Kanaris, A., Kosmatopoulos, E. B., & Loannou, IEEE Trans. Autom. Control, 43(4), 522–539.
P. A. (2001). Strategies and spacing requirements
Nagel, K., & Schreckenberg, M. (1992). A cellular
for lane changing and merging in automated
automaton model for freeway traffic. J. Phys. I
highway systems. IEEE Trans. Veh. Technol.,
France, 2(12), 2221–2229.
50(6), 1568–1581.
Nakamura, E. F., Loureiro, A. A. F., & Frery, A.
Kato, S., & Tsugawa, S. (2001). Cooperative driv-
C. (2007). Information fusion for wireless sensor
ing of autonomous vehicles based on localization,
networks: Methods, models, and classifications.
inter-vehicle communications and vision systems.
ACM Comput. Surv., 39(3), A9/1–A9/55.
JSAE Rev., 22(4), 503–509.
Papageorgiou, M., Diakaki, C., Dinopoulou, V.,
Kerner, B. S. (2004). Three-phase traffic theory and
Kotsialos, A., & Wang, Y. (2003). Review of
highway capacity. Physica A, 333, 379–440.
road traffic control strategies. Proc. of the IEEE,
Kesting, A., Treiber, M., & Helbing, D. (2007). 91(12), 2043–2067.
General lane-changing model MOBIL for car-fol-
Papageorgiou, M., & Kotsialos, A. (2002). Freeway
lowing models. Transp. Res. Rec., 1999, 86–94.
ramp metering: an overview. IEEE Trans. Intell.
Kesting, A., Treiber, M., Schönhof, M., & Helbing, Transport. Syst., 3(4), 271–281.
D. (2008). Adaptive cruise control design for ac-
Parker, D. S., & Stedman, T. C. (1994). Energy
tive congestion avoidance. Transport. Res. Part
efficient transportation for Florida. (http://www.
C, doi:10.1016/j.trc.2007.12.004.
p2pays.org/ref/12/11355.pdf, last accessed: Nov.
Lammer, S., Kori, H., Peters, K., & Helbing, D. 2007)
(2006). Decentralised control of material or traffic
Pedersen, M. M., & Ruhoff, P. T. (2002). Entry
flows in networks using phase-synchronisation.
ramps in the Nagel-Schreckenberg model. Phys.
Physica A, 363(1), 39–47.
Rev. E, 65(5), 056705.
Li, L., Song, J., Wang, F. Y., Niehsen, W., &
Reed, P. (2005). We test the tips: What really
Zheng, N. N. (2005). IVS 05: New developments
saves gas? and how much? (http://www.edmunds.
and research trends for intelligent vehicles. IEEE
com/ownership/driving/articles/106842/article.
Intell. Syst., 20(4), 10–14.
html, last accessed: Nov. 2007).
Liang, C., & Peng, H. (2000). String stability
Saha, A. K., & Johnson, D. B. (2004). Model-
analysis of adaptive cruise controlled vehicles.
ing mobility for vehicular ad-hoc networks. In
JSME Intl. J., Series C, 43(3), 671–677.
Proc. VANET ’04 (pp. 91–92). New York, NY,
Lu, X. Y., Tan, H. S., Shladover, S. E., & Hedrick, USA: ACM.
J. K. (2004). Automated vehicle merging maneu-
Sarvi, M., & Kuwahara, M. (2007). Microsimu-
ver implementation for AHS. Vehicle Syst. Dyn.,
lation of freeway ramp merging processes under
41(2), 85–107.
congested traffic conditions. IEEE Trans. Intell.
Lygeros, J., Godbole, D. N., & Sastry, S. (1996). Transport. Syst., 8(3), 470–479.
Multiagent hybrid system design using game

198
Proactive Traffic Merging Strategies for Sensor-Enabled Cars
Treiber, M., & Helbing, D. (2002). Microsimula-
tions of freeway traffic including control mea-
sures. (http://arxiv.org/abs/cond-mat/0210096)
Treiber, M., Hennecke, A., & Helbing, D. (2000).
Congested traffic states in empirical observations
and microscopic simulations. Phys. Rev. E, 62(2),
1805–1824.
Treiber, M., Kesting, A., & Helbing, D. (2006).
Delays, inaccuracies and anticipation in micro-
scopic traffic models. Physica A, 360(1), 71–88.
Varaiya, P. (1993). Smart cars on smart roads:
Problems of control. IEEE Trans. Autom. Control,
38(2), 195–207.
Victoria, E. (n.d.). How you can save on fuel costs.
(http://www.epa.vic.gov.au/air/savefuel/savefuel.
asp, last accessed: Nov. 2007)
Vlacic, L., Parent, M., & Harashima, F. (2001).
Intelligent vehicle technologies: Theory and ap-
plications. London: Butterworth-Heinemann.
Wang, Z., Kulik, L., & Ramamohanarao, K.
(2007). Proactive traffic merging strategies for
sensor-enabled cars. In Proc. VANET ’07 (pp.
39–48). New York, NY, USA: ACM.
199
200

Chapter XI
The Localisation Problem in
Cooperative Vehicle
Applications
Javier Ibañez-Guzmán
RENAULT S.A.S., France

ABSTRACT

The next paradigm towards enhancing vehicle safety and road transportation represent cooperative
systems. Advances in computer and communications technologies are facilitating the establishment of
vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) wireless communications links. This enables
the sharing and aggregation of information which results in an extension of the driver awareness horizon
in an electronic manner. In this chapter, V2V and V2I applications are considered as a spatio-temporal
problem. The tenet is that sharing information can be made only if this is time stamped and related to
a spatial description of the information sources. The chapter formulates the spatio-temporal problem
having as constraint the precision of the pose estimates of the vehicles involved. It regards the localisa-
tion problem and accuracy of digital road maps as a combined issue that needs to be addressed for the
successful deployment of cooperative vehicle applications. Two case studies, intersection safely and
an overtaking manoeuvre are included. Recommendations on the precision limits of the vehicle pose
estimations and the potential uncertainties that need to be considered when designing V2V and V2I
applications complete the chapter.

Introduction simple thermo mechanical devices to complex

mechatronics systems where networks of micro-
Over the years motor vehicles have provided a computers plus a plethora of proprioceptive and
rapid means of transportation, they are today exteroceptive sensors control multiple functions.
part of our daily lives. These have evolved from The dependency on motor transport has resulted in

Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.
The Localisation Problem in Cooperative Vehicle Applications
traffic congestions, pollution and an unacceptable
level of accidents (Broughton & Walter, 2007).
Whilst in countries with advanced economies, the
number of accidents has decreased, in countries
with emerging economies accident tendencies
are on the increase, in most cases with disastrous
consequences.
Since the late seventies, the number of safety
features in all types of vehicles has increased.
These are known as passive safety systems like
seat belts, deformable chassis, etc. At the same time
legislation has hardened and road infrastructure
has improved. Currently the number of accidents
and road fatalities has decreased particularly in
Europe, Japan and the USA; however over the
past years a plateau is being reached and further
reductions are difficult. At the same time conges-
tion and pollution continues to increase whilst
society is becoming more concerned on the costs
of pollution to global warming and health under
an uncontrollable raise in fuel costs. In the past
years, the trend has been to develop active safety
systems as part of advanced driving assistance
systems (ADAS) like the use of video cameras
for obstacle detection, Radar or Ladars (Laser
detection and ranging sensors) for vehicle fol-
lowing, etc. In other words, through the use of
exteroceptive and proprioceptive sensors, it is
feasible to estimate the vehicle ego-state and
perceive its immediate surrounding environment.
The combination of the captured data results in
digital models that represent the vehicle immedi-
ate surrounding environment and is enabling the
development of advanced safety systems as it is
possible to inform and warn drivers of potential
risks or even act on the subject vehicle in the
case of pre-crash braking systems (Laugier C.,
Vasquez D., Yguel M., Fraichard Th. & Aycard
O., 2008, Wang H., Xu Jian, Ibañez-Guzmán, J,
Jarvis, R., Goh, T. & Chang C.W., 2000),. Whilst
this represents a major progress and should yield
to better safety, there are two major constraints:
Sensors: these have limited capabilities and only
a combination of several of them could warrant
limited safety; Cost: prices are high and only a
limited number of vehicles would be equipped
with full safety features. Further, vehicle onboard
sensors due to their physics and the limits in their
layout on the vehicles, means that the perception
horizons are limited which reduces the effective-
ness of several active safety applications.
Advances in computer and communications
technologies are facilitating the establishment of
vehicle to vehicle (V2V) and vehicle to infrastruc-
ture (V2I) wireless communications links (Cook
J.A., Kolmanovsky I.V., McNamara D., & Prasad
K.V., 2007, Srini V.P., 2006) . This enables the
sharing and aggregation of information and hence
to extend the driver awareness horizon. Within
this context, it is feasible to share information
between vehicles and infrastructure, to combine
this information and allow for a better machine
and driver understanding of its situation and
thus to extend electronically the driving horizon
(Cook J.A., Kolmanovsky I.V., McNamara D., &
Prasad K.V., 2007, Srini V.P., 2006) . A typical
example could be cooperative adaptive cruise
control (CACC), an extension of ACC, where in
addition to measuring the distance to the leader,
a vehicle can exchange information with the
leader by wireless communications to improve
performance (Arem B.van., Driel C. van., and
Visser R., 2006). At present multiple worldwide
efforts exist in this domain like the SKY project
on intersection safety by NISSAN Corp., which is
already in its field operational test phase involv-
ing hundreds of passenger cars. In Europe there
are two major large scale demonstrators’ projects
under the 6th research framework of the European
Community, namely the co-operative vehicle in-
frastructure systems (CVIS) and the Cooperative
vehicles and road infrastructure for road safety
(SAFESPOT) projects, (CVIS Project Partners,
2008 & SAFESPOT Project Partners, 2008).
V2V and V2I applications result from the
sharing of information between the various enti-
ties involved. The purpose is to extend the driver
awareness horizon by associating the information
201
 The Localisation Problem in Cooperative Vehicle Applications
generated in one vehicle, transmitting it to the
relevant vehicles within its immediate environ-
ment so as to share information. The association
is based in two important parameters, time and
space. In this chapter, V2V and V2I applications
are considered as a spatio-temporal problem. The
tenet is that the sharing of information can be
made only if this is time stamped and related to
a spatial description of the information sources.
That is, all applications need to have a spatial
description with respect to something and for
this to be usable they need to be referred not
only with respect to an absolute reference frame
but also to a spatial description of the environ-
ment like for example the position on a road,
the position of the vehicles with respect to an
intersection and so on. The scope of this chapter
does not include the study of techniques used to
enhance the vehicles pose (position and attitude)
estimates as these are addressed in much detail
by work elsewhere (Schlingelhof M., Betaille D.,
Bonnifait P. & Demaseure K., 2008, Lee K.W.,
Wijesoma S. & Ibanez-Guzman J., 2007 & Cui
Y.J. & Ge S.S. 2002).
This chapter is organised as follows, Section
2 states that cooperative vehicle applications are
governed by the solution of a spatio-temporal
problem where the main constraint is the precision
of the pose estimates of the vehicles involved in
the application. Section 3 formulates the localisa-
tion problem and accuracy of digital road maps
as a combined issue that needs to be addressed
for the successful deployment of cooperative
vehicle applications. Section 4 completes the
problem formulation by studying in a simple
manner the case of the use of V2V or V2I com-
munications to traverse safely a road intersection
and an overtaking manoeuvre. Finally, Section
5 presents a summary of the chapter including
recommendations and comments on the implica-
tions on the precision limits of the vehicle pose
estimations and the potential uncertainties that
need to be considered when designing V2V and
V2I applications.
202
CoopERratihicl
application a SSpatio-TEMmporal
pro
All types of vehicles travel on networks of roads
provided by the infrastructure, and the control
of this network is through road and traffic signs
to various levels of complexity. Within these
networks, sets of vehicles travel guided inde-
pendently but whose response is constrained by
the state of the road network. Communications
between the vehicles and infrastructure provides
a different network that allows for the dynamic
sharing of information amongst all the entities in
the road network.
Independent of the wireless communications
technologies and protocols deployed, two types
of links can be established: Between vehicles,
that is Vehicle to Vehicle (V2V) communications,
and between vehicles and the infrastructure or
road side units (RSU), also known as Vehicle to
Infrastructure (V2I).
The main feature is that the information ac-
quired by the vehicles or infrastructure via their
own sensors is shared with other vehicles or with
another part of the infrastructure via wireless
communications links. This inherent property of
Intelligent Transportation Systems (ITS) provides
opportunities for developing new applications. In
order to define the need for a Spatio-Temporal
formulation and to focus on the localisation prob-
lem, in this section a succinct list of applications
is introduced. For this purpose, applications have
been classified into four families: Safety, Mobility,
Efficiency and Telematics. Details of each family
are given along the following structure: definition,
case study, illustrative examples, technological
challenges, and Localisation Information. The
classification follows the one defined by the Re-
search and Innovative Technology Administra-
tion (RITA), U.S. Department of Transportation
(Department of Transport, 2008). The emphasis
is according to the level of position accuracy
required by the applications, with those related
to Safety addressed first.
The Localisation Problem in Cooperative Vehicle Applications
Safety
A fundamental function for a driver is to be aware
of the relationships between its vehicle (subject
vehicle-SV), the other vehicles (principal other
vehicles-POV), and other entities within its sur-
rounding environment. Today, this awareness is
assisted through the use of a series of onboard
perception systems that detect obstacles, track
vehicles, assist in the vehicle control process, etc.
These are known as driving assistance systems
whose details are found in the literature (Gharavi
H., Prasad K. V. & Ioannou P.A. (Eds.), 2007).
These systems are designed to reduce risk during
travel by providing means to avoid collisions/
crashes. For this purpose, a series of sensors are
embarked onto the vehicles to different levels of
complexity.
Experience has shown that there is no ideal
sensor, as all types of sensors have limitations and
hence there is the need to use multiple sources of
information by combining their outputs. For this
purpose, sensor fusion methods are used to build
models of the environment that are then used for
decision making and hence informing or warn-
ing the driver, or even acting on the vehicle. As
demonstrated by the landmark event on vehicle
navigation, the DARPA Urban Grand Challenge,
today it is possible to generate enriched infor-
mation with a good degree of certainty so as to
guide autonomously ground vehicles in urban
like environments (Defense Advanced Research
Projects Agency, 2008). These technologies can
be transferred to passenger vehicles and are today
part of the technological roadmaps of vehicle
OEMs. A major constraint is that for this purpose,
a combination of sensors is needed operating at
fast sampling rates and having a high resolution.
Further, due to the geometrical characteristics
of passenger vehicles, the layout of these sensors
are such that they will always have a limited field
of view (Kelly, A. & Stentz, A., 1998), limiting
their capabilities.
In order to illustrate this problem, the following
use case is defined: Intersection Safety.
Use Case Generic Intersection Safety
Figure 1. presents a schematic view of a typical
intersection. For description purposes and to
ensure the consistency of the descriptions, the
vehicles have been classified as follows: Subject
vehicle (SV) is the centre of interest and with
respect to which the application or function is de-
rived. Intruder Vehicle (IV) represents the vehicle
that is to cross the path of the SV representing a
direct risk and thus the likelihood of a collision.
The Distracter Vehicle (DV) represents vehicle
not directly involved in the use case, but rather it
acts as an occlusion or is within the neighbour-
hood of the SV. There are other special vehicles
as listed in Figure 1, which are used to complete
the scenarios.
For the case of the generic intersection safety
use case, the subject vehicle (SV) arrives to it thus
the driver needs to be aware of the presence of all
the vehicles or vulnerable road users that represent
a risk. It is assumed that the SV is equipped with
front looking perception sensors (e.g. Ladars,
Vision Cameras) that are able to detect objects
obstructing its path. Making an abstraction of
reliability and costs of these systems, it can be
observed that the physics of the process limit
the driver awareness horizon. Although addi-
tional sensors could be included occlusion, their
limited field of view and layout on the vehicle
are major concerns (Gharavi H., Prasad K. V. &
Ioannou P.A. (Eds.), 2007). If speeds increase or
the environment is more cluttered, the problem
is compounded.
Observing the scenario in Figure 1 from the
driver or sensor point of view, it is clear that
the presence of the Distracter Vehicle (DV) and
that of the Two-Wheel vehicle is ignored; both
are occluded and thus constitute a risk. Further,
supposing that there is only the Intrusion Vehicle
(IV) and that this arrives at a prohibited speed, it
203
 The Localisation Problem in Cooperative Vehicle Applications

Figure 1. Schematic of a typical intersection in an urban setting

Figure 2. Schematic view of the SV environment using V2V communications

esv

SV
SV
POV

204
The Localisation Problem in Cooperative Vehicle Applications
is impossible for the driver of the SV to know that
the IV will not have sufficient space to brake at
the stop line due to the speed that it is travelling,
presenting a risk to the SV. Multiple situations
exist like for example if the High Tonnage Vehicle
(HTV) or truck is at a stop, the SV or the Power
Two-Wheel vehicle (PTW) are at a risk as they
are not aware of such situation. If each of the ve-
hicles can transmit its position, speed and other
status data, by associating this information onto
digital representation of the context (intersection),
an extended awareness can be created, as shown
in Figure 2. An application running in the SV
then can analyse and identify the possible risks
informing consequently the driver beyond what
current sensors could provide. For example the
DV and the two-wheeler are highlighted in red as
potential risks as shown in Figure 2. The vehicle
at the rear of the SV could also be warned that it
is driving too close, and warning information is
sent via a wireless link.
From this simplistic diagram, it can be ob-
served that by knowing the spatio-temporal rela-
tionships between all the actors, a full awareness
model can be created beyond the capabilities of
current sensor technologies and ideally avoiding
their need altogether, which results in the interest
in V2V and V2I technologies.
Applications
There are a plethora of applications related to
vehicle safety, which involve knowing the location
of the entities of interest at a given time and their
relative situations (e.g. speeds, inter-distances,
etc.). In general, these are known relative to the
environment or context which provides informa-
tion that leads to the establishment of relationships
between all objects, that is the complete perceived
model of the area of interest (Madhavan R., Mes-
sina E.R. & Albus J.S., 2007 & Papp Z., Brown
C.L., & Bartels C. 2008).
A good panorama of safety related applications
can be found at the SAFESPOT site (SAFESPOT
Project Partners, 2008). Sample applications
include:
Road Hazard Warning. The SV has detected
a hazard along its trajectory; this could be an
accident, a broken vehicle, road works, etc. This
information is geolocalised and broadcasted
through the VANET (Vehicular Ad-Hoc Network)
to the immediate vehicles.
Lane Change Assistance. Changing lanes
could create risk situations, due to the presence of
vehicles in the driver blind zone. By knowing what
occurs within the vicinity of the SV, such as the
distribution of vehicles on the appropriate lanes,
their relative speeds and inferring the intentions
of the driver, the system could generate an advice
to the driver on whether or not to concretise the
manoeuvre. The use of a blind spot detector will
be unnecessary.
Overtaking Assistance. The overtaking ma-
noeuvre involves a degree of risk. Drivers need
to have a clear line of sight of what is in front or
behind the vehicle, and the relative speeds of all
the observed vehicles, with respect to their own
speed as well as the shape of the road. When
drivers’ cognitive and perception capabilities are
reduced as they become older, or when visibility
is poor, overtaking becomes potentially more dan-
gerous. By knowing the position/direction/speed
of the various intervening vehicles, with relation
to the SV plus context (road), it will be possible
to infer whether there is a high risk involved and
consequently warn the driver.
Technical Challenges
For safety applications, it is necessary to know
precisely and on time the location of the SV and
its relationships with other road entities within
its immediate manoeuvring area. There is little
room for uncertainty or malfunctioning of the
system. That is no detection/location mistakes are
acceptable and such systems have to prove their
field reliability. Knowing the vehicle ego-state is
something achievable, however, determining the
205
 The Localisation Problem in Cooperative Vehicle Applications
vehicle position and projecting it onto a digital road
representation is much more difficult (Schlingel-
hof M., Betaille D., Bonnifait P. & Demaseure K.,
2008, Lee K.W., Wijesoma S. & Ibanez-Guzman
J., 2007 & Cui Y.J. & Ge S.S. 2002).
If data from on board exteroceptive sensors
is to be shared, this poses different challenges in
particular on the integrity of this data, in terms
of the degree of confidence on the detection ef-
fected, and its location with respect to the sensing
vehicle.
Other issues to take into account are for ex-
ample information latency, capacity of the com-
munication channels and sampling rates.
Mobility
Improving mobility by reducing delay and travel
times is a major objective of many ITS (Intelligent
Transportation Systems) components (Depart-
ment of Transport, 2008). Increases in traffic
density over the years have become problematic;
e.g. costly, increasing pollution, etc. By being
able to collect information from vehicles travel-
ling over a wide area, from networks of sensors
distributed strategically and from public service
organisations/road operators/public authorities,
it is possible to centralise this data to fuse it
and infer relevant information that leads to the
regulation of traffic, the broadcasting of location
dependent dynamic information and to address
traffic disturbances like accidents. In this case
there is observability of the traffic conditions in
an instantaneous manner and thus controllabil-
ity of the road network could be attained. Each
communicating vehicle could be considered as
a sensor, and the aggregation of the information
results in a distributed centralised data fusion
problem where statistics will be generated, and
control actions will be broadcasted to vehicles and
infrastructure. The sharing of information from
the infrastructure to all vehicles on an area through
the VANET to extend the driving awareness of
the drivers is an important characteristic.
206
Use Case: Geolocalised Broadcasting
of Information
Over a road segment it is necessary to control
virtually the flow of traffic. For example the infra-
structure could broadcast an event over an area,
such as the need for a free corridor to facilitate the
crossing of an Emergency Service Vehicle (ESV).
For this purpose, the information is broadcasted
from the infrastructure or through multi-hoping
between the relevant vehicles, as shown in Figure
3. The relevant vehicles will then form a blue cor-
ridor for the save crossing of the ESV.
The schematic in Figure 3 shows the manner
in which the information broadcasted can be
directed to a target area, so that the informa-
tion can be used to control the flow of vehicles
within that area. In this case, drivers within the
neighbourhood receive information on the arrival
of an ESV associated to a lane, and accordingly
the implicated drivers will leave the lane open
for the traversal of the ESV.
Applications
Controlling vehicle mobility in a geolocalised
manner is a group of applications that can be
implemented using ITS technologies. These as-
sociate information on other vehicles or events to
the road network in a precise manner. For example
in Figure 3, if there was a parallel road next to
the one where the ESV is travelling and thus the
information might be associated to the wrong road,
the result could be catastrophic, hence there is a
need for precise location information.
Most of the applications addressed in projects
like CVIS include this type of applications (CVIS
Project Partners, 2008). Sample applications
include:
Navigation and Traffic alerts. Information
broadcasts extend the situation awareness hori-
zon of drivers that should allow for vehicle speed
optimisation, and best trajectory recommenda-
tions based on up to date traffic information. To
The Localisation Problem in Cooperative Vehicle Applications
Figure 3. Traffic control broadcast information
Visualisation within the vehicles
Subject Vehicles
improve traffic flow, the usage of lanes will be
organised and speed recommendations will also
be issued, etc.
Presence of a Slow Moving Vehicle. Drivers
will be notified of the presence of a slow moving
vehicle on a given lane. Upon detection of such
vehicle the information is broadcasted including
geolocalisation data. This will be transmitted
to vehicles in the neighborhood. These vehicles
will then re-broadcast the information to other
vehicles, and those heading in the same direction as
the slow vehicle will warn drivers of this issue.
Monitoring of traffic conditions. In a way
all vehicles could be regarded as a collection
of wireless sensors that share their information
locally and globally. Locally detailed models of
the environment could be built using distributed
decentralized fusion techniques whilst globally
all data is collected by a server. Relevant infor-
mation will be generated using centralized data
fusion and data mining techniques. This will be
broadcasted to the relevant areas with its integrity
ensured via statistical techniques.
Efficiency
The sharing of information over wide areas al-
lows for the optimisation of resources, with the
transmission of relevant information on time in a
geolocalised manner. This results in an intelligent
infrastructure leading to intelligent environments
where it will be possible to deploy a collection
of vehicles with limited sensing and intelligent
capabilities, that is, the efficiency of existing
facilities is optimised so that mobility and com-
merce needs can be met while reducing the needs
to construct or expand new facilities (Department
of Transport, 2008). The capacity of the road
network is increased whilst at the same time by
optimising the travelling speeds of an ensemble
of vehicles so that it is possible to optimise their
own efficiency i.e. fuel consumption. Applications
that lie within the efficiency domain augment the
throughput of the road network that is the number
of persons, goods, or vehicles traversing under
safety constraints.
207

Use Case Traffic Jam: Speed Regulation
Over a road segment, compared to the theoretical
speed on that segment the actual speed of vehicles
on it is much slower. Ego state data from these
vehicles is transmitted automatically to the nearest
road side unit (RSU). This is compared statisti-
cally and as a result the RSU starts to transmit to
the vehicles travelling along the same direction
and within the neighbourhood of the segment
that there is a traffic jam. At the same time data
is uploaded to a central server in order to adapt
the speeds of the vehicles leading to that area or
proposing alternative routes. Figure 4 shows a
schematic representation of this situation.
A major question resides on the number of
vehicles necessary for such an application to be
viable. Floating car data could be available from
a limited number of vehicles (SV) as in Figure
4, which could transmit the relevant information
to other vehicles. In the figure it is possible to
observe that some SV vehicles establish links
with other vehicles, those traveling too close to
let them know the risk exists.
If more precise information is available with
respect to a lane in this section of the road, then
only vehicles along this lane will be informed,
however errors in the estimation of the alloca-
tion of the vehicles to a given lane could result
on an increase of risk as the information will be
transmitted to the wrong vehicles resulting in
more congestion or even accidents.
Figure 4. Traffic jam notification
IV IV SV
SV POV
POV SV SV POV POV
Traffic Jam
208
The Localisation Problem in Cooperative Vehicle Applications
Applications
Communications and position information of all
vehicles in the road network enable its overall
dynamic observability. Therefore, as in any
control problem, it could be possible to control
the flow and hence optimise its capacity. Sample
applications include:
Synchronisation of Vehicle Speeds to Traffic
Lights. The traversal of long roads controlled by
traffic lights could be optimised by synchronis-
ing the speed of the vehicles as they reach the
traffic lights so as to reduce the brusque speed
reductions and thus optimise flow and reduce
fuel consumption.
Dynamic Platoons. Ensembles of vehicles hav-
ing cruise control functions could be located in
platooning configurations so that they can travel
at faster speeds and uniformly. For this purpose,
the precise inter vehicle distances and speed in-
formation is important as well as certitude of the
lane in which the vehicles are travelling.
Customised Lane Assignment. Vehicles of a
certain class can be restricted or assigned to a
lane and their behaviour can also be monitored
if necessary. Long haul trucks will be kept on re-
stricted lanes at certain times of the year. Vehicles
participating in car pool schemes will be given
priority on the use of fast lanes, etc.
Telematics
Within the framework of this chapter, Telematics
applications comprise the conveying essentially
vast information over networks where nodes are
static and mobile, that is, information is provided
via service and content providers. The points of
interaction are infrastructure providers and to
SV
some extent Vehicle manufacturers. Figure 5
IV SV
shows a classical infrastructure for the provi-
sion of Telematics services where the consumer
POV
is centre on which the vehicle user will accept.
The interaction with the vehicle could be made
RSU
either with the vehicle in steady state conditions
The Localisation Problem in Cooperative Vehicle Applications
Figure 5. A sample structure for telematics ap-
plications
or in motion within the transmission range of
road side units.
In a typical application, there are transactions
either service or content which is communicated
to the vehicle driver; for this it is sufficient to
know that the vehicle is within communication’s
reach during the transaction period. At the same
time information from the vehicle could be up-
loaded.
Applications
The focus of the applications is transactions of all
types where there is the exchange of information
and services. It can be triggered at the request of
the driver or proposed as the vehicle traverses next
to a point of exchange in the infrastructure. In
general Telematics related applications operate on
transactions or information exchange and do not
include data linked to the immediate manoeuvr-
ing of the subject vehicle. The degree of position
information accuracy is low, as transactions are
mainly done on a per zone basis, unlike safety
related applications where precise information of
the vehicle whereabouts is required. In this section
sample applications are only included:
Location Based Electronic Commerce. As
the vehicle enters a zone a series of services are
provided e.g. the purchase of the local tourist
guide. This is downloaded upon payment via an
electronic wallet. The booking of an event within
the proximity can be made. The transaction is peer
to peer and should reduce deployment costs.
Downloading of Multimedia Information.
Infrastructure points allow for the reception of
large amounts of information. This can be in the
form of music files, films or specific documents.
For this purpose due to the volume involved, it is
preferred to be implemented next to road infra-
structure elements and perhaps through different
RF (Radio Frequency) channels to those used in
ITS applications.
Updating of Navigation Information. The
provision of digital representations of the road
network is a fundamental component in ITS appli-
cations. It allows for the use of a priori information
that is not only used for car navigation purposes
but also associated to information captured by
the vehicle onboard sensors so as to build models
for situation understanding purposes. A major
problem today is the relevance and integrity of
this stored digital map information, that is, map
updating is a major issue and hence is a major
Telematics application.
LocaliSAation CONSTtraint in
raltrafficENVIironm
From the panorama of applications presented in
Section 2, it can be observed that it is not suf-
ficient to know the position of the vehicles (with
respect to a global navigation reference frame
like the WGS 84 geodetic datum) but expressed
relative to the road segment travelled by the ve-
hicles. Current car navigation systems function
in that manner that is the absolute position of
the vehicle estimated by a combination of data
from a GPS receiver, odometry, and a gyroscope
is related to the spatial representation of digital
roadmaps (nodes and links) via a process known
as map-matching (Ardeshiri T. et al. 2006). This
allows the representation of the location of the
vehicle with respect to the context, such as the
digital maps.
209
 The Localisation Problem in Cooperative Vehicle Applications

In cooperative vehicle applications, this is
fundamental as all information shared from other
vehicles or infrastructure will be fused in the
subject vehicle. That is the common denominator
is the spatial representation of the road geometry
and its attributes with respect to which any ad-
ditional information is enriched so as to create a
model of the environment that is used for machine
interpretable situation awareness for driving
conditions.
Currently there are two major constraints on
the deployment of cooperative applications:
Map Accuracy
Maps are networks of nodes connected by links,
to which a series of attributes are assigned. Nodes
are described spatially with respect to an absolute
reference frame to facilitate association with the
location estimates. Maps are mainly provided by
two companies, namely TeleAtlas and Navteq.
In these maps, all types of public traversable
roads should be detailed, which are built through
several means including terrain surveys, the
purchase of existing maps and satellite data. It
should be pointed that: Nodes are represented
within an accuracy of 15m/48ft for detailed City
and Connector Roads and within 100m/328ft
for Inter-Town. Only roads known as Advanced
Driving Assistance Systems (ADAS) compliant
have a spatial description of equal or less than 5m.
Further, changes to the road infrastructure have to
be reflected in the digital maps. Currently this is
difficult and map updates are long to take effect,
for example, a year could have transpired before
this information could be available to the user.
If the map is no longer relevant, the applications
relying on it will collapse.
like GPS. GNSS based location estimates are
calculated by carefully timing the signals sent
by the constellation of satellites orbiting around
the Earth, which transmit messages based on a
common time plus the satellite orbit. In prin-
ciple, four satellites are used in order to take
into account time differences. The accuracy is
dependent on the satellite position and signal
delay, the latter occurs due to inconsistencies in
atmospheric conditions (ionosphere). There are
also ephemeris errors, satellite distortion and so
on, which will result in errors of the order of 10
to 15m. In urban conditions, as show in Figure
6, considerable errors could exist, which are the
results of multipaths (the bouncing of signals from
neighboring buildings) or occlusion, that is, there
is an insufficient number of visible satellites like
in urban canyons. Since the errors can be large
as shown in Figure 6, an application might not
work normally in such conditions.
Figure 6 shows the trajectory of a vehicle
recorded using a high-end Inertial Measurement
Unit (IMU) coupled to a surveyor level GPS plus
vehicle odometry in a loosely coupled manner.
The quality of the IMU gyroscopes and acceler-
Figure 6. Response of Automotive type GPS re-
ceivers in an Urban Canyon

Localisation Estimates

The location of a vehicle is estimated based on

the sole sources of absolute location information,
the Global Navigation Satellite Systems (GNSS),

210
The Localisation Problem in Cooperative Vehicle Applications
ometers allow for the estimation of the vehicle
trajectory with high precision, which is referred
to the ground truth, the continuous trajectory
in Figure 6. The other trajectories represent the
estimates of several COTS low cost GPS receiv-
ers recorded at the same time. The response of
each is very different and jumps up to 36 m were
recorded. The source of errors is mainly due to
multipaths and occlusion of GPS signals. To re-
duce for the lack of GPS signals and compensate
errors, the fusion with the vehicle odometers plus
gyroscope is made. However when large errors
occur or masking is over long periods, this type
of fusion is not sufficient, hence, the interest in
academia and industry to solve the localisation
problem (Lee K. W. 2007 & Toledo-Moreo R. et
al. 2007).
However, GPS alone is insufficient because
estimation errors exists due to position The
position calculated by a GPS receiver requires
the current time, the position of the satellite and
the measured delay of the received signal. The
position accuracy is primarily dependent on the
satellite position and signal delay.
Figure 7. Distribution of localisation accuracy requirements per application domain. (Adapted from
Gillièron P.Y. & Konnen J. 2003)
Localisation: A Combined Problem
for Cooperative Vhicles Applications
In the previous sections it has been shown that
the localisation problem for cooperative vehicle
applications is a combined problem as both map
and location accuracy introduces constraints. An
interpretation of these constraints on the applica-
tions can be represented in the study by Guilleron
and Konnen and shown in Figure 7 (Gillièron P.Y.
& Konnen J. 2003).
There is interdependency between the require-
ments in terms of vehicle location estimations
and accuracy of the digital maps, that is, both
requirements need to be complied with. Safety:
applications dealing with safety depend on the
precise localisation of the vehicle relative to the
digital map. In this case, map-matching errors
with longitudinal and lateral positioning errors
being inferior to 5 m are not allowed, particularly
in lane level accuracy where sub-meter level ac-
curacy is needed. Mobility: the requirements are
less stringent, and tolerated errors are less than
8 m. The main constraint is to ensure that the
correct roads are identified and the correspond-
211
 The Localisation Problem in Cooperative Vehicle Applications
ing vehicles are associated to them. Efficiency:
accuracy in this case should be around the 10 m
mark, and the dynamic sharing of information
between vehicles should reduce the need for
higher accuracy levels. Contrary to the Safety
related applications, Mobility and Efficiency ap-
plications are concerned with larger numbers of
vehicles distributed over large areas and thus the
requirements on knowledge of their presence in
these areas should be at times sufficient. Telemat-
ics: the requirements are different. They should
operate within a zone basis and also rely very
much on the identification of the vehicle within
the application zone. The accuracy should be
around 15m. Most applications at the transaction
level are mainly large volumes of data.
Analyof SPSpatial n on
rntati
The spatial and temporal relationships between
vehicles and their environment provide indications
on whether or not a potential safety risk (collision)
might exist. They are used as part of machine
inferred situational analysis to trigger different
driving assistance functions. Within this context
two important parameters could be considered:
Reaction time and Manoeuvre constraints. The
first takes into account the vehicle overall response
time including the driver and vehicle reaction
times; it is specific to the time taken by a vehicle
to brake. The second considers the ability of the
vehicle to change its trajectory by including the
vehicle kinematic and dynamic constraints. Most
safety related driving assistance functions are
based on the ability to warn drivers on time so
they can reduce their speed or move to a halt in
time. That is to inform drivers of potential haz-
ards with sufficient time that eventually they are
able to stop on time or reduce the vehicle speed
so as to avoid any safety hazard situation. The
same applies to the timely trajectory control that
is drivers need to manoeuvre at the correct time
212
and the position. Considerations of latencies in the
whole process are important as timing is crucial
for the relevance of the information. Further, most
information is spatial like the relative position of
the vehicles between each other or with respect to
a road feature like an intersection. Errors in both
time and location would render any application
of sharing this information across the vehicles
and infrastructure, which is no longer possible
as uncertainty introduces potential liabilities or
safety hazards.
To illustrate this section two sample cases
are considered: The first addresses the reaction
time and the vehicle location as applied to a road
intersection assistance application. In this case the
dependency of the relative position of the vehicles
related to their ability to stop in time is examined.
The purpose is to ensure that relevant vehicle(s)
stops to allow the respective vehicle to cross
the intersection safely. The second case relates
to the vehicle location to its trajectory change
capabilities. The purpose is to demonstrate the
importance of location information with regard to
the vehicle context in terms of time and space. In
this case, the relevant information has to be sent
to the vehicles concerned only in time so as to
become applicable. The scenario used is that of a
sudden blockage at a motorway exit that implies
that the pertinent vehicles within the immediate
environment are made aware of the situation so
they can take the most appropriate measures, like
changing their trajectories.
Road Intersection Assistance
The drawing in Figure 8 shows a typical situa-
tion that could occur as several vehicles converge
towards an intersection. The vehicle running
the application is labelled as the Subject Vehicle
(SV), while the vehicles involved in the study
are the Intruder Vehicle (IV), and the Power Two
Wheeler-motor bicycle (PTW). In the scenario,
the SV has the right of way and the IV and PTW
should respect the stop sign and reach a zero speed
The Localisation Problem in Cooperative Vehicle Applications
before the stop line. If the SV knows the position
of the IV, PTW and other vehicles, as well as its
own with respect to the intersection, plus their
corresponding speeds, it will be possible to know
if the vehicles could stop on time before entering
the intersection. For this purpose it is possible
to estimate the braking distance of the relevant
vehicles (IV and PTW) and infer whether or not
the vehicles could stop on time.
It is assumed that the vehicles’ positions and
the digital road map are accurate so it is possible
to estimate the vehicle speed and position with
respect to the intersection whether or not the
vehicles could stop on time when required as in
the case of the IV and the PTW.
For a V2V or V2I application, a major consid-
eration is the time allocated to interact with the
driver so a warning can be sent to the SV drivers
on time as they arrive to the intersection. There
Figure 8. Scenario depicting the arrival of several vehicles to an Intersection
DV
VRU
POV IV
S V = su b je ct ve h icle
IV = in tru sio n ve h icle
D V = d istra cte r ve h icle
H T V = h ig h to n n a g e ve h icle
V R U = vu ln e ra b le ro a d u se r
P T W = p o w e r tw o w h e e l ve h icle
POV
are two contextual parameters: One is the Time to
Intersection (T2I), which is the interval that the
SV or IV or PTW will need to arrive (stop) to the
intersection after a given event. The T2I depends
on the vehicle initial speed, the environment and
the action that the driver will take. The other is
the Response time that governs the T2I. That is
the timely response to external stimuli from the
driver and vehicle. It shows whether it is necessary
to look ahead for a given speed and response time
(Myopia), and whether the response time is too
large for a given speed and look ahead (Latency
Problem). A complementary issue is throughput
that is guaranteeing adequate sensory and pro-
cessing throughput. In a stand-alone perception
system, this refers to the processing needed in
terms of resolution, stabilisation and occlusion
compensation (Kelly, A. & Stentz, A., 1998).
Within a cooperative vehicle context, it should
include delays in the communication process.
VRU
DV
PTW
HTv
SV
213
 The Localisation Problem in Cooperative Vehicle Applications
The distance to the intersection (D2I) for a
stand-alone vehicle relates to the computational
reaction time and mechanical manoeuvrability.
It is the distance needed to arrive safely (stop or
under a controllable speed) to the intersection, as
a reaction to observations on the environment.
For this purpose, the vehicle dynamics and the
spatial extent of both vehicle and the intersection
are considered.
The response time governs most of the deci-
sions/actions; it can be defined as a perceive-think-
act loop that models the overall vehicle control
and planning system. The system response time,
is the time period between the instant an object
(entity) appears in the field of view (FOV) of the
perception sensor and the instant the vehicle has
completed the execution of the associate avoidance
trajectory. In a cooperative system the FOV will
be the electronic horizon (EH) which is larger
than the FOV. This time comprises:
• Tperc: perception time, includes detection/sens-
ing the environment, location of the detected
object, recognition and understanding
• Tplan: deciding what to do, which is the action
time allocated to the driver
• Tcom: commanding the actuators, e.g. brakes,
throttle, steering wheel
• Tact: actuator response time, i.e. any time
lag between the input command and the
actuator response
• Tveh: vehicle response, associated to the
vehicle dynamics.
Tresp = Tperc + Tplan + Tcom + Tact + Tveh (1)
The total system response time is defined by
equation (1):
Treact = Tperc + Tplan + Tcom + Tact (2)
It is possible to distinguish the time before the
driver response including the actuator response,
from the time after. The former is the reaction time
214
whilst the latter is the manoeuvre time, indicated
in equations (2) and (3), respectively.
Tmanoeuvre = Tveh (3)
The distance travelled during the reaction time
is calculated by the cruising speed and thus linear
in initial velocity, whilst the distance travelled
during the manoeuvre time will be quadratic, most
likely, as the vehicle reaches a zero speed.
For cooperative vehicles, if information is
received from distributed sources (vehicles and
infrastructure), the cooperative system time Tcoop
should be added to the response time Tresp.
The actuation space of a vehicle is covered by
a command vector (wrench) in terms of propulsive
rotational effort (heading), propulsive linear effort
(speed) and resistive linear effort (deceleration).
For the purposes of this analysis, the vehicle is
considered as a point object under the assumption
that there is an instantaneous response of actua-
tors to their commands.
The total travelled distance (Sawo) is defined
from the moment an event has been detected,
the awareness that an event has occurred (e.g.
a warning message received, an object appears
within the horizon, an intersection is detected,
etc.) to the time the vehicle moves to a halt. It is
proportional to the vehicle cruising speed, stop-
ping distance, and driver reaction time. It is the
distance travelled by the vehicle during the reac-
tion time Treact times the vehicle cruising speed Vac
plus the vehicle braking distance Sbrake.
To estimate the braking distance, it is assumed
that a full brake command is issued, when the
vehicle is travelling at a constant speed Vac in a
straight line. The sliding friction coefficient (µ)
between the vehicle and ground is taken as equal
to 0.65 when the road surface is wet, and equal
to 1 when it is dry. If there is a positive slope, the
vehicle will stop earlier and if this is negative, the
stopping distance will be longer. The road gradient
(f) is incorporated and expressed in m/100m, as
this affects the braking distance when driving in
The Localisation Problem in Cooperative Vehicle Applications
irregular terrain (this might not be as important
when considering intersections). The expression
for the Braking Distance will be:
Vac2
Sbrake = (4)
2g( ± f )
Vac2
S AWO = TreactVac + (5)
2g( ± f )
Therefore, the response or travelled distance
is: equation (5).
The first term is the reaction distance, and the
second the manoeuvre distance. The distances
increase in a quadratic manner as a function of
the vehicle speed. Therefore, the separation in
stopping distances increases at higher Vac , further
the effects of Tresp can become predominant. For
example, at a velocity of ~40 km/h the difference
of 2 s in reaction time implies a brake distance dif-
ference of ~22 m; whilst at ~60 km/h the difference
is ~33 m. This has implications for cooperative
systems as the reaction time is dominated by the
whole perception process, communications and
human response. A cooperative vehicle solution
should not only optimise the communications and
sensing component but also the driver reaction,
this is a systems problem.
A preliminary analysis indicates that the
driver should be notified, by considering its
reaction times say 2s for any individual and
3s for the elderly. Urban Vehicle speeds are 40
Km/h ±20% and at secondary country roads (the
equivalent of the D-type roads in France) are 70
Km/h ±20%.
When the vehicle is running in an Urban Set-
ting, the minimum Information distance is IUmin
= 28m and the maximum distance is IUmax = 55m.
The former occurs when the vehicle is approach-
ing the intersection at a low speed and assumes a
fast driver reaction time, whilst the latter assumes
that the vehicle is moving at a higher speed and
the driver has a slower reaction time. This limits
the distances at which information has to be
communicated to the driver. A similar analysis
can be made for a rural setting.
The simple analysis has shown that even as-
suming ideal conditions there is a certain degree
of uncertainty with respect to the vehicle response
and stopping times. However, when applications
are safety related, assumptions should be minimal
or non existent.
On an urban setting where road separation
could be as little as, say 20 m, through intelligent
map matching and assuming the availability of
digital maps with an ADAS type accuracy (5m),
it will be possible to discriminate roads travelling
parallel to each other if the location estimation
is under 5m.
The longitudinal error once map matching ex-
ists is a different problem. Latencies should exist
on the transmitted information. If the vehicle is
travelling at 40 km/h and the latency on the re-
ceiver data at another vehicle is 200 ms, then the
travelled distance will be 2.2m. So if the vehicle
needs information at approximately 30 m before
the intersection, the latency effects need to be
taken into account. If the error on the estimation
is 5m longitudinally, the total uncertainty will be
7m. On an urban setting an intersection would
could be as small as 10m, thus if a positive error
of 7m exist, the vehicle might stop not at the entry
point of the intersection but rather in the middle
of it. The application based on these assumptions
can result in a hazard rather than a driving aid.
Localisation has to be resolved locally, that is
to associate relative positions to a local model of
the environment based on the digital road map,
where features are used to improve the vehicle
position as landmarks, a kind of assisted simulta-
neous localisation and map building, as developed
in (Lee K. W. 2007).
Lane Change Manoeuvre
The drawing in Figure 9 shows a lane change
manoeuvre of the SV as it approaches motorway
exit junction. The driver could choose to change
215
 The Localisation Problem in Cooperative Vehicle Applications
Figure 9. Lane change manoeuvres
IV
SV
to a slow lane and then take the exit junction. An
alternative is for the SV to turn left and travel
along the fast lane. Both situations are examined
next with regard to a V2V driving assistance
function.
First Case. The SV wants to take the exit
junction, thus moves from the middle lane to
the slowest lane and then takes the exit junction
(blue trajectory in Figure 9). In the SV electronic
horizon or digital map, the presence of the motor
bicycle (PTW), the IV, the POV, and DVs will
be represented as in the drawing assuming that
there is a perfect estimation of the positions of
the vehicles and the representation of the road
geometry. Assuming that the position of the PTW
could be estimated with an accuracy of 5m and
it is riding close to the middle lane. The width
of each of the lanes will be 4m. The PTW posi-
tion uncertainty is shown in Figure 9, its current
position marked using a green link which was
matched to the second lane. In this case the elec-
tronic horizon will represent wrongly the PTW
and thus authorise the SV vehicle change lanes
with disastrous consequences.
Second Case. The SV wants so move onto
the fast lane, A correct representation of all the
components in the EH horizon will only be pos-
sible if the precision of the location estimates
and digital road map are accurate. A critical
entity for this manoeuvre is the lane position of
the IV vehicle. Again if the position uncertainty
is say 5m, the likelihood for the IV vehicle to be
216
POV
DV
HTv
DV
matched on to the intermediate lane as shown in
Figure 9 will be high. Therefore there might be
the wrong lane association of the IV vehicle in the
digital map and thus for the blind detection/lane
assistant application, it will be possible to give
the go ahead for the lane change manoeuvre; with
unexpected results.
This case study shows the importance of
good vehicle position estimates. In order to have
a good lane discrimination, this needs position
errors under 0.5m if safety related estimates are
needed. The combination of data from GNSS
receivers and proprioceptive sensors alone will
not suffice, as this provides mainly an absolute
position estimation with errors reaching with
much difficulty under 5m in all conditions par-
ticularly in urban areas. A different approach
is needed that measures relative position of the
vehicle and provides sufficient data for a better
discrimination like, for example, lane detection
that will narrow very much lateral positioning
errors and avoid for example the risks associated
to the given examples.
CONCLUS ION and DIS CUSS ION
This chapter has focused on the constraints im-
posed by the needs for precise vehicle localisation
estimates to be able to deploy vehicles with com-
munications capabilities in real traffic conditions.
It has been written from a practical perspective
The Localisation Problem in Cooperative Vehicle Applications
based on the analysis of use cases to infer some
results. These reflect the interest on them by major
stakeholders in the domain.
V2V and V2I safety applications are complex,
and the problem does not limit to communications
capabilities. It involves spatial information with
respect to the vehicles’ locations relative to each
other and the infrastructure, and their representa-
tion on the only source of common data across
the system, feature-rich digital road maps. Time
is a very important parameter, it imposes the
relevance of the information shared, as scenarios
are very dynamic, that is information corresponds
to specific instances of time. Information needs
to be associated in a spatial manner, for example
the location of a vehicle, an obstacle or road works
to a lane or road segment.
The deployment of advanced functions in the
automotive domain is constrained by cost, a major
issue for automotive OEMs. The estimation of a
vehicle position using low-cost components results
on estimates with errors reaching more 10 m in
urban conditions. Applications that require that
the estimated position should be less than one
metre (e.g. need for lane discrimination), implies
that their deployment will be difficult using cur-
rent solutions. Localisation estimates should be
relative to the area being traversed by the vehicle
and/or with respect to the other vehicles. For
this purpose, features on the environment in the
form of landmarks need to be used to increase
the localisation of the subject vehicle. The avail-
ability of beacons like those provided by UWB
communications should also contribute to solve
this problem. Another constraint resides on ac-
curacy of the digital road maps and their constant
update, as all positions are expressed with respect
to these maps. Currently only major roads have
accuracies less or equal to 5m, because it is eco-
nomically difficult to extend them to all roads as
well as to make them more accurate. Therefore
this constraint has to be also included as part of
the design of V2V and V2I applications.
Despite the referred constraints there is a strong
potential on the use of communications technolo-
gies for vehicle safety applications. Constraints
will be overcome through the use of embedded
machine intelligence as well as complementary
exteroceptive and proprioceptive sensors onboard
passenger vehicles, bringing opportunities not
only for safety applications but also to optimise
traffic flow, reduce fuel consumption and pol-
lution.
It is likely that within the next future links
between infrastructure and vehicles should led
to the creation of intelligent spaces, where the
sharing of information amongst all actors should
be widespread. Each actor has the potential of be-
coming a sensor providing sufficient information
to build digital representations of the world to be
used to enhance the drivers’ situational awareness
and lead to the deployment of advanced driving
assistance functions that may even lead to the
autonomous functions.
RE
Ardeshiri, T. et al. (2006). Offset Eliminative
Map Matching Algorithm for Intersection Ac-
tive Safety Applications. Proceedings Intelligent
Vehicles Symposium IV2006, Tokyo.
Arem, B. van, Driel, C. van., & Visser, R. (2006).
The Impact of Cooperative Adaptive Cruise
Control on Traffic-Flow Characteristics. IEEE
Transactions On Intelligent Transportation Sys-
tems, 7(4).
Broughton, J., & Walter, L. (2007). Trends in fatal
car accidents: Analysis of CCIS Data (Published
project report PPR 172). TRL Limiter.
Cook, J. A., Kolmanovsky, I. V., McNamara,
D., & Prasad, K. V. (2007). Control, Comput-
ing, and Communications: Technology for the
twenty-first century Model T. Proceedings of
217
 The Localisation Problem in Cooperative Vehicle Applications

the IEEE Special Issue Advanced Automobile Lee, K.W., Wijesoma, S., & Ibanez-Guzman, J.
Technologies, 95(2). (2007). A constrained SLAM approach to robust
and accurate localisation of autonomous ground
Cui, Y. J., & Ge, S. S. (2002). Autonomous ve-
vehicles. Robotics Autonomous Systems, 55.
hicle positioning with GPS in Urban Canyon
environments, IEEE Transactions in Robotics Madhavan, R., Messina, E. R., & Albus, J. S.
and Automation, 19. (2007). Intelligent Vehicle Systems: A 4D/RCS
Approach. New York: Nova Science Publishers.
CVIS Project Partners. (2008). Retrieved June
2008, from http://www.cvisproject.org Papp, Z., Brown, C. L., & Bartels, C. (2008). World
Modeling for Cooperative Intelligent Vehicles.
Defense Advanced Research Projects Agency
Proceedings 2008 IEEE Intelligent Vehicles
(2008). Retrieved June 2008, from http://www.
Symposium, IV2008, Eindhoven.
darpa.mil/grandchallenge/index.asp
SAFESPOT Project Partners. (2008). Retrieved
Department of Transport. (2008). Retrieved June
June 2008, from http://www.safespot-eu.org
2008, from http://www.itslessons.its.dot.gov/
Schlingelhof, M., Betaille, D., Bonnifait, P., &
Gharavi H., Prasad K. V., & Ioannou P.A. (Eds.),
Demaseure, K. (2008). Advanced positioning
(2007). Special issue Advanced Automobile Tech-
technologies for co-operative systems. IET Intel-
nologies. Proceedings IEEE , 95(2).
ligent Transportation Systems, 2(2).
Gillièron, P. Y., & Konnen, J. (2003). Enhanced
Srini, V. P. (2006, December). A vision for sup-
Navigation System for Road Telematics. In Pro-
porting Autonomous Navigation in Urban Envi-
ceedings the Third Swiss Transport Research
ronments, IEEE Computer.
Conference, (STR 03), Ascona/Ticino.
Toledo-Moreo, R. et al. (2007). High-Integrity
Kelly, A., & Stentz, A. (1998). Rough Terrain
IMM-EKF-Based Road Vehicle Navigation With
Autonomous Mobility - Part 1: A Theoretical
Low-Cost GPS/SBAS/INS. IEEE Transactions on
Analysis of Requirements. Autonomous Robots,
Intelligent Transportation Systems, 8(3).
5.
Wang, H., Xu, J., Ibañez-Guzmán, J, Jarvis,
Laugier, C., Vasquez, D., Yguel M., Fraichard Th.
R., Goh, T., & Chang, C. W. (2000). Real Time
& Aycard O., (2008). Geometric and Bayesian
Obstacle Detection for AGV Navigation Using
Models for Safe Navigation in Dynamic Environ-
Multi-baseline Stereo. In D. Rus & S. Singh
ments. Intelligent Service Robotics 1(1).
(Eds.), International Symposium on Experimental
Lee, K. W. (2007). Autonomous vehicle navigation Robotics Lecture Notes in Control and Informa-
in neighbourhood environments: A map assisted tion Sciences, 271. Springer.
and a feature-based approach. Unpublished
doctoral dissertation, Nanyang Technological
University.

218
 219

Chapter XII
An Overview of Positioning and
Data Fusion Techniques Applied
to Land Vehicle Navigation
Systems
Denis Gingras
Université de Sherbrooke, Canada

ABSTRACT

In this chapter, the authors will review the problem of estimating in real-time the position of a vehicle
for use in land navigation systems. After describing the application context and giving a definition of the
problem, they will look at the mathematical framework and technologies involved to design positioning
systems. The authors will compare the performance of some of the most popular data fusion approaches
and provide some insights on their limitations and capabilities. They will then look at the case of ro-
bustness of the positioning system when one or some of the sensors are faulty and will describe how the
positioning system can be made more robust and adaptive in order to take into account the occurrence
of faulty or degraded sensors. Finally, they will go one step further and explore possible architectures
for collaborative positioning systems, whereas many vehicles are interacting and exchanging data to
improve their own position estimate. The chapter is concluded with some remarks on the future evolu-
tion of the field.

INTRODUCTION are (vehicle position) and we also need to locate

The field of automotive positioning systems has
become a research topic in full rise these last few
years. Generally speaking, positioning informa-
tion in road transportation is of prime importance
for safety reasons. We need to know where we
obstacles and other objects/vehicles in the vicinity
and nearby environment of our own vehicle. So an
approach able to precisely localize our vehicle and
evaluate its environment on the road is required
to increase safety (Drane & Rizos, 1997). Apart
from safety issues, the next generation of vehicles

Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.
 An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems
will allow the driver and passengers to have access
to a broad range of services, such as path plan-
ning, navigation, guidance and tracking, which
are based on information technologies, telecom-
munications and telematics. Future vehicles are
likely to be as well mobile offices, information
centers on wheel, or e-nodes connected to the
web and other networks (Kim, Lee, Choi & al.,
2003). To supply those services and to provide the
adequate information contents to the driver and
passengers, it is required to determine in real-time
the vehicle’s position as accurately and efficiently
as possible. This function is an essential part of an
integrated navigation information system (INIS).
An INIS embedded in a vehicle is basically com-
posed of a geographic information system (GIS),
a database composed of roadmaps, cartographic
and geo-referenced data, a positioning module,
a human-machine interface (HMI) as well as
computing and telecommunication capabilities.
It provides useful functionalities to the driver like
path planning, guidance, digital map and points
of interest directory (Farrell & Barth, 1999). The
guidance module uses a planned trip to indicate
the driver which route to take. To avoid giving
wrong indications and impair driving safety,
the navigation system relies upon a positioning
module to know precisely and continuously the
localization of the vehicle. In this chapter, we will
focus on the positioning module of INISs.
Nowadays, the most popular sensor used for
positioning is a global positioning system (GPS),
also called a GPS receiver. However, GPS alone
based systems suffer from various problems such
as poor data latency, signal multipath errors and
occlusions of satellite signals. To alleviate these
problems and obtain the required performance of
the positioning module, we usually use a com-
bination of heterogeneous and complementary
sensors whose measurements are integrated or
fused in real time (Cannon, Nayak & al., 2001,
Grewal, Weill & Andrews, 2001). Among the other
types of sensors, we have Distance Measurement
Indicators (DMI), such as differential odometers,
220
and 2 or 3 axis inertial measurement units (IMU).
Two or more of these complementary positioning
sensing methods must be integrated together to
achieve the required performance at low cost.
The data integration, which implies the fusion of
noisy signals provided by each sensor, must be
performed in some optimal manner. Basically the
goal of data fusion in positioning systems is 1) to
fill in the time gaps whenever we face a loss of
position estimate and 2) to improve the position
estimate by ensemble averaging and exploiting
information redundancy. Historically, data fusion
for positioning has been applied in robotics (Rou-
meliotis & Bekey, 2000, Durant-Whyte, 1994) as
well as in military, space and avionic applications
(Suresh & Sivan, 2004). Multi-sensor data fusion
system design based on rapid prototyping software
platform and Kalman filtering for land navigation
has also been proposed (Abuhadrous, Nashashibi,
Laurgeau & Chinchole, 2004, Redmill, Martin &
Özgüner, 2000). Hence, with the advent of more
powerful, accessible and cheaper technologies
embedded in vehicles for road transportation, on-
board navigation systems are becoming feasible
and standard in the automotive industry.
The chapter is constructed as follow. After the
present introduction, we will look at the specific
problem of vehicular positioning and localization,
what are the systems and sensors available to
do it and what are the problems to be solved. In
the following section, we will consider the most
common data fusion architectures when multiple
positioning sensors are used. Positioning system
designers usually choose the Kalman filter or a
variation thereof as the data fusion method. We
will then look at the various implementations
possible, such as the extended Kalman filter
(EKF), the unscented Kalman filter (UKF) (St-
Pierre & Gingras, 2005) and particle filters (PFs)
(Arulampalam, Maskell & al., 2002), which are
generalizations or variations of the Kalman filter
used to cope with nonlinearities of the sensors or
with non-gaussianity of the signals. To complete
the survey, we will also present an interesting
An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems
alternative to classical Kalman filtering based
on the use of artificial neural networks (NNs)
(St-Pierre & Gingras, 2004).
After having surveyed those various data
fusion approaches and their respective perfor-
mances, we will then focus our discussion on the
reliability and robustness requirements for a real
time positioning system. In positioning naviga-
tion systems, at any time, any of the sensors can
break down or stop sending information properly,
temporarily or permanently. To ensure a practical
solution for use in guidance and navigation sys-
tems, faulty sensors must be detected and isolated
such that their erroneous data will not corrupt the
global position estimates. As we just mentioned,
Kalman filter is usually popular for data fusion
applications. However, an interesting idea is to
use it for fault detection architecture as well. We
will then evaluate the potential of combining fault
detection and data fusion into a single architecture
to make a robust positioning navigation system
(Guopei & Gingras, 2007).
In the subsequent section, we will then con-
sider the possibility to achieve a distributed col-
laborative architecture. Proliferation of real time
inter-vehicular communications provides new
sources of exploitable positioning data. Vehicles
can, under numerous situations, have GPS satel-
lite shortages but there will always be vehicles
in their vicinity, viewing different GPS satellites
line of sights (LOS), to provide them with useful
navigation information. We will thus look at a
cooperative positioning technique making use
of reliable positions of some vehicles to enhance
positioning estimates of some others (Abdessa-
mie & Gingras, 2008). The approach, based on
a simple idea, exploits inter-vehicle data flow to
extract good position measurements from vehicles
with good GPS satellite LOS, in order to enhance
low positioning accuracy of other vehicles, in the
neighborhood. The integration of such information
is done using geometric data fusion approach.
Finally, we conclude the chapter with a discus-
sion on future trends in the field, what are the chal-
lenges ahead and what can be expected for future
positioning systems. Some general comments are
provided along with proper references.
CONTEXT AND PROBLEM
SATEMENT
Modern automotive navigation applications
require position estimates with a precision of 1
meters or less and at a frequency of 1 Hertz or more.
Basically three types of positioning architecture
can be used to locate a vehicle in its environment
(Laneurit, Blanc & al., 2003)
The first type involves absolute localization,
that is, a global position of the vehicle is given in
the world reference by some exteroceptive sensors
such as a GPS receiver. The performance obtained
often depends on satellites visibility and actual
satellites configuration at the time of measure-
ment. Good satellites configuration are one which
present a sufficient elevation angle, a good signal
to noise ratio and a good geometrical configuration
of the satellites. A differential GPS receiver alone
could in principle estimate the position with the
required performance. However, it is still much
too expensive for automotive applications and
the fixed references are not always available. On
the other hand, a low cost GPS receiver cannot
always achieve the required precision and reli-
ability as it suffers multipath signal propagation
and frequent signal outages, most specifically in
urban areas.
The second type involves relative localiza-
tion, that is, when the position and orientation
of the vehicle are determined from its previous
position. Integrative sensors, like odometers and
inertial systems, are commonly used in relative
positioning. This type of measurement, however,
involves time derivation. We therefore observe a
rapid divergence of the localization when distance
increases due to cumulative errors in the estimate.
This is the main drawback of this family of sen-
sors. Two popular sensors used are the differential
221
 An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems
odometer and the inertial measurement unit (IMU)
(Abbott & Powell, 1999). A differential odometer
measures the distance traveled by a vehicle and
the current azimuth during a sampling period.
An inertial measurement unit (IMU) measures
the vehicle’s inertia characterized by its accelera-
tion and its angular velocity. An IMU measures
the acceleration and the angular velocity along
the axis of a Cartesian coordinate system. With
these two sensors, the position of the vehicle is
reckoned by applying basic kinematics’ equations
and using an initial position obtained from another
information source. By means of integration, the
traveled distance and the azimuth variation can
be computed and therefore a new position can
be obtained with these measurements and the
last known position by dead reckoning. The esti-
mated position will eventually drift from the real
position because of the accumulation of errors.
Indeed, the recursive nature of the positioning
computation with the IMU causes the positioning
error to grow proportionally with time. Typically,
DMIs (odometer/IMU) can cumulate a drift in
the position estimate of the order of 10m to 15m
after ten minutes of integration. A periodic reset
of those relative localization systems is therefore
needed.
The third type of positioning architecture
consists of a hybrid form of localization, that is,
this method combines the two previous ones in
order to cumulate their advantages and to allevi-
ate their drawbacks. The third type, which we
will consider from now on, implies data fusion
from the various sensors. A trivial way to fuse
the data from these two sensors is therefore to
reset periodically the IMU position estimate with
an absolute position estimate from the GPS. A
more complex fusion method than reckon/reset
positioning system described above is required
to improve the precision of the estimation. These
methods fuse continuously the available measure-
ments in some optimal sense, as depicted in figure
1 in a centralized architecture.
222
Figure 1. Centralized data fusion architecture
GPS
IMU Fusion Estimated
Odometer Method Position
Inclinometer
One important issue is data registration and
latency correction. Odometer and INS signals
are received with a nearly constant latency (this
error is usually far less than lms). However, GPS
positioning calculation time and reception delay
may vary largely. Fortunately, it is possible to
calculate those delays using the Pulse Per Second
GPS signal and introduce them in the data fusion
algorithms. When the latency of the GPS data is
low, the position information can be propagated
forward using the velocity data and a constant
vehicle velocity model. This is usually adequate for
slow moving vehicles. Problems occur, however,
when the latency of the GPS data is high and the
vehicle’s dynamic is fast. For example, if the GPS
latency is 0.2s and the vehicle is travelling at 30
m/s (110 km/hr), then the vehicle will move six
meters ahead while storing the IMU data. If the
vehicle is moving at constant speed following
a straight path, then the position estimate can
simply be adjusted forward accordingly. However,
any curve or any deviation from a straight line
will lead to incorrect position estimates. Simple
straight GPS correction is not sufficient in that
case and a more sophisticated latency correction
algorithm is required before applying and fusing
the GPS data with the current inertial state.
The sensors’ measurements are also distorted
by deterministic and random errors. Those sources
of random errors are usually described using
stochastic models in a statistical framework. The
estimate ideally maximizes the a posteriori prob-
ability of the random variable position, resulting
from the mathematical transformation of the
An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems

stochastic processes, which model the sensors’
imperfect measurements. The fusion method then
defines the mathematical transformation. The
architecture of data fusion system for positioning
can be decentralized or centralized. In the central-
ized architecture, all the sensor measurements are
fused by one fusion method only. So it is easy to
compare the performance of two different fusion
methods when the cluster of sensors is the same.
Before looking into the details of data fusion
architectures, we will now look at the different
sensors available for localization.
SENSORS MODELS
In a positioning navigation system, various sen-
sors are used. Generally speaking, a sensor is
a device that responds to or detects a physical
quantity and transmits the resulting signal to a
controller. Positioning sensors can be designed to
detect various parameters (coordinate, distance,
direction, or angular velocity) of the position of
vehicular mechanical systems. Sensors can pro-
vide absolute or relative navigation information.
GPS and magnetic compass provide absolute
position and angular direction (azimuth) respec-
tively, whereas all other sensors provide relative
navigation information using dead reckoning.
Detailed relationships between the position and
typical sensors are shown in Table 1.
GPS Model
The GPS model can be described by equation (1)
(Farrell & Barth, 1999).
ρ̂[t] = ρ[t] + δtrop[t] + δiono[t] + δwhite[t] + δmult[t] (1)
where ρ̂[t] is the pseudo-range measured by the
emulated receiver, ρ[t] is the real pseudo-range,
δtrop[t] is the tropospheric delay, δiono[t] is the iono-
spheric delay, δwhite[t] is the white noise generated
by the receiver’s electronic components and δmult[t]
is the multipath problem.
IMU Model
The 3-axis inertial measurement unit has a single
model for every gyroscope and accelerometer
described by equation (2), (Grewal, Weill &
Andrews, 2001).
dˆi [t ] = di [t ] + si (di [t ] + d j [t ]*sin(δij )
t
+ d k [t ]*sin(δik )) + bi [0] + ƒ wbi [n]
n =1

+ d j [t ]*sin(δij ) + d k [t ]*sin(δik )
Table 1. Relationship of vehicle position and
sensor outputs +ci (di [t ] + d j [t ]*sin(δij ) + d k [t ]*sin(δik )) 2
+ wi [t ],
Sensor Output Relation to position
(2)
Directly output posi-
GPS Rover position
tion coordinates
Outputs can be where dˆi [t ] is the measurement, di[t]is the real
IMU
Accelerations and integrated by an INS value, si is the scale factor error, dj[t] is the real
angular rate to obtain the vehicle
position value relative to the axis j, dk[t] is the real value
relative to the axis k, δij is the misalignment angle
Distance or incre- Position coordinates
Odometer
ment of distance are determined by between the axis i and j, δik is the misalignment
dead reckoning from angle between the axis i and k, bi[0] is the turn-
Inclinometer Inclination the distance and
direction relative to on bias, wbi[n] is the random walk white noise
Magnetic
compass
Azimuth a known location characterizing the bias drift, ci is the non-linear

223
 An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems
scale factor and wi[t] is the additive white noise
component.
The scale factor error, the bias parameters,
the non-linear scale factor and the additive white
noise component have different values depending
on the sensor type.
Differential Odometer Model
DMIs consist of wheel rotation sensors, such as
differential odometers and ABS (anti-breaking
systems) sensors (Bonnifait, Bouron, Meizel, &
Crubille, 2003). Those are the one usually used
in automotive localization systems. The wheel
revolutions are integrated from them to measure
the total distance traveled by the vehicle. Given
time information traveled, a forward velocity is
determined. As opposed to GPS receivers, DMIs
are not subject to signal masking or outages.
However the positioning errors cumulate with
time, as is the case for the IMUs. A differential
odometer is constituted of two sensors measuring
the number of rotation of each wheel situated on
the same axle. The total traveled distance and the
azimuth of the vehicle can be computed with the
equations (3) and (4) respectively.
dˆ[t ] = dˆ[t − 1] +
(1 + vd * v[t ] + sd ) * cd [t ] + r [t ]
+ minimizing the mean square error (Gelb, 1974).
2
(1 + vg * v[t ] + sg ) * cg [t ] + r[t ]
(3)
2
θˆ [t ] = θˆ [t − 1] +
(1 + vg * v[t ] + sg ) * cg [t ] + r [t ]
−
l Many other filter designs can be shown to
(1 + vd * v[t ] + sd ) * cd [t ] + r[t ] (4)
,
l constraints. The recursive sequence involves pre-
where d̂[t] is the total traveled distance at time t,
d̂[t – 1] is the total traveled distance at time t-1,
v[t] is the car’s velocity, vd and vg are the gains
224
characterizing the tire dilatation, sd and sg are
the scale factors for the right and the left wheel
respectively, cd[t] and cg[t] are the numbers of
rotations measured within the interval [t-1,t]
for the right and left wheel respectively, r[t] is a
uniform random variable describing the resolu-
tion error, θ̂[t] is the azimuth estimated at time t,
θ̂[t – 1] is the azimuth estimated at time t-1 and l
is the axle length.
Data fuSIion TECHNIQUES
The Kalman Filter
The most basic data fusion method for localization
is based on the Kalman filter (Kalman, 1960). It
is a set of mathematical equations that provides
an efficient computational means to estimate the
state of a process, in a way that minimizes the
mean square error. This filter is very powerful
in several aspects: it supports estimations of
past, present, and even future states, and it can
do so even when the precise nature of the mod-
eled system is unknown. The Kalman filter is an
optimal linear estimator, which uses the a priori
information on the sensor noise sources, the ve-
hicle dynamic and the kinematics equations to
compute recursively an optimal position, while
The filter is optimal when the process noise and
the measurement noise can be modeled by white
Gaussian process. The recursive implementation
of Kalman filters is well suited to the fusion of
data from different sources at different times in
a statistically optimal manner.
be equivalent to the Kalman filter, given several
diction and update steps. The prediction step used
a dynamics model that describes the relationship
between variables over time. A statistical model of
this dynamic process is also necessary. A predic-
tion is usually done to estimate the variables at the
An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems
time of each measurement, as well as in between
measurements when an estimate is required. The
measurement update combines the historical data
passed through the dynamics model with the new
information in an optimal fashion. The Kalman
filter addresses the general problem of trying
to estimate the state x ≠  R n of a discrete-time
controlled process that is governed by the linear
stochastic difference equation We then have,
xk + 1|k = Φk + 1|k[xk|k] + wk (5)
zk + 1 = Hk + 1|k[xk + 1|k] + vk + 1, (6)
where xk + 1|k is the predicted process state vector,
xk|k is the estimated process state vector, Φk + 1|k is
the discrete state transition matrix from k to k+1,
wk is the process noise vector, vk + 1 is the measure-
ment vector, Hk + 1|k is the observation matrix and
vk + 1 is the measurement noise vector. Supplied
with initial conditions PO and xO, the prediction
equations can be given by
xˆk−+1 = φk xˆk (7)
Pk−+1 = φk Pk φTk + Q k (8)
and the update equations are given by
K k = Pk− H kT (H k Pk− H kT + Rk )
−1
(9)
xˆk = xˆk− + K (zk − H k xˆk− ) (10)
Pk = (I − K k H k )Pk− , (11)
Where x̂ is the vector of estimated states, P is the
covariance of estimated states, Q is the dynam-
ics noise matrix, K is the Kalman gain matrix,
R is the covariance matrix and z is the vector of
observations.
Although an error state Kalman filter is desir-
able in some applications, a state space model will
be preferred here as it most clearly illustrates the
operation of the filter. Rigorous development of
a Kalman filter requires a great deal of work in
understanding the physics and electronics involved
in each sensor to understand what the error sources
will be. A comprehensive model of all of the nec-
essary variables to match as closely as possible
the real world phenomenon is needed. Given the
specifications of the project, a sensitivity analysis
would then be done to decide which variables
may be ignored or lumped together. This often
involves Monte Carlo simulations with several
likely candidate filters and repeated tuning of the
statistics. The filter must be able to operate with
the allowed throughput and processing restric-
tions. Finally, blunder detection, adaptive filter
gains, and practical limits to covariance must be
set to achieve optimum performance.
The Centralized Kalman Filter (CKF) is the
most common filter design implemented in inte-
grated navigation systems. However, it behaves
poorly in the presence of nonlinearities. In the
following sections, we will look into more details
of the use of the various improvements possible
over the basic Kalman filter for an integrated
navigation information system.
The Extended Kalman Filter
Improvement of the basic Kalman filter architec-
ture can be achieved in nonlinear situations with
the extended Kalman filter (EKF) (Anderson B.
& Moore J.,1979, Brown R. G. & Hwang P. C.,
1997). This filter is based upon the principle of
linearizing the state transition matrix and the
observation matrix with Taylor series expansions.
The extended Kalman filter has been very popular
for land navigation system (Abbott & Powell,
1999, Wu, Gao & Wan, 2002, Redmill, Kitajima
& Özgüner, 2001). The equations of a centralized
data fusion architecture based on an extended
Kalman filter for land navigation positioning
system are described in (Abbott & Powell, 1999).
As with the Kalman filter, the extended Kalman
filter predicts the states of the random process us-
ing equation (5). The predicted states are updated
with the measurements in equation (6).
225
 An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems
Box 1.
 aR   cos(Φ y )*cos(Φ p ) − sin(Φ y ) cos(Φ y ) *sin(Φ p )   aN 
… …
… aP ⁄⁄ = … sin(Φ y ) *cos(Φ p ) cos(Φ y ) sin(Φ y ) *sin(Φ p ) ⁄ * …
 aY ⁄ …
…  − sin(Φ p ) 0 cos(Φ p )
In our study presented in (St-Pierre & Gingras,
2005), we have 13 states to describe the random
process. A position-velocity-acceleration model is
used for each component of the position (Brown
& Hwang, 1997). The last four states include the
slope, the pitch, the azimuth and the yaw velocities.
The state transition matrix Φk + 1|k is linear. Only
the observation matrix Hk + 1|k contains nonlinear
equations, the most relevant for horizontal posi-
tioning is described by equation (12). (See Box
1), where aR, aP, aY are the acceleration vector
components along the roll, the pitch and the yaw
axis respectively, Φp and Φy are the Euler angles for
the pitch and the yaw axis respectively, aN, aE, aD
are the acceleration along the north axis, the east
axis and the down axis respectively. The extended
Kalman filter approximates the nonlinear matrix
H based on the Taylor series expanded about the
estimated state vector with:
∂H [ xˆk k ]
H [ xˆk +1 k ] ≈ H [ xˆk k ] + ( xˆk +1 k − xˆk k )
∂xˆk k
(13)
The linear approximation often introduces large
errors in the estimated state vector and can lead
to the divergence of the filter. As expected, the
linearization can lead to poor performance and
divergence of the filter for highly non-linear
problems. In addition, the performance analysis
of the extended Kalman filter presents some
difficulties due to the recurrence of the measure
sequence into the states of the filter (Grewal,
Weill & Andrews, 2001). Finally, implementa-
226
⁄
aE ⁄⁄ (12)
…
⁄ …
  aD ⁄
tion of the extended filter can be quite laborious
depending on the number of states required to
model the system. For all these reasons, a recent
improvement to the EKF, named the “unscented”
Kalman filter (UKF) has been proposed (Julier
& Uhlmann, 1997).
The Unscented Kalman Filter
The UKF approximates the posterior probability
density resulting from the nonlinear transforma-
tion of a random variable instead of approximat-
ing the nonlinear functions with a Taylor series
expansion. The approximation is done by evaluat-
ing the nonlinear function with a minimal set of
carefully chosen sample points using a determin-
istic sampling approach. The posterior mean and
covariance estimated from the sample points are
accurate to the second order for any nonlinearity
(Merwe, Freitas & Wan, 2000). If the priori ran-
dom variable is Gaussian, the posterior mean and
covariance are accurate to the third order for any
nonlinearity (Wan & Merwe, 2000). The first use
of an unscented Kalman filter for land navigation
positioning system is described in (Julier, 1997).
Apart from our work (St-Pierre & Gingras, 2005),
one paper has been recently written on the use of
the unscented Kalman filter as the fusion method
in an integrated navigation information system
(Li & Leung, 2003). An unscented Kalman filter
has also been used for GPS positioning (Mao,
Wada & al., 2003).
The unscented Kalman filter is based on the
unscented transformation, which is a method
for reckoning the statistics of a random variable
 An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems
undergoing a nonlinear transformation. A set of
2*nx+ 1 weighted samples are deterministically
chosen to capture the true mean and variance of
the prior random variable.
nχ = nx + nw + nv (14)
where nx is the number of process states, nw is
the dimension of wk and nv is the dimension of
vk. The unscented Kalman filter approximates the
nonlinear observation matrix by
2*nχ
H [ xˆk +1 k ] ≈ ƒ Wi * H [χix,k +1 k ] + χiv,k +1 (15)
i =0
2*nχ
Where, ˆk +1are
H [ xW i k ƒ
] ≈the weights,
i =0
x
Wi * H [χi ,k +1 k ]are v
+ χthe
i , k +1sigma
2*nχ points describing the prior predicted states and
ƒi =0
Wi * H [χix,k +1 k ] + χiv,k +1 are the sigma points describing the mea-
surement noise.
In order to obtain statistically reliable data on
the performance of both algorithms, one hundred
Monte Carlo simulations have been run for each
sensor fusion method (St-Pierre & Gingras, 2005).
For each sampling time, the estimated positions
from the Monte Carlo simulations form the sam-
pling distribution. There were 26639 measurement
vectors for each Monte Carlo simulation. These
sampling distributions approximate the truth
continuous distributions of the posteriori random
variables describing the estimated positions. The
first moment of each sampling distribution has
been computed and used for the computation of
the performance metrics. The two performance
metrics usually encountered in data fusion systems
analysis are the accuracy/precision of the fusion
and the computational time to perform the fusion.
The accuracy is evaluated by taking the Euclidian
distance between the estimated position and the
true position. The mean and the variance of the
Euclidian distances for the whole simulation are
reckoned. The variance describes the precision
of the fusion method. The horizontal position is
described by the tangential plane located at the
real vehicle position whose coordinates are given
by the latitude and the longitude.
In (St-Pierre & Gingras, 2005) we gave detailed
results showing that the unscented Kalman filter
has slightly better results for horizontal positioning
than the extended Kalman filter. The estimated
position is less biased for the unscented Kalman
filter than for the extended Kalman filter. It shows
also that the unscented Kalman filter is more pre-
cise than the extended Kalman filter. However,
contrary to the claim in (Merwe, Freitas & Wan,
2000), (Wan & Merwe, 2000), we found that the
computational cost of the unscented Kalman
filter is significantly greater, by a factor larger
than 20, than the one of the extended Kalman
filter. The significant execution time difference is
related to the number of times equations (5) and
(6) are evaluated for each fusion algorithm. With
the unscented Kalman filter, these equations are
evaluated 75 times, once for each sigma point.
With the extended Kalman filter, the Taylor series
expansion of these equations is evaluated only
once at each iteration. Furthermore, the Jacobian
of the matrix H used in the Taylor series expansion
is calculated only once because the observation
equations are static. Thus the multiple computa-
tions of equations (5) and (6) by the unscented
Kalman filter at each iteration is responsible for
the larger computational cost.
Surprisingly, we found that the unscented Kal-
man filter is less performing than the extended Kal-
man filter when there is no GPS solution available
(St-Pierre & Gingras, 2005). In that situation, the
acceleration of the vehicle measured by the IMU
is used to estimate the vehicle’s position described
by equation (12). This equation represents the
nonlinear transformation of the estimated states
which are assumed to be Gaussian random vari-
able in order to predict the IMU measurement.
The performance of both filters depends on their
capacity to estimate the mean of the resulting
random variable. An empirical analysis has been
made to evaluate this capacity. In this experiment,
each state has been modeled by a discrete Gaussian
227
 An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems
random variable with 100 realizations distributed
uniformly in the range of possible values with a
99% probability of realization. Each realization
is present a number of times proportional to its
probability of realization in the statistical data
representing the probability function. Thus,
24060 samples modeled each random variable.
The nonlinear function described by equation
(12) is then applied to these random variables and
the means of the resulting random variables are
computed. The same discrete random variables
have been used with the Taylor series expansion
of equations (12) and (13). In the extended Kal-
man filter, the linearization occurs around the
states estimated at the previous iteration. The
linearized equation is applied to the predicted
states at the current time. The linearization error
is directly proportional to the difference between
the estimated states and the predicted states. For
the empirical analysis, the mean variation between
the estimated value and the predicted value ob-
tained with the extended Kalman filter for one
Monte Carlo simulation has been taken. Table 2
shows the percentage variation between the real
mean of the a posteriori probability density and
the estimated mean of the a posteriori probability
density obtained with the Taylor series expansion
and the unscented Kalman filter respectively. As
can be seen, the unscented Kalman filter provides
no significant improvement over the extended
Kalman filter and even brings some degradation
in performance for two acceleration components
(roll and pitch).
The superiority of the unscented Kalman
filter happens only when the variation between
the predicted states and the estimated states is
Table 2. Difference between the real mean and the
estimated mean of the a posteriori density
Estimated state EKF UKF
Roll acceleration 0.0064 % 0.8070 %
Pitch acceleration 0.0218 % 1.2876 %
Yaw acceleration 0.2482 % 0.0754 %
228
important. However, due to the low dynamics of
the vehicle, most of this variation is not important
enough to generate a significant linearization er-
ror for the EKF.
The Particle Filter
More recently, a new class of sequential stochastic
methods (Monte Carlo) for Bayesian filtering,
called particle filters (PF), has been proposed
to solve the multi-sensor fusion problem in land
navigation, tracking and positioning systems
(Caron F, Davy M. & al. in press, Arulampalam
M. S., Maskell S. & al., 2002). Generally speaking,
Kalman filters are special cases of Bayesian filters
suitable to estimate posterior probability distribu-
tions when linearity or gaussianity are assumed.
However, this is not always a valid hypothesis.
The usual standard approach consists of making
model simplifications or crude analytic approxi-
mations to obtain algorithms that can be easily
implemented. However these simplifications bring
undesired behaviour of the systems as we have
seen with the Kalman filter in real situation. With
the recent availability of high-powered comput-
ers, numerical-simulation based approaches can
now be considered and the full complexity of
real problems can be addressed. Although these
integration and/or optimization problems can be
tackled using analytic approximation techniques
(ex. EKF) or deterministic numerical integration/
optimization methods (ex. UKF) these classical
methods are often not sufficiently precise or
robust, or they are simply too complex to imple-
ment. In these situations, Monte Carlo algorithms
become an attractive alternative. These algorithms
are remarkably flexible and extremely powerful
(Doucet A. & Wang X. 2005). The basic idea is to
draw a large number of samples (named particles
here) distributed according to some probability
distributions of interest so as to obtain consistent
simulation based estimates.
As opposed to the Kalman filter based algo-
rithms, those particle filters are more general and
An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems
do not assume Gaussian posterior probability
distributions (Andrieu & Doucet 2002). In general,
those distributions have a very complex shape,
and cannot be calculated in closed-form as it re-
quires integration and optimization of complex
multidimensional functions. They can be arbitrary
multimodal or highly skewed posterior probability
distributions. This kind of distributions may arise
in positioning systems when faced with nonlinear
observations, evolution models or non-Gaussian
noise sources. This occur for instance in the
presence of sensor failures or abrupt changes in
sensors’ functioning conditions. Particle filters
or sequential Monte Carlo methods for Bayesian
filtering provide a numerical approximation of
those posterior and marginal posterior probabil-
ity distributions using a set of weighted random
samples, called particles.
Particle filters can deal with hybrid (both
continuous and discrete). state vectors. In most
cases, the trade-off between estimation accuracy
and computational burden comes down to adjust-
ing the number of particles used. The standard
particle filter algorithm is based on sequential
importance sampling, with weight normaliza-
tion. A detailed mathematical description, of the
standard particle filter algorithm can be found in
(Caron, Davy. & al., 2007) and (Doucet & Wang,
2005). Results and performances reported in the
literature (see references above) so far show that
significant improvements can be achieved in
general dynamic systems, where non Gaussian
posterior distribution and nonlinearities have to
be dealt with (Crisan & Doucet, 2002).
Data Fusion Using Neural Networks
A centralized fusion usually implied the use of
only one algorithm, which fuses all the measure-
ments provided by the sensors. The engineering
labour required to construct a centralized Kalman
or particle filter can be very high. For this reason,
most of the GPS/INS systems have decentralized
filters, which gain in simplicity. However, the
price to pay is a loss of precision compared with
a centralized architecture (Knight, 1997). An
interesting alternative is to use a neural network
(NN) for the data fusion. Indeed, a centralized
NN is no more difficult to realize than a decentral-
ized one. NNs have already been applied to data
fusion related to positioning problem in robotic
with success (Kobayashi & Arai, 1998, Beattie,
1998, Hagan and Menhaj, 1994). The major ad-
vantage of an NN compared to a standard Kalman
filter resides in the fact that it does not need any a
priori statistical and mathematical model to find
a function, which maps optimally the inputs with
the outputs, in our case the absolute position of
the vehicle. The most difficult task however, is to
gather sufficient training sensor data, which can
cover adequately the different manoeuvres to be
encountered by a road vehicle and its dynamic. The
manoeuvres are a function of the road geometry
and the vehicle’s performance. For example, a NN
trained with measured data coming only from a
straight road segment may not be able to give a
good position when the vehicle meets a curve.
As an example, we will consider the archi-
tecture described in more details in (St-Pierre &
Gingras, 2004). We used a feed-forward back-
propagation neural network that was trained with
14939 training data set for 2000 epochs. The
NN has 4 layers composed of 20, 20, 20 and 3
neurons respectively. The corresponding transfer
functions are linear, log-sigmoid, tan-sigmoid and
linear. This architecture was the most promising
of the 12 architectures investigated with various
numbers of layers, transfer functions and number
of neurons trained initially for 100 epochs. Batch
training was preferred over iterative training for
its computation efficiency. The mean square er-
ror gave the training performance at each epoch.
The second order training method was the scaled
conjugate gradient. This method requires only
O(N) operations per epoch compared to others
methods like Gauss-Newton O(N3) and Leven-
berg-Marquardt O(N3) (LeCunn & Bottou, 2001,
Hagan & Menhaj, 1994). When considering the
229
 An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems

Table 3. contains the NN’s inputs Table 4. Mean of the positioning errors by the
Kalman Filter and the NN
Data Sensor
Latitude GPS Position Kalman (m) ANN (m) Improvement
Longitude GPS (%)

Altitude GPS Latitude 2.73 9.34 -242.90

Percent Dilution Of Precision GPS Longitude 12.44 3.56 71.38

(PDOP) Altitude 49.21 0.38 100.78
Linear acceleration INS
Horizontal centripetal ac- INS Table 5. Variance of the positioning errors by the
celeration
Kalman filter and the NN
Vertical centripetal accelera- INS
tion
Position Kalman ANN Improvement
Yaw angular velocity INS (%)
Pitch angular velocity INS Latitude 1073.2 606.1 43.52
Total traveled distance Differential odometer Longitude 994.8 1029.7 -3.51
Azimuth Differential odometer Altitude 12952.0 2.4 99.98
GPS solution availability GPS
Sampling time IMU

of the longitude errors is 5 percent more for the NN

network size, the other training methods were
impractical.
The GPS receiver sampling frequency is usu-
ally lower than those of the inertial measurement
unit or the differential odometer. For synchroniza-
tion purpose, a Boolean entry specifies to the NN
if a GPS solution is available or not. A centralized
Kalman filter has also been realized as a refer-
ence for the comparative evaluation of the NNs
performances. The centralized Kalman filter has
13 states and 10 measurements. The simulation
generated 26430 data samples.
The mean and variance of the positioning er-
ror during the simulation was computed. Table 4
indicates that the positions estimated by the NN
are less biased for the longitude and altitude but
more biased for the latitude than the same posi-
tions estimated by the Kalman filter. The NNs
estimation biases do not exceed 10 meters. So
even if the NN is a biased estimator, it still meets
the required performance.
As shown in Table 5, the variances of the lati-
tude errors and the altitude errors for the NN are
less than those of the Kalman filter. The variance
than for the Kalman filter. In our particular case,
the performance of the NN is generally better than
the performance of the Kalman filter in a mean
square error sense. The important improvement
for the altitude is caused by the NNs ability to
estimate the bias of the GPS altitude.
It can be seen here that NNs can be used as a
centralized fusion method. The results show that
neural networks may be an attractive alternative to
the Kalman filter as a centralized fusion method. In
(St-Pierre & Gingras, 2004), we showed also how
NNs can also be used prior to fusion as nonlinear
pre-processing filters for the land navigation posi-
tioning problem. Similar use of NNs was applied
successfully to telecommunications problems
(Chuah & Sharif, 2001). In our paper, we showed
the NNs capability to successfully learn nonlinear
functions when applied to GPS and differential
odometer measurements. The major difficulty
with an NN is to have access to sufficient ground
truth data for the supervised training. Usually, the
only solution is to have access to some data coming
from a reference, usually given by high precision
and high cost sensors. Some further research on
this topic would include the evaluation of vari-

230
An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems
ous NNs with real sensors and the replacement
of the feed-forward back-propagation NN with
other types of NN such as recurrent NNs. More
exhaustive comparisons with other data fusion
techniques are also required.
ROB AND ADAPTIVE POS ITION
ESIMATE IN PRESENCE OF
SERS FAULTS
So far, we considered mainly the case of continu-
ously operating noisy sensors. But what happen to
the data fusion process and the position estimate
if one of the sensors fails? Either from a user’s
safety point of view or a designer’s perspective,
all automotive navigation systems should be fully
reliable and prevent faults or failures. In all but the
most trivial cases the existence of a fault may lead
to situations with safety, health, environmental,
financial or legal implications. Although good
design practice tries to minimize the occurrence
of faults and failures, it is recognized that such
events do occur. In such cases, faulty sensors
must be detected and the system must be able to
reconfigure itself so as to overcome the deficiency
caused by the fault. In brief, a navigation system
must be robust and adaptive.
Faults can cause the loss of the overall perfor-
mance of a system, which may present hazards to
personnel or lead to unacceptable economic loss.
In order to minimize the impact, fault detection
schemes must be developed. In addition to the
potential use of particle filters described in the
Figure 2. Fault detection architecture
S ens or Outputs
R es idual
G eneration
previous section, several other fruitful research
efforts in the field of fault detection and filter based
adaptive architectures, combining fault detection
and data fusion, have been proposed to improve
the reliability and adaptability of various control
systems (Patton, Frank & Clark, 1989, Pouliezos
& Stavrakakis, 1994, Isermann, 2005). However,
little has been published so far in the area of au-
tomotive navigation systems.
Fault Detection Architecture
A fault is usually defined as an undesired change
in system estimated parameters that degrade par-
tial or overall performance. Fault detection is a
binary decision making process. Either the system
is functioning properly, or there is a fault present.
Generally speaking, fault detection consists of two
processes: residual generation and decision mak-
ing, as shown in Figure 2 (Hashimoto, Kawashima
& al. 2001, Magrabi & Gibbens, 2000).
Residual Generation
Residuals are defined as the resulting differences
between analytically redundant quantities in the
system model. These are similar to innovations
generated by a Kalman filter, which are the dif-
ferences between the measured and estimated
outputs. Under normal conditions, residuals are
small or zero mean; while the occurrence of a
fault causes the residuals to go to non-zero or
unusually large values.
no fault
Decis ion
Making fault
231
 An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems
Decision Making
The decision making process, which acts as an
arbitrator, involves assessing the residuals and
identifying when and where any abnormalities
occur. This is done through threshold testing
both static and dynamic residual behaviors, and
various statistical tests, where the thresholds are
typically based on signal/residual variance.
S classification is that DGPS can effectively remove
So far, although the precision and reliability
of sensors are improved significantly with the
development of the technology, various sensor
faults driven by different situations do exist. In
the following, several faulty scenarios of sensors
are investigated and discussed.
Global Positioning System (GPS)
Faulty Model
A low-cost GPS receiver can output the vehicle
position and driving speed. However, the measure-
ment is likely to be corrupted by time-correlated
noise and the GPS signal is susceptible to jamming.
However, the position and velocity measurements
do not drift over long periods of time. A GPS
faulty model can be based on four particular parts:
typical error budget, environmental interferences,
signal loss, and hardware malfunction.
Table 6. GPS Error sources and their approximate
Deviation [27]
Source of errors Standard deviation (m)
Common
Ionosphere 7.0
Clock and ephemeris 3.6
Troposphere 0.7
Non-common
Receiver noise 0.1~0.7
Multipath 0.1~5.0
232
Typical Error Budget
The main error sources in GPS are listed in Table
6. These errors can be divided into two catego-
ries (Farrell &Barth, 1999, Kronander, 2004):
common and non-common. Common errors are
approximately the same for receivers operating
within a limited geographic region. Non-common
errors are unique to each receiver and depend
on the receiver type and multipath mitigation
technique being used (if any). The point of this
the common errors.
Environmental Interferences
GPS satellite signals, as with any other radio sig-
nals, are subject to some forms of interference and
jamming. It is known that GPS satellite currently
transmit position information in the 1,500-MHz
frequency band with a typical accuracy under 100
meters to anyone in the world who has a simple
receiver costing as little as $100. Any electronic
systems generating radio signals in this frequency
band, main lobe or side lobe, will tend to be a
source of inference to the GPS receiver. With the
popularization of personal radio and Wi-Fi de-
vices, electromagnetic interferences, intentional
or unintentional, are more and more serious. As
an example, the proliferation of ultra-wideband
(UWB) devices intended to be mass-marketed
to the public could cause harmful interference
to GPS.
Signal Loss
GPS is a line-of-sight sensor, and therefore GPS
measurements are subject to signal outages. If it
cannot “see” four satellites, then it will not produce
the expected output. This case is called signal loss.
It may include the following scenarios:
• Urban environments with all buildings (the
so-called urban canyons).
• Inside parking structures.
• In a long tunnel without any relay station.
An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems
Figure 3. GPS faulty model
Real Environmental Typical
world interference error
source
• Under heavy foliage.
• Under bridges.
Hardware Malfunction
A GPS receiver hardware malfunction can be
caused by any abnormality of its components,
such as antenna, amplifier, reference oscillator,
frequency synthesizer, wire disconnection, and
power lost, resulting to no output, or provide an
unstable or incorrect signal. Compared to the
other sources of GPS faults, the probability of a
hardware malfunction of the GPS receiver is rather
low and can be considered negligible. Taking all
the above into account, the GPS faulty model can
be described as in Figure 3.
Inertial Measurement Unit (IMU) Faulty
Model
A low-cost IMU can output the vehicle accelera-
tions and angular rate which can then be integrated
by an Inertial Navigation System (INS) to obtain
the vehicle position, velocity, and attitude. The
advantage of an INS is low sensitivity to high-
frequency noise and external conditions. But the
measurement error of INS will accumulate if it is
not calibrated on-line. Several IMU error models
have been derived (Kim, Lee, Choi & al., 2003),
which are all equivalent in the Kalman filtering
context. Error state vectors consist of navigation
parameters, and accelerometer and gyroscope er-
ror states (Jekeli, 2001). In this model, the errors
Random
noise
Signal Position
loss information
GPS
receiver
modulated by the Earth’s spin rate are neglected
because of short-range nature of our present ap-
plication. In this paper, the error or faulty model
used is presented in figure 4 and the scenarios
driving the IMU to a faulty state are discussed
below:
IMU Error Sources and Faulty Scenarios
For this sensor, we find the following sources:
• Bias due to bearing torques (for momentum
wheel types), drive excitation feedthrough,
electronics offsets and environmental tem-
perature fluctuations. Intuitively, bias is
any nonzero sensor output when the input
is zero.
• Scale factor error, often resulting from aging
or manufacturing tolerances.
• Alignment errors: Most stand-alone IMU
implementations include an initial transient
period for alignment of the gimbals (for gim-
baled systems) or attitude direction cosines
(for strapdown systems) with respect to the
navigation axes. Errors remaining at the end
of this period are the alignment errors. These
include tilts and azimuth reference errors.
Tilt errors introduce acceleration errors
through the miscalculation of gravitational
acceleration, and these propagate primarily
as Schuler oscillations plus a non-zero-mean
position error approximately equal to the tilt
233
 An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems
Figure 4. IMU faulty model diagram
Alignment Scale factor
Real
error error
world
Nonlinearity
error
error in radians times the radius from the
earth center. Initial azimuth errors primarily
rotate the system trajectory about the starting
point, but there are secondary effects due
to coriolis accelerations and excitation of
Schuler oscillations.
• Cross coupling error (non-linearity).
• Quantization error, which is inherent in all
digitized systems.
• Fault due to one or multiple of the moving
parts wear out or jam, or gimbals lock.
The IMU faulty model can be illustrated as
in Figure 4.
Odometer Faulty Model
An odometer is one of the most common devices
used for tracking and to provide relative position-
ing information of vehicles. In the transmission-
based odometer, the distance to be determined is
based on the number of counts for the wheel and
calibration constants which are proportional to the
radius of the tire. Thus any potential trends that
change the radius and the number of counts can
drive the odometer to a faulty output. Hence, in
term of calibrating the odometer, the scale factor
error is the most critical element, as it affects the
total distance traveled and the forward speed of
the vehicle. Hence, adequate modeling for this
error state could be either a random constant or
a first order Gauss-Markov process.
234
Quantization
error
IMU
Tire Radius Change
The detailed sources of tire radius variation are
listed in the following reference (Lezniak, Lewis
& McMillen, 1977). Mainly:
• Tire radius tends to increase as vehicle
velocity increases because of increasing
centrifugal force on the tire.
• Tire radius tends to increase as air pressure
within the tire increases due to increase tire
temperature or other factors.
• Tire radius tends to increase as tread is worn
off during the lifetime of the tire.
Road Situation
These kinds of error sources depend on the road
situation, including:
• Running over objects on the road, slips or
skids involving one or more wheels when
the vehicle accelerates or decelerates too
rapidly or travels on a snowy, icy, or wet
road.
• In sharp turns, the contact point between
each wheel and the road can change, so that
the actual distance between the left and right
wheels will be different from the one used
to derive the heading.
Gears Tooth Lost
An odometer can operate by counting the pass of
teeth or tabs of the ferrous wheel mounted on the
An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems
rotating shaft of the vehicle. If one or more teeth
are lost, then the value will abate n/t, where n is
the number of the lost teeth, t is the total teeth of
the wheel. In the real situation, the occurrences
of losing three or more teeth are so puny that they
can be omitted.
In brief, the wear out of the tire, the pressure of
the tire, the velocity of the vehicle, the slippage of
the tire, and the gear teeth lost will all contribute
to the imperfect operation of the odometer.
Inclinometer Faulty Model
The error sources of the inclinometer may con-
sist of:
• Error caused by thermal expansion or tem-
perature changes. A normally distribution
band-limited white noise is used to demon-
strate the thermal noise.
• Drift, calibration error or quantization error
due to analog to digital converter resolu-
tion.
• Electromagnetic interference (the major
component of the inclinometer faulty model).
It can be a uniform or a Gaussian distribution,
or a combination of both. The variance and
amplitude depend on the traveling environ-
ment.
• Power lost or hardware malfunction: A per-
manent fault, but since it is only in a very low
possibility, it is omitted in this simulation.
Magnetic Compass Faulty Model (Fluxgate
Compass)
A magnetic compass is an inexpensive absolute
direction sensor. The main drawback with this
device is that the quantity measured, i.e. the
intensity and direction of the magnetic field, can
be distorted in the presence of metals and other
electrical or magnetic fields, such as power lines,
transformers and cars’ powertrain systems.
Compass operations include the following
error sources (Zhao, 1997):
• Hilly road error: When the vehicle is
traveling over a hilly road, the compass plan
will not be parallel to the plane of the Earth
surface. The compass measures only the
projection of the vector components. This
is a short-term magnetic anomaly.
• Random noise error: a) In the situation of
traveling nearby power lines, big trucks, steel
structures (such as freeway underpasses and
tunnels), reinforced concrete buildings, or
bridges (short-term magnetic anomalies); b)
In an environment of electrical or magnetic
noise, or magnetization of the vehicle body
(long-term magnetic anomalies).
• Calibration error: Misalignment of the
compass with respect to the vehicle frame
simply results in a constant error. This type of
error can also be attributed to an inaccurate
estimation of the current declination.
• Permanent fault: power lost or interface
cable disconnected (very low possibility,
they are being omitted in the simulation).
Adaptive and Robust Data Fusion
Architecture
In positioning navigation systems, high precision
and reliability with low cost are always pursued.
Actually, for road navigation, the benefits of the
information obtained by the fusion process make
it possible to use multiple less powerful, lower
cost sensors to achieve as good a performance
as those much more expensive ones.
Kalman filter and its derivatives, the most
popular data fusion methods, have been used
extensively in autonomous or assisted navigation
system for several years. But almost all of these
applications are based on the assumption that all
sensed data are complete and reliable. If anyone
or more sensors are faulty, then the fusion filters
235
 An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems
Figure 5. Adaptive sensor fusion system
Fault
Gaussian Detection
noise Process
Fusion
Sensors
measurements Process
will tend to malfunction as well. In order to en-
sure a reliable positioning estimate in the case of
faulty sensors, an adaptive approach is proposed
as in Figure 5.
As most of the navigation systems, the objec-
tive of the above system is to make the position
estimate of the vehicle as accurate and reliable
as possible. Sensors (GPS, IMU, odometer, in-
clinometer and compass) data which are used to
compute the position and attitude of the vehicle,
often involve sources of uncertainties. Meanwhile,
a state space model can be constructed from the
vehicle dynamic to perform the function of sen-
sor fusion. Both their outputs (measurements and
estimates) can be combined together through a
particular function so as to generate a residual
signal. Passing this signal through a detection
process, a decision is made: either the system is
running properly or there is a fault occurring,
which leads to the fusion process rerunning to
optimize the position estimates.
From the diagram above, we can see how
a residual signal generator and a fault detector
are embedded into the conventional data fusion
architecture. A Kalman filter approach is used,
since it is simpler and more effective to attain the
residual signal via state estimation. The key idea
is to reconstruct the outputs of the process with
the aid of a Kalman filter and to use the estimation
error, or some particular statistical functions of
them to assess the residual signal.
236
residuals Residual
Generator
output
estimates
Then the faultiness can be detected by con-
sidering the residual properties against some
threshold. If any information from the sensors
is detected to be faulty (residual properties go
over a given threshold), then the corresponding
measurements are discarded. In that case, the
position estimate needs to be re-updated. In this
approach, fault detection and data fusion are
combined into a single Kalman architecture to
construct a fault tolerance, robust and adaptable
vehicle positioning system.
The detailed flow chart is shown in Figure 6.
Note that the shaded blocks are part of the fault
detection process. Detailed tests results presented
in (Guopei & Gingras, 2007) show that the pro-
posed Kalman filter based state estimation scheme
ensures that the position estimate is always optimal
and brings significant benefit to the data fusion
system compare with the conventional fusion
architecture without fault detection, in particular
for the frequent GPS signal loss case. Performance
analysis and more details on this architecture can
be found in (Guopei & Gingras, 2007).
COLLABORATIVE DATA FUSION
S
So far we have considered the case of a single
vehicle only. In real life however, cars are not
alone on the road. We may therefore ask ourselves
An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems

Figure 6. Adaptive sensor fusion architecture flow chart

Initialize x̂k− and Pk−

Compute the Kalman gain

K k = Pk− H kT ( H k Pk− H kT + Rk ) −1

Update estimate with measurement zk

xˆk = xˆk− + K(zk − Hk xˆk− )

Residual generation
z k − H k xˆ k
r=
z k2 + ( z k − H k xˆ k )
2

Decision making Y Discard ith

measurement
ri > Threshold ?
Ki = 0

Update the error covariance

Pk = (I − K k H k )Pk−

Project the error covariance ahead

Pk−+1 = φ k Pk φ kT + Qk

Project the state ahead

xˆ k−+1 = φ k xˆ k

Next iteration

the following questions: can we take advantage
of the other vehicles proximity and positioning
data to improve our own position and, if yes, how?
The intelligent vehicle systems (IVS) envisioned
here would be able to communicate their position
and navigation information through inter-vehicle
communications (IVC). IVS will be evolving in
mobile ad hoc networks called (MANET) that
give access to valuable real time data, especially
high precision positioning information. However,
in terms of global navigation systems (GNS),
individual vehicles in a given MANET would
not have access to the same constellation of satel-
lites. And certainly vehicles with good lines of

237
 An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems

Figure 7. Collaborative positioning architecture: In this simplified example, the vehicle at far right and
at far left have a good line of sight (LOS) and can share and provide their good GPS position estimate
to the vehicle in the center who has a bad LOS and therefore a poor GPS position estimate. All vehicles
are assumed equipped with GPS receivers

sight (LOS) have more precise position estimates
than vehicles with a poor LOS. Moreover, some
vehicles may possess high precision DGPS or
beacon-based positioning information to share
in the network.
In this section, we briefly present a collabora-
tive positioning concept that uses some of the
above IVC features, along with additional range
measurement capabilities, to ameliorate position-
ing estimates of neighboring vehicles in a MANET
(Abdessamie & Gingras, 2008). Two vehicles at
different locations can have different sets of visible
satellites and by mean of inter-vehicle commu-
nications, the satellite information can be shared
between vehicles (Berefelt, Boberg & al., 2004).
Many research activities are being conducted in
IVS collaborations. We ref to (Nirupama, 2005) for
a good introduction to IVS research and develop-
ment. For a little more technical reference on the
subject and still worth reading, see (Jurgen, 1998).
Designing a collaborative system for accurate
estimation of relative positions of neighboring
vehicles based on real-time exchange of individual
GPS coordinates, while using vehicle-to-vehicle
radio communications is a challenging task (Kuk-
shya, Krishnan & Kellum, 2005). Research in the
domain of collaborative navigation was usually
aimed for collaborative driving systems (CDS),
where car platoons or mobile robots are involved
(Gaubert, Beauregard & al, 2003, Crawford &
Cannon, 2005). However, most CDS previously
reported require either costly vehicle-to-vehicle
relative dependence or vehicle-to-infrastructure
dependence.

238
An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems

Figure 8. Principle of geometric data fusion approach in collaborative positioning of vehicles using
GPS as shown in Figure 7. The far right and far left vehicles are sharing their good GPS position es-
timates using appropriate fusion in order to reduce the uncertainty ellipse of the vehicle in the center
of the platoon

In a recent paper (Abdessamie & Gingras,
2008) we have proposed a less expensive navi-
gation approach, exploiting existing positioning
resources embedded in vehicles and independent
of road infrastructures. To illustrate the concept
here, we will treat only the case of three vehicles,
as shown in figure 7, which, although limited, is
sufficient to understand the underlying principle
in a more general case. In order to consider the
design of collaborative localization architectures,
such as the one we present in (Abdessamie & Gin-
gras, 2008), we need to consider a set of various
assumptions in order to simplify the model. In
our case, we supposed the following conditions
to be valid:
• All vehicles in the MANET are equipped
with necessary navigation items (GPS, DR,
IVC sensors…etc.)
• All vehicles have range measurement radars,
to provide precise inter-vehicle distance.
Millimeter wave radars MMW for automo-
tive, studied in [3], would be appropriate
solution to our application for they have a
high Doppler sensitivity and a 200 meters
range with a good precision which is very
suitable for our case.
• No vehicle dynamics are considered in
the CPA, we instead used a simple motion
model.
• Error covariance matrix on position, head-
ing, and inter-vehicle distance contains the
global errors of vehicle systems.
• No special vehicle frame is considered
• Coordinate reference system is geodetic
altitude, latitude and azimuth.
• No altitude difference is considered, this
corresponds to the case where all three ve-
hicles are located on a relatively flat plane
with a constant altitude.
• Inter-vehicle communications are real time
and safe.

239
 An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems
Collaborative Uncertainty
Minimization
In a distributive, collaborative approach, each
vehicle has its own set of position estimates as
well as other information such as range informa-
tion from other neighboring vehicles.
To achieve a collective improvement of position
estimates, we have considered a geometric data
fusion approach (Abdessamie & Gingras, 2008).
This method is based on the geometric analysis of
the sensing uncertainty and is motivated by the
geometric idea that the volume of the uncertainty
ellipsoid should be minimized, as illustrated in
figure 8. The resulting fusion equations coincide
with those obtained by Bayesian inference, Kal-
man filter theory, and by weighted least-squares
estimation (Abidi & Gonzales, 1992). When
systematic error or bias is made small, the uncer-
tainty ellipsoid encloses a region in space where
the true value most likely exists. The center of the
ellipsoid is the mean of the measurement and the
ellipsoid boundary represents a distance of one
standard deviation from the mean.
Geometric data fusion has been used in many
research applications (Elliott, Langlois & Croft,
2001, Lee & Ha, 2001). It had proven to be a
powerful uncertainty management data fusion
technique. In (Abdessamie & Gingras, 2008), we
showed that a collaborative approach can be used
in navigation systems and further improve the
position estimate over a conventional data fusion
approach. More mathematical details and perfor-
mance analysis can be found in the paper.
Concluding rmark and
futurtr nd
In this paper, we have surveyed the positioning
estimation problem applied to land navigation
systems and reviewed some of the various sen-
sor fusion techniques usually encountered in
such systems. We have discussed their relative
240
performance and limitations. The extended
Kalman filter (EKF) in data fusion centralized
architectures remains a design of choice for most
applications. We have explained how to make such
systems more robust by detecting and identifying
sensor faults. Finally we looked at the possibility
to exploit the presence of several vehicles in the
vicinity, in order to improve one’s own position
estimate using a collaborative and geometric data
fusion approach.
A main trend can be seen in the field of dense
sensor networks, where multilateration can render
good location accuracy despite significant errors in
range estimates between sensors. It can therefore
be useful for vehicle navigation. This has launched
a research area known as sensor location using
wireless telecommunication techniques, which
seeks to process potentially enormous quantities
of data collectively to achieve optimal positioning
results. Some papers (Biswas & Ye 2004, High-
tower & Boriello, 2001) provide good surveys of
available techniques for sensor location. Usually,
the techniques achieving the best performance
process the input data in a centralized fashion;
however distributed versions are also available.
Savvides, Park & Srivastava (2002) solve a global
nonlinear optimization problem for localizing
sensors through Kalman filtering, which yields
good results, but requires a high number of anchor
nodes. The multi-dimensional scaling technique
employed by (Shang, Rumi & al., 2003) for sensor
location achieves good results with few anchor
nodes, but requires high radio connectivity. In
(Savarese, Rabaey & Beutel, 2001) good results
are presented, based on an architecture employing
a two-phase initiation and refinement approach,
and, as opposed to the aforementioned papers,
deals with relatively high levels of noise.
According to results previously reported, the
algorithms achieving the best performance in
sensor location techniques use quadratic con-
straints to minimize a linear objective function
(Biswas & Ye, 2004, Doherty, Pister & Ghaoui,
2001, Hightower & Boriello, 2001). Since some of
An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems
the constraints are non-convex, the papers above
differ primarily in their relaxation approaches to
render the problem convex. According to those
publications, the solution provided by Biswas et
al. (2004) has greater applicability and yields bet-
ter results than the one by Doherty & al. (2001).
The paper by Gentile, (2005) follows their same
centralized approach, maintaining the efficiency
of convex optimization. However by applying
linear triangle inequality constraints as opposed
to quadratic ones, the problem is automatically
convex. As a result, a tighter solution is guaranteed
since no relaxation of the constraints is needed.
With the current evolution of automotive
technologies, embedded positioning systems
will become more and more easily available at
a lower cost, thus allowing all vehicles to be
equipped with such technologies. In addition, all
vehicles are becoming networked and equipped
with wireless communication capabilities, thus
allowing the use of distributed and collabora-
tive techniques for navigation and positioning.
Wireless communications networks are becom-
ing attractive to localize vehicles using various
radio-based range technologies such as received
signal strength indicators (RSSI), power signal
attenuation or time-of-arrival (TOA) techniques
(Parker & Shahrokh, 2006, Kukshya, Krishnan
& Kellum, 2005, Chan, Tsui, So & Ching, 2006,
Tsukamoto, Fujii, Itami. & Itoh, 2001, Qi, Ko-
bayashi & Suda, 2006). Originally, most existing
radio-based ranging technologies for position es-
timation has been applied to fixed wireless sensor
networks, because it is not possible, for practical
and economic reasons, to equip all sensor nodes
in a network with GPS receivers. Previously
proposed radio-based ranging approaches have
been either distributed or centralized, providing
a more or less precise location estimate of the
nodes, depending on the scheme used (also called
course grain and fine grain schemes). Course grain
localization, usually configured in a centralized
fashion, requires less radio connectivity between
nodes but needs a central processing node in the
network to establish localization of the remaining
nodes. It works fine only if a rough estimate of
a node’s position is sufficient (Doherty, Ghaoui,
& Pister, 2001). Conversely, fine grain localiza-
tion requires more resources and radio traffic,
but provides a relatively accurate node position
estimates. Fine grain localization may operate in
either a distributed or centralized approach. The
main challenges in using these techniques for
vehicular applications, originally developed for
fixed sensor networks, are related to the continu-
ous high mobility of the nodes.
In a more theoretical context, extensions of
the Bayesian framework, which copes with the
different types of imperfections of the various
sensors and sources of information, such as
imprecision, uncertainty, incompleteness, and
ignorance, have also been proposed for the data
fusion problem in vehicle localization. For ex-
ample, Fisher’s fiducial probability, fuzzy logic,
credal inference, Demster’s belief functions and
transferable belief model (TBM) are being used
and described in various papers, such as (Caron
& Smets, 2005, Caron, Duflos & al., 2006, Smets
& Ristic, 2004) to name a few.
Finally, in a more global context, the advent of
new satellite networks for navigation in the years
to come, such as GALILEO in Europe, are likely to
boost vehicle localization worldwide capabilities
and provide more flexibility in the design of future
positioning systems for automotive applications
(Vejrazka, 2007). Clearly, more robust, networked
and distributed intelligent positioning systems are
likely to be encountered in future positioning and
navigation systems.
AcknowlEDdgmnt
I would like to thank the Canadian NCE AUTO21
for funding this research project, the University
of Calgary for providing the acceleration data,
Geomatic Canada for the topographical informa-
tion and E. A. Wan, A. D. R. Merwe and OGI
241
 An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems
School of Sciences for the Rebel Toolkit. I would
like also to thank my former students Mathieu
St-Pierre, Guopei Liu and Rabah Abdessamie for
their original contributions in this area.
REFERENCESnc
Abbott, E., & Powell, D. (1999). Land-Vehicle
Navigation Using GPS. IEEE Proceedings of the
Institute of Electrical and Electronics Engineer,
1(87), 145-162.
Abuhadrous, I., Nashashibi, F., Laurgeau, C., &
Chinchole, M. (2004, June). Multi-Sensor Data
Fusion for Land Vehicle Localization Using MAPS
TM
. Paper presented at the IEEE Intelligent Vehicle
Conference, Columbus, Ohio, USA.
Abdessamie, R., & Gingras, D. (2008). A col-
laborative navigation approach in intelligent
vehicles. Paper presented at the Society of Au-
tomotive Engineering (SAE) Conference, paper
no 1249 Detroit USA.
Abidi, M. A., & Gonzalez, R. C. (1992). Data
Fusion In Robotics And Machine Intelligence.
NY, USA: Academic Press.
Anderson, B., & Moore, J. (1979). Optimal Filter-
ing. Englewood, NJ: Prentice Hall.
Andrieu, C., & Doucet, A. (2002). Particle filter-
ing for partially observed Gaussian state space
models. Journal Royal Statistical Society B.,
4(64), 827–836.
Arulampalam, S., Maskell, S., Gordon N., &
Clapp T. (2002). A Tutorial on Particle Filters
for Online Nonlinear/Non-Gaussian Bayesian
Tracking. IEEE Transactions on Signal process-
ing, 50(2), 174-188.
Beattie, J. B. (1998). Self-localisation in the senario
autonomous wheelchair. Journal of Intelligent and
Robotic Systems, 22, 255-267.
242
Berefelt, F., Boberg, B., Nygards, J., Stromback,
P., & Wirkander, S.-L. (2004). Collaborative GPS/
INS Navigation in Urban Environment. Swedish
Defence Research Agency ION-NTM FOI, (pp.
1114-1125). Sweden.
Biswas, P., & Ye Y. (2004, April). Semidefinite
Programming for Ad Hoc Wireless Sensor Net-
work Localization. Presented at the IPSN-04
Conference on Information Processing in Sensor
Networks, Berkeley, CA, USA, (pp. 46-54).
Bonnifait, P., Bouron, P., Meizel, D., & Crubille, P.
(2003). Dynamic Localization of Car-like Vehicles
using Data Fusion of Redundant ABS Sensors. The
Journal of Navigation, 3(56), 429–441. UK.
Brown, R. G., & Hwang, P. C. (1997). Introduction
to Random Signals and Applied Kalman Filtering.
New York, NY, USA: John Wiley and Sons.
Caron, F., Smets, P., Duflos, E., & Vanheeghe, P.
(2005, July). Multisensor data fusion in the frame
of the TBM on reals. Application to land vehicle
positioning. Paper presented at the International
Conference on Information Fusion (FUSION’05),
Philadelphia, PA, USA.
Caron, F., Duflos, E., Pomorski, D., & Vanheeghe,
P. (2006). GPS/IMU Data Fusion using Multisen-
sor Kalman Filtering: Introduction of Contextual
Aspects. Information Fusion, 7(2), 221-230. Hol-
land: Elsevier.
Caron, F., Davy, M., Duflos, E., & Vanheeghe,
P. (2007). Particle Filtering for Multisensor Data
Fusion with Switching Observation Models:
Application to Land Vehicle Positioning. IEEE
Transactions on Signal Processing, 6(55), 2703-
2719.
Cannon, M. E., Nayak, R., Lachapelle, G., Saly-
chev, O. S., &Voronov, V. V. (2001) Low-cost
INS/GPS Integration: Concepts and Testings,
Journal of Navigation, 54(1), 119-134.
Chan, Y-T, Tsui, W-Y., So, H-C. & Ching, P-C.
(2006), Time-of-Arrival Based Localization
An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems
Under NLOS Conditions. IEEE Transactions on
Vehicular Technology, 1(55), 17-24.
Chuah, T. C., & Sharif, B. S. (2001). Robust adap-
tive spread-spectrum receiver with neural-net
preprocessing in non-gaussian noise. IEEE Trans-
action on Neural Networks, 3(12), 546-558.
Crawford, S., & Cannon, M. E. (2005). Perfor-
mance Evaluation of Sensor Combinations for
Mobile Platoon Control; Positioning, Location,
and Navigation (PLAN) Group, (Tech. Rep.) Uni-
versity of Calgary, Calgary Alberta, Canada
Crisan, D., & Doucet, A. (2002). A survey of
theoretical results on particle filtering for practi-
tioners,” IEEE Transaction on. Signal Processing,
3(50), 736–746.
Doherty, L., Pister, K. S. J., & El Ghaoui, L. (2001,
April). Convex Position Estimation in Wireless
Sensor Networks. Paper presented at the INFO-
COM 2001. 20TH Annual Joint Conference of the
IEEE Computer and Communications Societies.
Proceedings (pp. 1655-1663).
Doherty, L., Ghaoui, L. E., & Pister, K. (2001).
Convex position estimation in wireless sensor
networks. Paper presented at the INFOCOMM
Conference., (3), 1655-1663. Anchorage, Alaska,
USA
Doucet, A., & Wang, X. (2005). Monte Carlo
methods for signal processing: A review in the
statistical signal processing context. IEEE Signal
Processing Magazine, 6(22), 152-170.
Drane, C., & Rizos, C. (1997). Positionng systems
in intelligent transportation systems. Artech
House, Boston, Ma., USA.
Durant-Whyte, H. (1994). Where am I? Industrial
Robot, 21(2), 11-16. MCB University Press
Elliott, J. D., Langlois, D., & Croft, E. A. (2001,
June). Sensor Uncertainty Management for an
Encapsulated Logical Device Architecture: Part
I. American Control Conference, Arlington, USA,
Proceedings (pp. 4282-4387).
Farrell, J., & Barth, M. (1999). The Global Po-
sitioning System and Inertial Navigation. NY,
USA: McGraw-Hill.
Gaubert, N., Beauregard, M., Michaud, F., & de
Lafontaine, J. (2003, October). Emulation of Col-
laborative Driving Systems Using Mobile Robots.
Presented at the IEEE International Conference
on Systems, Man and Cybernetics, 1, 856-861.
Washington DC, USA.
Gelb, A. (1974). Applied Optimal Estimation.
Cambridge, MA: The MIT Press.
Gentile, C. (2005, May). Sensor Location through
Linear Programming with Triangle Inequality
Constraints. Paper presented at the ICC-IEEE
International Conference on Communications,
Proceedings, 5, 3192-3196.
Grewal, M. S., Weill, L. R., & Andrews, A. P.
(2001). Global Positioning Systems, Inertial
Navigation, and Integration. New York, NY,
USA: John Wiley & Sons.
Guopei, L., & Gingras, D. (2007, April). Adaptive
and Reconfigurable Data Fusion Architectures
in Vehicle Positioning Navigation Systems.
Paper presented at the Society of Automotive
Engineering (SAE) conference. Paper no. 1108,
Detroit, USA.
Hagan, M. T., & Menhaj, M. (1994). Training
feedforward networks with the Marquardt algo-
rithm. IEEE Transactions on Neural Networks,
6(5), 989-993.
Hashimoto, M., Kawashima, H., Nakagami, T.,
& Oba F. (2001, October). Sensor Fault detection
and Identification in Dead-Reckoning System
of Mobile Robot: Interacting Multiple Model
Approach. Paper presented at the IEEE/RSJ In-
ternational Conference on Intelligent Robot and
Systems, Mani, Hawaii, USA.
243
 An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems

Hightower, J., & Borriello, G. (2001, August).
Location Systems for Ubiquitous Computing.
IEEE Computer, 8(34), 57-66.
Isermann, R. (2005). Fault-Diagnosis Systems:
An Introduction-From Fault Detection To Fault
Tolerance. Springer Verlag, NY, USA.
Jekeli, C. (2001). Inertial Navigation Systems with
Geodetic Applications, De Gruyter, (pp. 308-318).
New York, USA
Julier, S. (1997). Process Models for the Navi-
gation of High-Speed Land Vehicles. Wadham
College, Oxford, UK (p. 184).
Julier, S. J., & Uhlmann, J. K. (1997). A New Ex-
tension of the Kalman Filter to Nonlinear Systems.
Paper presented at the 11th International Sympo-
sium on Aerospace/Defense Sensing, Simulation
and Controls. Orlando, FL, USA
Jurgen, R. K. (1998). Navigation and Intelligent
Transportation Systems. Progress In Technology,
Automotive Electronics Series. SAE International,
Detroit, USA.
Kalman, R. E. (1960). A new approach to linear
filtering and prediction problems. Transactions of
the American Society of Mechanical Engineering
(ASME). (82) (Series D), (pp. 35-45).
Kim, S.-B., Lee, S.-Y., Choi J., & et al. (2003).
Multi-Sensor data fusion for future telematics
applications. Journal Astronomical Space Sci-
ences, 20(4), 359–364.
Knight, D. T. (1997). Rapid development of
tightly-coupled GPS/INS systems. IEEE Aero-
space and Electronic Systems (AES) Magazine,
12(2), 14-18.
Kobayashi, F., & Arai, F. (1998). Sensor fusion
system using recurrent fuzzy inference. Journal of
Intelligent and Robotic Systems, (23), 201-216.
Kronander, J. (2004, December). Robust Automo-
tive Positioning: Integration of GPS and Relative
Motion Sensors. PhD Dissertation, Linköping,
Sweden.
Kukshya, V., Krishnan, H., & Kellum, C. (2005,
June). Design of a System Solution for Relative
Positioning of Vehicles Using Vehicle-to-Vehicle
Radio Communications During GPS Outages.
Paper presented at the IEEE Vehicular technology
Conference, 2, 1313-1317. Dallas, Tx, USA.
Laneurit, J., Blanc, C., Chapuis, R., & Trassou-
daine L. (2003, June). Multisensorial data fusion
for global vehicle and obstacles absolute position-
ing. Paper presented at the IEEE Intelligent Vehicle
Conference, San Francisco, USA.
LeCunn, Y., & Bottou, L. (2001). Gradient-Based
Learning Applied to Document Recognition. IEEE
Press, NY, USA.
Lee, J., & Ha, I. (2001). Real-Time Motion Capture
for a Human Body using Accelerometers. Ro-
botica, 19, 601–610. Cambridge, UK: Cambridge
University Press.
Lezniak, T. W., Lewis, R. W., & McMillen, R.
A. (1977). A Dead Reckoning/Map Correlation
System for Automatic Vehicle Tracking. IEEE
Transactions on Vehicular Technology, 1(VT-
26), 47-60.
Li, W., & Leung, H. (2003). Constrained Unscented
Kalman Filter Based Fusion of GPS/INS/Digital
Map for Vehicle Localization. Paper presented at
the IEEE International Conference on Intelligent
Transportation Systems, Shangai, China.
Magrabi, S. M. & Gibbens, P. W. (2000, June).
Decentralized Fault Detection and Diagnosis
in Navigation Systems for Unmanned Aerial
Vehicles. Paper presented at the IEEE Posi-
tion Location and Navigation Symposium, San
Diego,USA.
Mao, X., Wada, M., Hashimoto, H., Saito, M.
& Suzuki, I. (2003, November). EM/Unscented
Smoothing Based Parameter Learning for Nonlin-

244
An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems
ear Models for GPS Positioning. Paper presented
at the 10th World Congress on Intelligent Transport
Systems, Madrid, Spain.
Merwe, A. D. R., Freitas, N., & Wan, E. (2000,
August). The Unscented Particle Filter, Tech
report CUED/F-INFENG/TR-380. Cambridge
University Engineering Department, Cambridge,
England.
Nirupama, B. (2005). Wireless Sensor Networks,
A Systems Perspective, 2005 Artech House,
Boston, USA
Parker R., & Shahrokh V. (2006, September). Ve-
hicle Localization in Vehicular Networks. Paper
presented at the IEEE 64th Vehicular Technology
Conference VTC-2006, Melbourne, Australia.
Patton, R., Frank, P., & Clark, R. (1989). Fault
Diagnosis in Dynamic System: Theory and Ap-
plications. Prentice Hall, USA.
Pouliezos, A. D., & Stavrakakis, G. S. (1994). Real
Time Fault Monitoring of Industrial Processes.
NY, USA: Kluwer Academic Publishers.
Qi, Y., Kobayashi, H., & Suda, H. (2006). On
Time-of-arrival Positioning in a Multipath En-
vironment. IEEE Transactions on Vehicular
Technology, 5(55), 1516-1526.
Redmill, K. A., Martin, J. I., & Özgüner, U. (2000,
October). Sensor and data fusion design and
evaluation with a virtual environment simulator.
Paper presented at the IEEE Intelligent Vehicles
Symposium. Dearborn, Mi, USA, Proceedings
(pp. 668-674).
Redmill, K. A., Kitajima, T., & Özgüner, U.
(2001, June). DGPS/INS Integrated Positioning
for Control of Automated Vehicle. Paper presented
at the IEEE Intelligent Transportation Systems,
Oakland, USA.
Roumeliotis, S., & Bekey, G. (2000, April). Bayes-
ian estimation and Kalman filtering: A unified
framework for mobile robot localization. Paper
presented at the IEEE International Conference
on Robotics & Automation, San Francisco, CA,
USA.
Savarese, C., Rabaey, J. M., & Beutel, J. (2001,
May). Location in Distributed Ad-Hoc Wireless
Networks. Paper presented at the IEEE Confer-
ence on Acoustics, Speech, and Signal Process-
ing, Salt Lake City, Utah, USA, Proceedings (pp.
2037-2040).
Savvides, A., Park, H., & Srivastava, M. B. (2002,
September). The Bits and Flops of the N-hop
Multilateration Primitive For Node Localization
Problems. Paper presented at the ACM conference
on Wireless Sensor Networks and Applications.
WSNA’02, Atlanta, Georgia, USA Proceedings
(pp. 112-121).
Shang, Y., Rumi, W., Zhang, Y., & Fromherz, M.
P. J. (2003, June). Localization from Mere Con-
nectivity. Paper presented at the ACM conference
on Mobile Ad Hoc Networking and Computing.
Annapolis, Maryland, USA, Proceedings (pp.
201-212).
Smets, P., & Ristic, B. (2004, April). Kalman
filters and joint tracking and classification in
the transferable belief model framework. Paper
presented at the 7th International Conference on
Information Fusion, (pp. IF04–0046), Stockholm,
Sweden.
St-Pierre, M., & Gingras, D. (2005, June). Com-
parison between the unscented Kalman filter
and the extended Kalman filter for the position
estimation module of an integrated navigation in-
formation system. Paper presented at the Institute
of Electrical and Electronics Engineering (IEEE)
Intelligent Vehicles Conference, Parma, Italy.
St-Pierre, M., & Gingras, D. (2004, March). Neural
network based data fusion for vehicle position-
ing in land navigation system. Paper presented
at the Society of Automotive Engineering (SAE)
conference. Paper no. 345, Detroit, USA.
245
 An Overview of Positioning and Data Fusion Techniques Applied to Land Vehicle Navigation Systems
Suresh, B., & Sivan, K. (2004, April). Perfor-
mance evaluation of multi-sensor data-fusion
systems in launch vehicles. Sadhana , 29(2),
175–188. India
Tsukamoto, K., Fujii, M., Itami, M., & Itoh K.
(2001, May). Study on acquisition and tracking
of vehicle location using IVC and RVC dynamic
networks. Paper presented at the IEEE Intelli-
gent Vehicle Symposium, (pp. 467-470), Tokyo,
Japan.
Vejrazka, F. (2007, April). Galileo and the Other
Satellite Navigation Systems. Paper presented at
the IEEE 17th Radioelektronika International
Conference.
246
Wan, E. A., & Merwe, R. V. D. (2000). The Un-
scented Kalman Filter for Nonlinear Estimation.
Paper presented at the Symposium on Adaptive
Systems for Signal Processing, Communication
and Control (AS-SPCC-2000), Lake Louise,
Alberta, Canada.
Wu, Q., Gao, Z., & Wan, D. (2002, April). An
adaptive information fusion method to vehicle.
Paper presented at the IEEE Position location
and navigation symposium. Proceedings (pp.
248-253).
Zhao, Y. (1997). Vehicle Location and Navigation
Systems. Artech House, Norwood, USA
 247

Chapter XIII
Efficient and Reliable
Pseudonymous Authentication
Giorgio Calandriello
Politecnico di Torino, Italy

Antonio Lioy
Politecnico di Torino, Italy

ABSTRACT

Privacy, security, and reliability are key requirements in deploying vehicular ad-hoc networks (VANET).
Without those the VANET technology won’t be suitable for market diffusion. In this chapter, the authors
are concerned with how to fulfill these requirements by using pseudonym-based authentication, designing
security schemes that don’t endanger transport safety while maintaining low overhead. At the same time
our design improves the system usability by allowing nodes to self-generate their own pseudonyms.

INTRODUCTION Such applications are sensible to attacks, for

Vehicular ad-hoc networks (VANET) are com-
posed by mobile nodes, that is, vehicles, and
road-side units (RSU) equipped with sensing,
processing and communication capabilities,
and are deployed to increase road safety and the
efficiency of transportation systems through a
number of applications, most of which will be
based on beacons, that is, vehicles disseminat-
ing frequently their position, state and eventual
situation-dependent warnings (for example, ice
on the road).
example by injecting false beacons into the net-
work, and can disclose users’ personal information
or track vehicles, as an attacker collecting VANET
data could do. Thus there’s the need of security
mechanisms and privacy-enabling technologies,
which actually are a prerequisite for deployment.
Three recent research efforts, the IEEE 1609.2
working group (IEEE1609.2, 2006), the NoW
project (Gerlach et al., 2007), and the SeVeCom
project (Papadimitratos, Buttyan, Hubaux, Kargl,
Kung & Raya, 2007), are currently working on
VANET security architectures, whose common

Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Efficient and Reliable Pseudonymous Authentication
mechanism is the use of public key cryptography
to protect V2V and V2I messages.
Message authentication, integrity, and non-re-
pudiation, as well as protection of private user in-
formation are identified as primary requirements.
On one hand, traditional public key cryptography
in combination with periodic key changing is
known as pseudonymous authentication, in which
nodes are equipped with multiple credentials,
termed pseudonyms that do not reveal the node
real identity, e.g., as those proposed by Chaum
(1985). This prevents messages signed under dif-
ferent pseudonyms to be linked (Armknecht et
al., 2007; Raya & Hubaux, 2005; Gerlach, 2005;
Raya & Papadimitratos, 2006). On the other
hand, group signatures, that is, cryptographic
primitives for anonymous authentication, are also
applicable to VANET (Ateniese & Tsudik, 1999;
Brickell et al., 2004; Chaum & van Heyst, 1991;
Syverson & Stubblebine, 1999). This guarantees
a stronger level of anonymity as an eavesdropper
cannot link any two signatures from any node.
This property comes at a higher security overhead
as it will be explained in the rest of the chapter,
while pseudonym-based authentication gathered
a warmer and wider acceptance by the research
community (Gerlach et al., 2007; IEEE1609.2,
2006; Papadimitratos, Buttyan, Hubaux, Kargl,
Kung & Raya, 2007), reason for which our focus
in this chapter is on pseudonym-based systems.
In this chapter we investigate technical issues
related to the use of pseudonyms in vehicular
communications.
First, we ask ourselves how the security over-
head can be reduced while maintaining acceptable
levels of security and application robustness.
Second, do safety applications remain reliable
in presence of security mechanisms? Finally,
can self-generation of pseudonyms still meet the
security and privacy mechanisms? In this chapter,
we provide answers to these questions. First we
define more precisely the problem at hand and
we outline our approach to solve it. Then we
analyze the costs of the proposed solutions and
their impact on transportation safety.
248
BACKGROUND
We present the idea of self-generation of pseud-
onyms in our previous work (Calandriello et al.,
2007). Zeng (2006) presents the same approach
independent of our work, and Armknecht et al.
(2007) apply it to VANET. Calandriello et al.
(2007) show that the proposed optimizations can
also be applied to the cryptosystem presented by
Zeng (2006).
The work of Raya & Hubaux (2005) presents
the main problems and challenges of VANET
security, and outlines in detail the Baseline
scheme which we summarize in the following.
Papadimitratos, Gligor & Hubaux (2006) survey
the security requirements of VANET and pro-
vide models for the system and the adversaries.
We invite the non-specialized reader to refer to
the above works for an in-depth introduction to
VANET security.
Chaum (1985) presents the original idea of
using pseudonyms, while their applicability to
VANET has been explored by other studies in the
context of the SeVeCom and NoW projects: (Ger-
lach, 2005; Gerlach et al., 2007; Papadimitratos,
Buttyan, Hubaux, Kargl, Kung & Raya, 2007;
Raya & Hubaux, 2005). Papadimitratos, Kung,
Hubaux & Kargl (2006) discuss in general the
issue of privacy in vehicular communications.
A number of works are concerned with differ-
ent aspects of security and privacy of vehicular
networks, either outlining challenges (El Zarki
et al., 2002; Parno & Perrig, 2005), describing
particular attacks (Blum & Eskandarian, 2004;
Jakobsson et al., 2004) or more general attack
overviews (Aijaz et al., 2005), offering general
suggestions towards solutions (Gerlach, 2005;
Raya, Papadimitratos & Hubaux, 2006), or propos-
ing mechanisms (Golle et al., 2004; IEEE1609.2,
2006; Li et al., 2005; Raya & Hubaux, 2005).
The work of Laberteaux & Hu (2006) proposes
to authenticate messages through the combined
use of public and symmetric key cryptography.
Freudiger et al. (2007) approach the problem
Efficient and Reliable Pseudonymous Authentication
privacy in vehicular networks from another
perspective, proposing the use of mix-zones to
change pseudonyms, and analyze the privacy
level achieved.
Many recent works propose the use of bilinear
pairings to provide privacy in VANET. Lin et al.,
(2007) use group signatures directly on VANET
messages, as we survey in the Group Signature
approach below. We discuss the advantages and
disadvantages of this scheme later in the paper. The
approach from Lu et al., (2008) is similar to ours
(Hybrid) but heavily relies on the infrastructure
to generate pseudonym, which is a requirement
hardly achievable in practice.
Torrent-Moreno, Jiang & Hartenstein (2004),
Torrent-Moreno, Corroy, Schmidt-Eisenlohr &
Hartenstein (2006), Chen et al. (2007) extensively
analyze the VANET channel conditions, provid-
ing metrics to measure the channel performance.
Chen et al. (2007) provide extensions to the ns-2
simulator to use realistic propagation models and
to reliable assess the channel conditions. Torrent-
Moreno, Santi & Hartenstein (2006) provide a
distributed algorithm to fairly share the available
bandwidth by effectively adjusting the beacon
power level.
On the safety application side, Xu et al. (2004),
Yang et al. (2004), Biswas et al. (2006), ElBatt
et al. (2006) propose schemes for safety mes-
saging and analyze their impact with respect to
delay and transportation safety, although they do
not consider the costs of security. Our previous
work (Papadimitratos, Calandriello, Hubaux &
Lioy, 2008) analyzes the impact of security on
transportation safety.
SAFETY AND SECURITY
REQIREMENTS
Vehicular communications primarily aim at en-
forcing transportation safety. This means saving
thousands of lives otherwise lost in accidents and
saving pollution, time and fuel waste due to traffic
jams and unorganized traffic.
Applications for transportation safety are
real-time: they require low delay in message
transmission and processing as emergencies
happen suddenly and drivers have to react fast
in order to avoid troubles. Human reflexes are in
the order of magnitude of 1 s. Safety applications
should be much faster (e.g. one order of magni-
tude), both on communication and processing
sides, in order to allow the driver to take quick
response to an emergency event. Their goal is
to disseminate alerts fast, in order for the driver
to be warned much earlier than he is nowadays,
without such applications. Current envisioned
safety protocols (Biswas et al., 2006; ElBatt et
al., 2006; Xu et al., 2004; Yang et al., 2004) rely
on consecutive broadcast retransmissions of the
emergency warning, and are essentially based
on redundancy, with optimizations designed to
reduce the channel load and to extend the message
dissemination area beyond the range of a given
transmitter. It’s important that the communica-
tions overhead is kept low: in a scenario with
high node density every vehicle will contribute
to saturate the channel, and this will affect the
performance of safety applications, for example
by increasing the transmission delay. Each vehicle
will contribute regardless of the content of the
messages it send; essentially, the offered load
depends on the beaconing rate, on the payload
size and on the security overhead. While the
former depends on the specific application, the
latter is directly affected by the security design;
given its relevance, it’s then critical that it’s kept
to a minimum.
As these are life-critical applications they must
be reliable, both in terms of fault-tolerance and
security, otherwise the system becomes useless.
Safety applications must be trusted as they can
endanger drivers’ life, and in general the system
must be secured to guarantee its efficiency and
avoid a formidable set of possible abuses. On the
networking side, this means that a vehicle must
trust the messages it receives. Consider, for ex-
ample, nodes that “contaminate” large portions
249
 Efficient and Reliable Pseudonymous Authentication
of the network with false information: a single
vehicle can transmit false hazard warnings (e.g.
ice formation on the pavement), which can then
be taken up by all vehicles in both traffic streams.
Or, similarly, a vehicle that meaningfully modi-
fies messages of other vehicles, or even a vehicle
that forges messages in order to masquerade as
an emergency vehicle to mislead other vehicles to
slow down and yield. These simple examples of
exploits indicate that vehicular communications
must be secured: otherwise these systems could
make anti-social and criminal behavior easier, in
ways that would actually jeopardize the benefits
of their deployment.
Thus, a comprehensive set of security mecha-
nisms integrated into the VC system is critical
for their deployment (Papadimitratos, Gligor &
Hubaux, 2006). We outline here a subset of re-
quirements which apply to safety messaging:
Message authentication and integrity. Mes-
sages must be protected from any alteration and
the receiver of a message must be able to verify
that the sender of the message is a legitimate node
of the network.
Message non-repudiation. The sender of the
message cannot deny having sent a message
Privacy and anonymity. Vehicle communica-
tion systems should not disclose or allow interfer-
ence on the personal and private information of
their users. This requirement is becoming increas-
ingly important as is the sensitivity of personal
information. In VANET, examples of “personal
information” are the most frequent areas where
a car is parked, or the most used paths between
two different points. If no privacy protection
mechanism is in place, any (passive) eavesdropper
could obtain such information and connect them
to the driver’s real identity, becoming thus able
to track her at any given point in time and space.
This Big Brother scenario is easier to achieve
than one could imagine, as it only needs a series
of receivers placed in key locations and enough
time to collect the needed information; in fact,
without privacy protection, any single message
from a given vehicle will reveal its identity.
250
In particular we require anonymity for the ac-
tions (e.g. messages or transactions) of the network
nodes with respect to a set of observers. At best
we require full anonymity: an action performed
by a given node could have been performed by
any other node in the system. Full anonymity can
be however a drawback in some circumstances as
well, for example during an emergency where it
would be desirable to track a sender of emergency
messages in order to facilitate the management
of alerts. As full anonymity is extremely costly
to achieve, as it will be explained in the follow-
ing of the chapter, we soften this requirement
to limited traceability: a node can be tracked
and its messages linked, but only for a limited
amount of time. We emphasize that this does not
automatically mean that the node is identified,
that is, its long-term identity disclosed. This lat-
ter requirement allows for a node to be tracked
during an emergency and thus to better manage
the stream of alerts.
Liability identification. Users of vehicles are
liable for their deliberate or accidental actions
that disrupt the operation of the network, and the
latter should help in the attribution of liability, for
example by providing meaningful information.
This is a requirement that implies that anonym-
ity could be removed and the node identified if
necessary. It’s a matter of policy to specify the
observer with the legal power to do so.
EXAMPLE SCENARIO
We outline an example scenario to show why the
above security requirements are important. Let’s
assume a vehicle V moving from two locations A
and B. As V moves, it receives beacons and alerts
from other vehicles. V needs some mechanism
to trust the content of the received messages,
otherwise the system would not be usable at all.
Without security, any pedestrian, car, base station
enabled with the DSRC protocol can send mes-
sages, and there would be absolutely no way to
Efficient and Reliable Pseudonymous Authentication
tell true warnings to false ones. Without security,
V could be for example forced to change route
because it receives fake alerts of “traffic jam
ahead”. As explained above, messages need to
be authenticated, and their integrity ensured, in
order to avoid false messages and alerts.
Let’s now imagine that two other vehicles in
the radio range of V have an accident. As they are
VC-enabled, they would send messages up to the
moment of crash, which V can store and reveal
during later investigation. Some other mechanism
is then needed to tell which message was sent by
which vehicle, otherwise the investigation would
be useless. This mechanism would guarantee the
property of non-repudiation.
Let’s now imagine that someone is eavesdrop-
ping on the network. Without security, it can learn
that V is on the street at a given moment, match
it with the owner data, learn for example that A
is the V’s home and B its workplace, and when
the house is empty.
ADVERSARY MODEL
We characterize the adversary with 2 dimensions:
geographic extension and method, with the former
being local or global and the latter active or pas-
sive. We note that we do not attempt to define all
possible adversaries and attacks against VANET,
but we limit ourselves to a subset which targets
the security requirements specified above.
The first adversary to consider is the global
passive, thus an eavesdropper with the ability to
overhear messages on a “significant” portion of
the network, that is, on a geographic area wide
enough for the eavesdropper to collect meaning-
ful information, and attempting of recovering the
drivers’ personal data.
The second model is the active adversary, either
local or global, which controls a radio transceiver
and can thus inject messages on the network. This
adversary is a threat for safety applications as it
disseminates bogus information on the network,
for example “car crash ahead, slow down”, when
such event did not occur. If such information is
picked up by the nearby vehicles the attacker
take over the functionality of the VANET safety
system and cause accidents or traffic jams. To
continue the previous example, nearby vehicles
will most likely abruptly slow down, starting a
potential dangerous situation.
Both types of adversaries do not have access
to legitimate credentials or, in other words, we
choose not to consider internal adversaries. Such
choice implies that vehicles are physically pro-
tected against tampering, and that such protection
is adequate, at least against the high majority of
users.
SYSTEM MODEL
VANET employ the Dedicated Short Range
Communications (DSRC) data link technology
(DSRC, 2006), according to which nodes use one
common channel to transmit beacons, that is,
periodic updates about their state. Beacons enable
safety applications, and their rate ρ range from
3 to 10 per second, with the latter value deemed
mandatory; this is also the value we assume in
the following of the chapter. Beacon payload
includes at minimum the position and the speed
of the vehicle, as well as a timestamp and eventu-
ally different warnings. Safety applications run
on each vehicle and identify dangerous situations
from the beacon content. Their response is to send
warnings to other vehicles and alerts to the driver,
but they will not override the latter.
We assume that alerts, when transmitted,
are part of the beacon payload. The timestamp
combined with digital signatures also gives pro-
tection from replay attacks. For the scope of this
chapter, we don’t consider multi-hop routing and
focus on broadcast transmissions only. Finally
we consider vehicle-initiated transmissions only,
as fixed infrastructure nodes don’t require their
privacy to be protected.
251
 Efficient and Reliable Pseudonymous Authentication
We assume that each vehicle has a long-term,
unique identity that combines a number of attri-
butes (Papadimitratos, Gligor & Hubaux, 2006)
and a set of cryptographic keys. We assume for
simplicity one global and unique Certification
Authority (CA) managing credentials with which
all legitimate nodes are registered, and with the
legal power to revoke credentials and disclose
identities of the nodes, if deemed necessary for
liability attribution.
PROB AND APPROACH
OVERVIEW
We want to secure VANET by fulfilling the re-
quirements described above. Regarding message
unlinkability, we require that any two messages
m1, m2, generated at times t1, t2 respectively cannot
be linked if t2 > t1 + τ, where τ is a user-specified
parameter. This means in practice switching to
a new set of keys every τ seconds, and that only
signatures generated with the same key can be
linked to each other. This is the basic idea behind
pseudonymous authentication, and τ is termed the
pseudonym lifetime.
Our first proposal, in order to simplify the
credentials management, is to allow vehicles
to self-generate their own pseudonyms and the
corresponding private keys. This eliminates the
need of pre-loading, storing, refilling, as well as
obtaining pseudonyms over an “out-of-band”
channel, and improves the system usability, as
vehicle’s privacy is not compromised if no new
pseudonyms are available (e.g., the initial supply is
exhausted) and it’s not necessary to over-provision
a vehicle with pseudonyms. Self-generated pseud-
onyms must be certified: in the baseline scheme
presented below the certification is provided by
means of digital certificates created by the CA
and installed on the vehicle. We introduce the use
of Group Signatures as a mean to self-certify the
self-generated pseudonyms while achieving the
same level of trust.
252
Self-generation of pseudonyms imposes higher
transmission and processing costs, due to the use
of Group Signatures. Our second contribution
consists in a number of optimizations to signifi-
cantly reduce the overhead without compromising
security, that is, by keeping signing and verifying
each message. Finally, we identify the problem
of changing pseudonyms during an emergency
situation, and solve it by proposing an additional
scheme.
Our goal in this chapter is to analyze the ef-
fectiveness of pseudonymous authentication, and
how it could be used in practice. We do not aim at
formal definitions of privacy, nor to demonstrate
that the proposed schemes are provably secure.
We refer the reader to the “Background” section
for related work in the field with in-depth security
analysis.
SCHEME OUTLINE
We describe here our proposal for pseudonymous
authentication and compare it with pre-existing
schemes, the first of which is termed Baseline
Pseudonym (BP) and also relies on pseudonyms,
while the Group Signatures (GS) relies on a dif-
ferent type of cryptosystem. Our scheme, named
Hybrid, essentially combines the two and allows
the on-the-fly generation of own pseudonyms
with self-certification through the use of bilinear
pairings.
Baseline Pseudonyms. Each vehicle V is
equipped with a set of anonymous public keys,
termed pseudonyms, that is, keys certified by
the CA and not holding any information about V
itself. The CA generates a certificate CertCA(KiV )
on the pseudonym KiV , i-th pseudonym for node
V, which uses the corresponding private key kiV
to digitally sign its messages and attaches KiV
and CertCA(KiV ) to them to allow their verifica-
tion by other vehicles. The message format is as
follows:
Efficient and Reliable Pseudonymous Authentication
m, σiV(m), KiV, CertCA(KiV ) (1)
with σiV () denoting V ‘s signature generated with
its i-th private key kiV and m the message payload.
To identify the sender from Msg. (1) when needed,
the CA needs to maintain a mapping between the
set of pseudonyms { KiV } provided to V and V’s
long-term identity.
Assuming the public key of the CA available, a
node receiving Msg. (1) first validates CertCA(KiV
), then if the validation is successful it verifies σiV
(m). The validation of CertCA(KiV ) requires to check
if KiV is not revoked, and this check is performed
through a Certificate Revocation List (CRL) dis-
tributed to vehicles via the fixed infrastructure
(Raya, Papadimitratos & Hubaux, 2006).
This scheme meets all the above security re-
quirements: messages are authenticated and their
integrity protected, and digital signatures ensure
non-repudiation. As explained above, the CA can
perform an inverse mapping and disclose V’s
identity when needed. A global eavesdropper can
follow V for the lifetime of a given pseudonym,
but cannot link the set of pseudonyms used by V
to each other, while an active adversary cannot
forge m as it does not have access to the needed
credentials.
This scheme however requires periodic pseud-
onym refilling, which essentially is a burden for
the driver, and an index of poor system design:
the driver should not care about the details of
vehicular communications security.
Group Signatures. This cryptosystem relies
on the concept of group, whose members are all
the legitimate vehicles registered with the group
manager (whose role is played by the CA), which
installs a secret group signing key gskV and a public
group public key gpkCA. The key property of the
cryptosystem is that any member V can issue a
group signature ΣCA,V on a message on behalf of
the group, and any receiver with the gpkCA can
verify that a legitimate group member signed
the message but cannot tell which one. Another
property is that it’s impossible to link any two
signatures from a group member. The message
format is as follows:
m, ΣCA,V (m) (2)
The concept of group signatures was originally
introduced by Chaum and van Heyst (1991) and
revisited in subsequent works, e.g., (Ateniese &
Tsudik, 1999; Brickell et al. 2004; CVIS; Syverson
& Stubblebine, 1999). See (Bellare, Miccinaccio
& Warinschi, 2003; Bellare, Shi & Zhang, 2004)
for formal definitions and security analysis. For
the rest of the discussion, we assume and utilize
the group signature scheme proposed by Boneh
& Shacham (2004), including the revocation
method they propose, which is again based on
the concept of Revocation List (RL). To allow li-
ability attribution, the scheme allows for an Open
operation (Bellare, Miccinaccio & Warinschi,
2003; Bellare, Shi & Zhang, 2004), that the CA
only can perform in order to disclose the signer’s
identity if presented with Msg. (2)
A node receiving Msg. (2) verifies ΣCA,V (m),
which tells that a legitimate member generated
m. The node itself cannot determine the identity
of the node. Being the group signature a digital
signature, the integrity of m is also protected and
non repudiation ensured.
This scheme meets all the above requirements,
and guarantees full anonymity as well, at the
cost of high channel and processing overhead.
The scheme cannot “scale down” the provided
anonymity level: nodes are always fully anony-
mous.
Hybrid Scheme. The scheme we propose here
is a combination of the two previous ones. Each
node is equipped with group keys as in the GS
scheme, and generates a set of pseudonyms { KiV }
and corresponding private keys. It then generates
a group signature ΣCA,V () on each pseudonym KiV,
which essentially is a self-certification CertH (KiV)
of the pseudonym, with the superscript H indi-
cating that it’s a different type of certificate if
compared to the one in BP. V signs the message
253
 Efficient and Reliable Pseudonymous Authentication
m with the self-generated private key kiV, and then
attaches the corresponding pseudonym and self-
certificate CertH (KiV ) as shown below:
m, σiV (m), KiV, CertH (KiV ) (3)
Any node with the gpkCA can verify the
CertH (KiV ). This validates the pseudonym KiV,
which enables verification of the σiV (m). The
properties of group signatures guarantee that the
receiver can only verify that a legitimate group
member signed Msg. (3), but can neither identify
V nor can link this message to any previous other
CertH (KiV ) sent by V. Messages signed under the
same kiV can be linked, but as in the Baseline
scheme this can be done only during the lifetime
of the pseudonym in use.
The σiV (m) provides message integrity, authen-
tication and non-repudiation. As in GS, the Open
operation on the CertH (KiV ) allows to disclose the
identity of the signer.
OPTIMIZATIONS
Optimization 1
The CertH (KiV ) is computed only once at the
sender side, as it’s a digital signature computed
on a piece of information (the KiV) that doesn’t
change (during its lifetime τ). For the same reason
it’s validated only once at the receiver’s side and
then stored, even if it’s sent in principle with every
message. As per the group signature properties,
two signatures computed on the same informa-
tion would look different, but both of them will
be accepted as valid at the receiver. Since the KiV
is sent in the clear, it’s sufficient to use one single
valid CertH (KiV ). This optimization is effective
because τ >> ρ-1.
254
Optimization 2
The signature σiV (m) is appended to every mes-
sage, but the pseudonym and the certificate only
to every α messages. We term the messages like
Msg. (3) as LONG, and the remaining α – 1, car-
rying only the σiV (m), as SHORT, and as Cer-
tificate period. To bind a SHORT message to the
corresponding pseudonym, a 4-byte keyID field
(a random number) is added to all messages.
This Optimization is designed to reduce the
packet overhead. However it can harm proto-
col robustness, if the message carrying Ki+1V,
CertHCA(Ki+1V ) is not received. Then, nodes in
the range of V must wait for α messages before
the next LONG message, while being unable to
validate any message from V. The danger arises
especially if the vehicles are close to each other
and/or move at high relative speeds. Thus, we
propose the following scheme to mitigate this
problem:
Optimization 3
V repeats the transmission of the new Ki+1V,
CertHCA(Ki+1V ) when the pseudonym is changed,
for β times. β is named the Push period.
CRYPTOGRAPHIC OVERHEAD
In the following of the chapter we’ll use ECDSA
(IEEE 1363a, 2004) as signature algorithm for
σiV (m), and the Group Signatures algorithm as
proposed by Boneh and Shacham (2004), with
security levels of t=80 bits for σiV (m) and t=128
for CertH(KiV ) and CertCA(KiV ). In fact, high se-
curity is needed only for the long-lived GS keys.
A comparison of the chosen algorithms is made
in Table 1, with figures for GS derived from an
estimate of the needed number of scalar multipli-
cations (Brown et al. (2001); Koblitz & Menezes,
2005) and a benchmark of the same operation
Efficient and Reliable Pseudonymous Authentication

Table 1. Computation costs and communication overhead for different signing algorithms
Algorithm Security level Signature Sign (ms) Verify
(bits) (bytes) (ms)
GS 80 151 17.8 15.6
GS 128 225 53.7 49.3
ECDSA- 80 48 0.5 3
192
ECDSA- 128 64 0.8 4.2
256

on a 1.5GHz Centrino PC, on which we ran the 52.3 ms/msg for verification. Table 2 summarizes
OpenSSL speed benchmarks for ECDSA. the costs per message.

Computation Overhead
Without considering Optimization 2, the security
Table 1 compares the performances of the dif- overhead sums to 137 bytes for BP, given by one
ferent algorithms in terms of computation time, KiV, one CertCA(KiV ) and one σiV (m). In GS the
signature and key sizes. overhead is given by one ΣCA,V (m), that is, 225
To compute the processing delays for indi- bytes per message. For Hybrid, the overhead is
vidual messages it suffices to sum the individual σiV (m), KiV, CertHCA(KiV ), in total 298 bytes per
delays of the corresponding primitives. In the BP message. The keyID field adds 4 bytes to these
scheme V computes one ECDSA signature per figures. For the α - 1 SHORT messages the over-
message, and the receiver validates one ECDSA- head is the sum of one σiV (m) plus the keyID,
192 (the σiV (m)) per message and one ECDSA-256 that is, 52 bytes.
(the CertCA(KiV )) for each KiV. The overall costs
are thus 0.5 ms/msg for signing, and 7.2 ms/msg Revocation
for verification.
In the GS scheme the costs are given by one Revocation is an important side of any public key
signature/verification of ΣCA,V (m), thus 53.7 ms/ cryptosystem. Traditional approaches are based
msg for signing and 49.3 ms/msg for verification. either on the offline verification of the public key
In the Hybrid scheme the costs are given by the against a revocation list (RL) or the online verifi-
generation and verification of one σiV (m) plus one cation of the key status (OCSP). In VANET, the
CertH (KiV ), totaling 54.2 ms/msg for signing, and second approach requires constant connection to
a central server and is thus not feasible in prac-
tice, while a revocation list can be periodically
disseminated through the RSU. We assume such
Table 2. Processing costs and overhead for different an approach in the rest of the chapter.
packet types. © 2008 IEEE, used with permission We note that the size of the CRL is much
higher in BP than in the other approaches, as it
Sign (ms) Verify (ms) Overhead
(bytes) includes revoked pseudonyms, not nodes. For BP,
BP LONG 1.3 7.2 141 each KiV must be checked against a CRL through
Hybrid LONG 54.2 52.3 302 a number of string comparisons linear to the size
SHORT 0.5 3 48 of the CRL. For GS, the revocation is performed

255
 Efficient and Reliable Pseudonymous Authentication
for each ΣCA,V (m) and involves two pairing calcu-
lations plus one string comparison, with the RL
linear in the number of revoked group members.
In the Hybrid scheme the revocation requires
two pairing calculations (as in GS), per KiV (as
in BP); the RL is again linear in the number of
revoked group members. Assuming that a string
comparison takes 1e-6 s. and a pairing calculation
5e-3 s., if RLCheck is the cost of comparing one
entry in the RL for GS, for BP the same operation
costs 1e-4* RLCheck. The cost in Hybrid is again
RLCheck as the revocation check in the scheme
requires pairing computation.
As said before, the CRL for BP includes all
the not yet used pseudonyms, with a cost in the
order of 1e5 per node. The performances depend
on the data structure and search algorithm used to
handle the revocation data in memory. Assuming
a hash map, the cost is an O(1) search time, with
a prior hash map construction needed each time
a new RL is received, bearing a O(n) cost. The
cost would then be 1e-4 * RLCheck. In the GS
scheme, the revocation will take τ * ρ * RLCheck
(e.g. for the assumed values of τ = 60 s. and ρ = 10
msg/s. the expression becomes 600 * RLCheck).
The comparison is made by considering the total
number of signatures that have to be verified
against the RL during one pseudonym lifetime:
each message requires such verification in GS,
while it suffices to do it once per pseudonym in
BP and Hybrid.
OPTIMIZATION EVLUATION
Simulation Environment
Our simulator is partly custom-built in C and
partly relies on ns-2, and implements the Dedi-
cated Short Range Communications (DSRC) data
link technology (DSRC, 2006). Vehicles transmit
beacons, which include vehicle’s direction and
location, on a common channel at a rate ρ = 10
beacons/sec. The beacon’s payload is set to 200
256
bytes. Table 2 provides the cryptographic com-
munication overhead per message. We implement
the physical layer and a realistic radio propagation
model (Nakagami, 1960), as in Torrent-Moreno,
Jiang & Hartenstein (2004) and Chen et al., (2007)
with an intended (nominal) transmission range of
d = 200 meters (DSRC, 2006; Xu et al., 2004)
We simulate scenarios of four- and eight-lane
highways, with vehicles moving in two opposing
two- and four-lane flows. Vehicles are randomly
placed along each lane with an average inter-ve-
hicle spacing of s meters, and a random-drawn v.
We set s = 20 meters and v = 80km/h. We model
wet road conditions by setting the friction coef-
ficient μk to 0.41, which corresponds to vehicle
braking capabilities of 4 m/s2. We consider an
“emergency braking alarm” application, with
such messages transmitted along with beacons,
and we investigate how many vehicle collisions
occur when safety messaging is used with and
without security.
Transportation Safety Simulation
Stup
We focus on a platoon of one hundred cars in
the highway scenario described above. Vehicles
in the platoon, denoted as V1 to V100, are placed
along a single lane, with V1 at the head and V100
at the back. At a given point in time the leading
vehicle V1 performs an emergency brake and starts
sending warning messages. Once some Vi, with i
> 1 receives the warning, it warns its driver and
starts sending warnings itself. We recall that Vj,
with j ≠ i, can accept the warning as valid if and
only if it already validated Vi’s pseudonym. As
proposed by Xu et al., (2004), Yang et al. (2004),
Biswas et al. (2006), ElBatt et al. (2006) when
Vi receives a warning from a Vj with j > i (e.g. a
vehicle behind itself), it stops transmitting warn-
ings. The rationale is that at least one vehicle Vj
behind Vi has already been warned, thus it’s in a
better position to keep warning Vk, for k > j. We
consider here all braking actions as “emergency”.
Efficient and Reliable Pseudonymous Authentication
Braking has two effects: (i) it turns on the ve-
hicle rear red lights that visually warn vehicles
within line of sight (as we assume a platoon on
a single lane, Vi will only warn Vi+1 with its rear
red lights), and (ii) it triggers the transmission of
own warning messages. Besides warning other
vehicles, a warned Vi clearly warns its driver
who starts braking according to how fast are his
reflexes. Driver reactions, modeled as a random
reaction delay between 0.75 and 1.5 seconds, are
triggered by both VC-enabled and visual (red
light) warnings.
We choose the pseudonym lifetime τ to be 60
s., and we consider the first 60 s. of the simulated
time as a warm-up period, during which no emer-
gency conditions arise. By the end of the warm up
period, this allows for one pseudonym change and
discovery of neighbors by validating their corre-
sponding certificates. This is a realistic situation:
at any given point in time, when an emergency
arises, vehicles have already discovered some of
their neighbors and can immediately validate their
warnings. The simulation concludes when all ve-
Figure 1. Probability of successful packet reception, Highway scenario, Baseline scheme, 8 lanes. ©
2008 IEEE, used with permission
hicles in the platoon are immobile, as V1 does not
resume any motion after its emergency braking,
and any Vi cannot physically go beyond Vi-1.
These simulated conditions are particularly
challenging, with high vehicle density and high
average velocities for the given vehicle spacing.
High density results in high network load, given
the assumed beaconing rates, and thus increased
packet loss rates. We do not consider any other op-
timization to reduce processing overhead beyond
considering messages from vehicles with the same
heading. This is reasonable for shoulder-separated
highways, but we acknowledge that it’s hard to
achieve in an urban or rural setting. Finally, we
remark that we do not simulate lane-changing
as a means to avoid collisions, but leave it as an
item for future work.
Our results are shown in Figures 1 and 2
respectively: the y axis is the probability of suc-
cessful reception of a packet, as a function of the
distance between the packet sender and receiver
(the x-axis).
257
 Efficient and Reliable Pseudonymous Authentication

Figure 2. Probability of successful packet reception, Highway scenario, Hybrid scheme, 8 lanes. © 2008
IEEE, used with permission

Although the message processing load may
be reduced, e.g., by using the node mobility
dynamics, this is not the case for the channel
load, because any transmission “contributes” in
generating interference and increases the number
of packet collisions. This is why the probability
of message reception is overall lower in Figure
2 than in Figure 1. Looking at each plot indi-
vidually, safety messaging with “No Security”
achieves the highest reliability consistently as the
distance changes, while the setting α = 1 achieves
the lowest reliability in both schemes, due to
transmission overhead: “No Security” employs
a constant packet size of 200 bytes, while each
of the different α settings leads to a different per-
packet average overhead, as Table 3 shows. Op-
timization 2 improves communication reliability
as α increases (and the corresponding overhead
is reduced), and the effect is more evident for
Hybrid, which employs larger certificates. Table
3 shows the average communication overhead
due to cryptography as a function of α. for both
the Hybrid and BP schemes. We observe that
“over-optimizing” does not significantly improve
channel reliability.

Transportation Safety
Table 3. Average packet overhead, in bytes per
packet, given different security schemes and We compare the impact of safety messaging on
different α settings. © 2008 IEEE, used with transportation safety when used without (the “No
permission Security” plot) and with security protocols. In the
latter case we set two values of Optimization 3,
α 1 5 10 15 30 50
β=0 (no use of such optimization) and β=5. With
Hybrid 502 299 273 265 257 253
the underlying assumption that nodes are able to
process all messages they receive, Figures 3, 4
BP 341 266 257 254 251 250

258
Efficient and Reliable Pseudonymous Authentication

Figure 3. Collisions in a platoon of 100 cars, 8-lane highway scenario, baseline pseudonym scheme. ©
2008 IEEE, used with permission

Figure 4. Collisions in a platoon of 100 cars, 8-lane highway scenario, hybrid scheme. © 2008 IEEE,
used with permission

259
 Efficient and Reliable Pseudonymous Authentication
show the percentage of crashes in the platoon as
a function of α.
We note that safety messaging reduces crashes
to around 10% of the platoon vehicles: this is a
significant result as in the absence of vehicle
communications, for exactly the same scenarios,
80%-100% of the vehicles crash (we performed
the simulations but we do not show it to keep the
figures readable). Safety messaging with no se-
curity achieves the lowest percentage of crashes,
confirming the intuition. We use this case as a
benchmark; α is not a relevant parameter and
curves in Figures 3, 4 are flat with minor vari-
ability due to the randomly seeded simulations.
The reasons of its efficiency are the lowest network
overhead and no restrictions on which alert mes-
sage can be validated. In contrast, the tuning of
the secure VC protocols affects the effectiveness
of the safety application. Figures 3 and 4 show
that fewer crashes happen as α increases, but then
a more accidents for high α values. One reason
is that the increase of α reduces the channel load
and thus increases the per-packet reception prob-
ability; another reason is that when α is high (e.g.
30 or 50), the authentication delay for a receiver
missing a LONG message increases: for example,
for α=50 the authentication delay after the loss of
one LONG message is 5 seconds, and even more
if more LONG messages (and not just one) are
missed. Optimization 3 is effective in reducing
the number of crashes for a given α compared to
the non-optimized case (β=0), by significantly
reducing authentication delays at the cost of a
slightly higher overhead. Figures 3 and 4 show
that β=5 yields an improvement exactly for high
α values, where the authentication delay is higher.
In the end, Optimization 3 brings the effectiveness
of the application close to the unsecured protocol
and overall, the investigated safety application,
if tuned appropriately, is equally effective on top
of both Hybrid and BP schemes.
260
CONCLUSON AND FUTUREWORK
In this chapter we propose to manage security
credentials in VANET through self-generation
and self-certification of pseudonyms. This greatly
simplifies the security management and makes
a step towards a usable system. We propose to
employ group signatures to generate certificates
which satisfy the requirements of anonymity and
liability attribution, and we show that the compu-
tational cost and the overhead are comparable to
the baseline approach thanks to the optimizations
we propose. We identify a potential problem in
the case of a pseudonym change when vehicles
are close, and we propose a mechanism that ad-
dresses it.
Next, we analyze the costs imposed by security
on the transportation systems. First we analyze
data link performance, obtaining packet recep-
tion probability curves for the pseudonym-based
security systems, then we analyze the impact
of safety messaging and security and privacy-
enabling technologies on transportation safety,
showing that secure communication schemes
achieve safety levels comparable to those with
no security at all.
Overall, we perform a detailed investigation
on pseudonym-based authentication, analyzing
several system issues and showing how these
security mechanisms can be applied in practice.
Future research directions are the dynamic ad-
aptation of the security parameters α and β to the
traffic conditions, an effective management of
certificates, multiple CA and CRL, and a detailed
analysis on the computation overhead. Regarding
the latter, we performed the initial steps towards
this direction in our previous work (Papadimitra-
tos, Calandriello, Hubaux & Lioy, 2008).
RE
Aijaz, A., Bochow, B., Dötzer, F., Festag, A.,
Gerlach, M., Kroh, R., & Leinmüuller, T. (2006).
Efficient and Reliable Pseudonymous Authentication
Attacks on intervehicle communication systems
- an analysis.In 3nd International Workshop on
Intelligent Transportation, Hamburg.
Armknecht, F., Festag, A., Westhoff, D., & Zeng,
K. (2007). Cross-layer privacy enhancement and
non-repudiation in vehicular communication. In
WMAN 2007, Bern.
Ateniese, G., & Tsudik, G. (1999). Group sig-
natures à-la-carte. In SODA ‘99: Proceedings
of the tenth annual ACM-SIAM symposium on
Discrete algorithms, (pp. 848-849), Philadelphia,
PA, USA, Society for Industrial and Applied
Mathematics.
Bellare, M., Micciancio, D., & Warinschi, B.
(2003). Foundations of group signatures: for-
mal definition, simplified requirements and a
construction based on trapdoor permutations.
In Eli Biham, (Ed.), Advances in cryptology -
EUROCRYPT 2003, vol. 2656 of Lecture Notes
in Computer Science, (pp. 614-629), Warsaw,
Poland, Springer-Verlag.
Bellare, M., Shi, H., & Zhang, C. (2004). Founda-
tions of group signatures: The case of dynamic
groups. In CT-RSA’05, Lecture Notes in Computer
Science, Springer-Verlag.
Biswas, S., Tatchikou, R., & Dion, F. (2006).
Vehicle-to-vehicle wireless communication proto-
cols for enhancing highway traffic safety. In IEEE
Communications magazine, 44 , 74-82).
Blum, J., & Eskandarian, A. (2004) The threat of
intelligent collisions. IT Professional, 6, 24-29.
Boneh, D., Boyen, X., & Shacham, H. (2004).
Short group signatures, In Crypto ‘04, LNCS
3152 (pp. 41-55).
Boneh, D., & Shacham, H. (2004). Group sig-
natures with verifier-local revocation. In CCS
‘04, (pp. 168–177), New York, NY, USA, ACM
Press.
Brickell, E., Camenisch, J., & Chen, L. (2004).
Direct anonymous attestation. In CCS ‘04, (pp.
132-145), New York, NY, USA, 2004. ACM
Press.
Brown, M., Hankerson, D., Lopez, J., & Menezes,
A. (2001). Software implementation of the NIST
elliptic curves over prime fields. In CT-RSA 2001,
(pp. 250-265), London, UK. Springer-Verlag.
Calandriello, G., Papadimitratos, P., Hubaux, J.-P.,
& Lioy, A. (2007). Efficient and robust pseudony-
mous authentication in VANET. In Proceedings
of the Fourth ACM international Workshop on
Vehicular Ad Hoc Networks (Montreal, Quebec,
Canada, September 10th, 2007). VANET ‘07.
ACM, New York, NY, (pp. 19-28).
Chaum, D. (1985). Security without identification:
Transaction systems to make big brother obsolete.
Commun. ACM, 28(10), 1030-1044.
Chaum, D., & van Heyst, E. (1991). Group signa-
tures. In EUROCRYPT, (pp 257-265).
Chen, Q., Schmidt-Eisenlohr, F., Jiang, D., Tor-
rent-Moreno, M., Delgrossi, L., & Hartenstein,
H. (2007). Overhaul of IEEE 802.11 modeling
and simulation in ns-2. In Proceedings of the
10th ACM Symposium on Modeling, analysis,
and simulation of wireless and mobile systems
(Chania, Crete Island, Greece, October 22-26,
(pp 159-168).
CVIS: The CVIS project, Retrieved September,
2008, from http://www.cvisproject.org/.
DSRC (2006). Dedicated Short Range Com-
munications. Retrieved September, 2008, from
http://grouper.ieee.org/groups/scc32/dsrc/index.
html.
ElBatt, T., Goel, S., Holland, G., & Krishnan, H.
(2006). Cooperative collision warning using dedi-
cated short range communications. In Proceed-
ings of the Third ACM International Workshop
on Vehicular Ad Hoc Networks (Los Angeles,
261
 Efficient and Reliable Pseudonymous Authentication
CA, USA, September 29th, 2007). VANET ‘06.
ACM, New York, NY, (pp. 1-9).
Freudiger, J., Raya, M., Feleghhazi, M., Papad-
imitratos, P., & Hubaux, J.-P. (2007). Mix Zones
for Location Privacy in Vehicular Networks. In
Proceedings of the First International Workshop
on Wireless Networking for Intelligent Transpor-
tation Systems (WiN-ITS 2007)
Gerlach, M. (2005). VaneSe - an approach to
VANET security. In V2VCOM 2005, San Diego,
California, USA.
Gerlach, M., Festag, A., Leinmüller, T., Goldacker,
C., & Harsch, C. (2007). Security architecture
for vehicular communication. In 2nd Interna-
tional Workshop on Intelligent Transportation,
Hamburg.
Golle, P., Greene, D., & Staddon, J. (2004). De-
tecting and correcting malicious data in VANETs.
In VANET ‘04, (pp. 29-37). New York, NY, USA:
ACM Press.
IEEE 1363a (2004). IEEE standard specifications
for public-key cryptography- amendment 1: Ad-
ditional techniques.
IEEE1609.2 (2006). IEEE trial-use standard for
wireless access in vehicular environments - se-
curity services for applications and management
messages.
Jakobsson, M., Wang, X., & Wetzel, S. (2004).
Stealth attacks in vehicular technologies. In
VTC-Fall 2004.
Koblitz, N., & Menezes, A. (2005). Pairing-based
cryptography at high security levels. Cryptology
ePrint Archive, Report 2005/076.
Laberteaux, K., & Hu, Y.-C. (2006). Strong
VANET security on a budget. In ESCAR 2006,
Berlin, Germany.
Li, M., Poovendran, R., Sampigethaya, K., &
Huang, L. (2005). CARAVAN: Providing location
262
privacy for VANET. In ESCAR 2005, Cologne,
Germany.
Lin, X., Sun, X., Ho, P., & Shen, X. (2007). GSIS:
A Secure and Privacy-Preserving Protocol for
Vehicular Communications. In IEEE Transactions
on Vehicular Technology 56, 3442-3456.
Lu, R., Lin, X., Zhu, H., Ho, P., & Shen, X. (2008).
ECPP: Efficient Conditional Privacy Preservation
Protocol for Secure Vehicular Communications.
In The 27th IEEE International Conference on
Computer Communications (INFOCOM 2008),
Phoenix, Arizona.
Nakagami, M. (1960). The m-distribution, a
general formula of intensity distribution of the
rapid fading. In W. G. Hoffman (Ed.), Statistical
methods in radio wave propagation, (pp. 3-36),
Oxford, UK. Pergamon.
OpenSSL. Retrieved September, 2008, from
http://www.openssl.org.
Papadimitratos, P., Buttyan, L., Hubaux, J.-P., Kar-
gl, F., Kung, A., & Raya, M. (2007). Architecture
for secure and private vehicular communications.
In ITST’07, Sophia Antipolis, France.
Papadimitratos, P., Calandriello, G., Hubaux, J.-
P., & Lioy, A. (2008). Impact of Vehicular Com-
munication Security on Transportation Safety.
In IEEE MOVE 2008, Phoenix, (AZ), US, April
17, 2008.
Papadimitratos, P., Gligor, V., & Hubaux, J.-P.
(2006). Securing Vehicular Communications -
Assumptions, Requirements, and Principles. In
ESCAR 2006, Berlin, Germany.
Papadimitratos, P., Kung, A., Hubaux, J.-P., &
Kargl, F. (2006). Privacy and Identity Manage-
ment for Vehicular Communication Systems: A
Position Paper. In Workshop on Standards for
Privacy in User-Centric Identity Management,
Zurich, Switzerland, July 2006.
Efficient and Reliable Pseudonymous Authentication
Parno, B., & Perrig, A. (2005). Challenges in
securing vehicular networks. In HotNets-IV, Col-
lege Park, MA, USA.
Raya, M., & Hubaux, J.-P. (2005). The security
of vehicular ad hoc networks. In SASN ‘05, (pp.
11-21), New York, NY, USA: ACM Press.
Raya, M., Papadimitratos, P., & Hubaux, J.-P.
(2006). Securing vehicular communications. IEEE
Wireless Communications magazine, 13(5).
Syverson, P. F., & Stubblebine, S. G. (1999).
Group principals and the formalization of ano-
nymity. In World Congress on Formal Methods,
(pp. 814-833)
Torrent-Moreno, M., Corroy, S., Schmidt-
Eisenlohr, F., & Hartenstein, H. (2006). IEEE
802.11-based one-hop broadcast communications:
understanding transmission success and failure
under different radio propagation environments.
In MSWiM ‘06, (pp. 68-77), New York, NY, USA,
ACM Press.
Torrent-Moreno, M., Jiang, D., & Hartenstein, H.
(2004). Broadcast reception rates and effects of
priority access in 802.11-based vehicular ad-hoc
networks. In VANET’04, (pp. 10-18), New York,
NY, USA. ACM Press.
Torrent-Moreno, M., Santi, P., & Hartenstein, H.
(2006). Distributed fair transmit power adjust-
ment for vehicular ad hoc networks. In SECON
‘06, (pp. 479-488).
Xu, Q., Mak, T., Ko, J., & Sengupta, R. (2004).
Vehicle-to-vehicle safety messaging in DSRC.
In VANET ‘04, (pp. 19-28), New York, NY, USA.
ACM Press.
Yang, X., Liu, J., Zhao, F., & Vaidya, N. H. (2004).
A vehicle-to-vehicle communication protocol for
cooperative collision warning. In First Annual
International Conference on Mobile and Ubiqiot-
ous Systems: Networking and Services.
El Zarki, M., Mehrotra, S., Tsudik, G., & Ven-
katasubramanian, N. (2002). Security issues in
a future vehicular network. In Proceedings of
EuroWireless 2002
Zeng, K. (2006). Pseudonymous pki for ubiq-
uitous computing. In Atzeni A. & Lioy, A.
(Ed.) EuroPKI 2006, (pp.207-222), Turin, Italy,
Springer-Verlag.
263
264

Chapter XIV
Simulation of VANET
Applications
Valentin Cristea
University Politehnica of Bucharest, Romania

Victor Gradinescu
University Politehnica of Bucharest, Romania

Cristian Gorgorin
University Politehnica of Bucharest, Romania

Raluca Diaconescu
University Politehnica of Bucharest, Romania

Liviu Iftode
Rutgers University, USA

ABSTtract

This chapter systematically presents actual issues regarding the simulation of VANET applications.
Some of them refer to challenges in developing VANET simulators. The chapter discusses simulator
architectures, models used for representing the communication among vehicles, vehicles mobility
features, and simulation tool implementation methods. A critical analysis of the solutions adopted in
some well-known actual simulators is also included. Other issues relate to the use of simulation in the
evaluation of applications that aim at improving the traffic safety and control. Representative city and
highway application scenarios are discussed, and results that can be obtained by simulation, along with
ways these results can be exploited by VANET developers and users are highlighted. Future trends in the
development of simulators that produce more accurate results and their use for the evaluation of more
sophisticated traffic control solutions are also included.

Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.
Simulation of VANET Applications
Introduction
Vehicle-to-vehicle communication is a very
actual and challenging topic. Vehicles equipped
with devices capable of short-range wireless
connectivity can form a particular mobile ad-hoc
network, VANET—Vehicular Ad-hoc NETwork.
The existence of such networks opens the way
for a wide range of applications. Two of the most
important classes of such applications are those
related to route planning and traffic safety. Route
planning aims to provide drivers with real-time
traffic information, which, in the absence of a
VANET, would require an expensive infrastruc-
ture. By contrast, the VANET approach is scal-
able and has low maintenance costs. Moreover,
short-range wireless communication technologies
(such as 802.11) have no associated cost, other than
the communication devices. Safety applications
involve disseminating urgent information, which
is not present in the driver’s field of view, or it is
difficult to notice for reasons such as fog or other
vehicles obstructing the line of sight. For instance,
a lot of accidents occur in foggy conditions be-
cause drivers notice too late that some kind of
incident has occurred in front of them. Safety at
intersections could also be enhanced, since the
risk of collisions could be detected in advance and
the driver could be warned seconds before what
would otherwise be an imminent accident.
The evaluation of VANET protocols and ap-
plications could be made through real outdoor
experiments, which are time-costly and claim
for a large number of resources in order to obtain
significant results. Instead, simulation is a much
cheaper and easier to use method. Obviously, this
leads network and application developers to use
simulation in order to evaluate different simple
or complicated and innovative solutions before
implementing them. In turn, this stimulated the
interest for the development of simulators that
easily integrate the models and respond to the
requirements of VANET applications. Simula-
tors have become indispensable tools at least
in the initial phases of the VANET application
engineering process.
This chapter aims to present the features
and trends in VANET applications simulation,
to explain the sources of its complexity and to
describe actual solutions. The main challenges
are presented in the next section, which intro-
duces the three models involved in VANET
simulations: the network and traffic models are
determined by the currently used technology;
the third model corresponds to the application
itself. Then, general features of VANET simula-
tors are presented. We describe an architecture
that integrates the first two models and allows
an easy interfacing with the third model. Details
about the network / communication model are
also given, and several components of the traf-
fic / mobility model are then discussed. Current
challenges in developing VANET simulators,
and the description of solutions adopted so far
are subjects of a separate section, in which issues
regarding the performance of different simulation
tools are presented. Another section is dedicated
to application scenarios, the results that can be
obtained by simulation, and the ways these results
can be exploited by VANET developers and users
to improve the car-to-car communication and the
traffic control. The chapter ends with conclusions
and future trends.
Background
Simulating a VANET involves two different
aspects. First, there are issues related to the com-
munication among vehicles. Network simulators,
like The Network Simulator—ns-2 (2008) and Jist/
SWANS (2008) cope with communication issues
and focus on network protocol characteristics.
The second very important aspect is related
to the mobility of the VANET nodes. Traffic
simulators take into account the traffic model,
not necessarily in conjunction with VANETs. For
example, TRANS—Traffic Network Simulator
265

(Katz, 1963) was used to optimize the timing
of traffic lights in the absence of inter-vehicular
communication.
In the majority of VANET applications, ve-
hicles react to messages. For instance, if a driver
receives a message saying that the road ahead
is congested, that driver will probably change
her/his route. In order to study such reactions, an
integrated simulator is needed, one in which the
relationship between the vehicle mobility module
and the network communication module is bidirec-
tional. Choffnes and Bustamante (2005) showed
that the vehicular mobility (traffic) model is very
important, and its integration with the wireless
network model could produce more significant re-
sults. The authors present an integrated simulator
that uses an original vehicular traffic model called
Street Random Waypoint (STRAW) implemented
on top of JiST/SWANS (2008). The authors have
used the simulator to show that studying routing
protocols for a vehicular network without an ac-
curate vehicular traffic model is a wrong approach.
In this respect, they compared their own results
with those obtained with the Random Waypoint
model (Broch, 1998), which is a very inaccurate
representation of a vehicular network.
The mobility model implemented in some
simulators is not a sufficiently accurate repre-
sentation of actual vehicle mobility. For example,
in the model of Saha and Johnson (2004), each
vehicle moves completely independent of other
vehicles, with a constant speed randomly chosen.
Multi-lane roads or traffic control systems are not
taken into consideration. Other authors (Mang-
haram et al., 2005) make similar simplifying as-
sumptions and do not consider multi-lane roads
or car following models. The mobility model of
Choffnes and Bustamante (2005) is more com-
plex: the motion of a vehicle is influenced by the
preceding vehicle, and traffic control systems are
considered. However, multi-lane roads are not
taken into consideration.
A more accurate model has been used in the
development of an integrated vehicular networks
266
Simulation of VANET Applications
simulator, VNSim (Gorgorin, 2006), which in-
cludes a complex model for vehicles mobility, a
wireless network simulator, and an interface for
the emulation of vehicular applications. The simu-
lator can be used to analyze networks of several
thousands of nodes in complex city scenarios as
well as in highway scenarios. The simulator al-
lows the evaluation of a large range of vehicular
computing applications, which cannot be studied
by using other simulators, and can be used to
improve both car-to-car communication protocols
and traffic control applications (Gradinescu, 2007;
Diaconescu, 2007).
GENERral FEAatur of VANET
Simulator
VANET simulators can be classified as microscop-
ic or macroscopic. Microscopic traffic simulators
emphasize local behavior of individual vehicles
by representing the velocity and position of each
vehicle at a given moment. This type of simula-
tion is especially helpful for studying localized
traffic interactions, but it comes with the price
of reduced scalability. Macroscopic simulators
capture traffic conditions in a global manner and
may use concepts from wave theory. Usually, the
traffic is represented in terms of flows (vehicle/
hour), density (vehicles/km) or average speed.
Similarly, microscopic network simulators take
into account the transmission of each individual
packet (or message), while the macroscopic net-
work simulators consider transmission statistical
properties.
Most authors agree that a macroscopic mobil-
ity model is not sufficient to allow the study of a
vehicular network, therefore microscopic simula-
tion, although more complex, is required.
The evaluation of vehicular applications for
online traffic control (that produce real time
feedback to the driver), such as safety applica-
tions, also requires a model of driver behavior.
Models for lane changing or car-following are
Simulation of VANET Applications

some of the features that a microscopic VANET
simulator has to implement to allow the analysis
of such applications.
Simulator Architecture
A VANET simulator has two main components:
a network component, capable of simulating the
behavior of a wireless network, and a vehicular
traffic component, able to provide an accurate
mobility model for the nodes of a VANET. De-
pending on the simulation type, the simulator can
contain other components as well. To describe
them, we refer to the VNSim presented in (Gor-
gorin, 2006).
Figure 1 shows the general view of VNSim.
This is a microscopic discrete-event simulator.
Consequently, it includes components for event
management: an Event Queue, where the events
are registered in the order of their occurrence
times, the modules for events processing, and a
logical clock for the management of the simulation
time. The simulation time is quantified in fixed
pieces so that each event can be associated to a
specific timestamp with an acceptable resolution.
At every moment of the simulation time, the Event
Engine pulls the current events from the queue,
and handles them in a random order by calling
the event processing modules. The vehicular
traffic component is represented by the Mobility
Module, while the wireless network component
is incorporated into the Network Module and
Scheduler.
The Event Queue can hold three types of
events: send, receive or GPS. A send event for
a specified node triggers the call to the node’s
procedure responsible for preparing a message. It
also schedules the corresponding receive event(s)
for the message receiver(s) that are determined by
the simulator according to the wireless network
model. The receive event is associated with a
node (group of nodes) to which the message is
transmitted (broadcasted). Its action is to call
the appropriate handler in each of the receiving

Figure 1. General VNSim architecture

Vehicle 1

Receive Event
Receive Receive Handler Vehicle 2 Vehicle n
Message
Own Data

Send Event Send Handler

Send Prepare message
Engine
Event

Message
GPS data
GPS Event

Vehicular Mobility Model

Network Module and Mobility Module

Scheduler
Event Queue

New Events

3.2s 4 3.3 s .0s VANET SIMULATOR

267

nodes. The GPS event is scheduled at a regular
time interval, for each node, in order to simulate
the way an actual VANET application periodically
collects GPS data. The mobility module repeatedly
updates the position of each node representing a
vehicle, according to the vehicular mobility model.
This model takes into account vehicle interactions
(passing by, car following patterns, etc.), traffic
rules, and the behavior of different drivers.
The main advantage of this architecture is
the possibility to execute (or emulate) the actual
code of a vehicular application without signifi-
cant changes, by using appropriate simulator
interfaces.
In other cases, the vehicular mobility and
wireless network models can be incorporated
in different simulation tools. For example,
MOVE—MObility model generator for VEhicu-
lar networks (2008) uses SUMO—Simulation
of Urban MObility (2008) for simulating a traf-
fic scenario and generating an output file with
vehicular mobility traces. This file is offered as
an input to a network simulator, such as ns-2 or
Qualnet (2008), to obtain the final simulation
results. A similar solution is adopted in (Eichler,
2005), where the authors propose the coupling of
the CARISMA Traffic simulator with ns-2 and
with the application module, in order to study the
vehicle-to-vehicle warning-message distribution.
Synchronization data is exchanged periodically
between the two simulators. The challenge is to
ensure the “coupling of simulators”, i.e. the con-
sistency of vehicle positions in both simulators,
at synchronization points.
The simple architecture mentioned above
makes use of existing traffic and network simu-
lators. The drawback is that it does not support
the simulation of complex VANET applications
in which the driver behavior would be influenced
by the application itself, as already discussed be-
fore. An innovative architecture aiming at solving
this problem is currently under evaluation by the
authors of TraNS (Piorkowski, 2008). They are
268
Simulation of VANET Applications
trying to develop a general framework for con-
trolling the mobility, without being limited to a
specific traffic simulator or to a specific network
simulator. This architecture promises to offer new
possibilities for simulating VANET applications,
but the framework has not yet been released.
A slightly different architectural approach is
the hybrid simulation (Killat, 2007). The simula-
tor architecture incorporates a previously imple-
mented vehicular traffic simulator VISSIM and
two new components: VCOM, used to simulate
the inter-vehicle communication on the basis of a
statistical model proposed by the authors, and an
application module. The simulator has to respond
to specific user requirements regarding the func-
tionality (it was aimed to investigate the impact of
car-to-x communications on transport efficiency)
and the performance (scalability, short response
time, and easy of use). The relevant character-
istics of a VANET simulator are the simulation
efficiency and accuracy for large scale VANETs,
which we will address in a separate section.
Communication Models
Communication models highlight the information
flows between vehicles and are of special interest
when vehicles move with high relative speeds.
VANET applications are affected by wireless
networking aspects such as transmission delay,
packet loss or network access scheme. However,
accurate network simulation introduces additional
complexity and makes several large scale VANET
applications unsuitable for simulation. In this
section, we discuss the network characteristics
that are mostly relevant for VANET simulation,
as well as the accuracy that can be reached with
their implementation.
At the physical layer, the reliability of the wire-
less network, in terms of packet loss and trans-
mission delays, is relevant to simulation. Several
models are used for the radio waves propagation,
which consider estimates of the power of a received
signal. This signal can be received by an antenna
Simulation of VANET Applications
only if the Signal-To-Noise (SNR) ratio is over a
specific SNR Threshold (SNRT).
Wave propagation models have been studied
extensively for decades in telecommunication
and recently have been correlated with vehicular
short range communication. The radio signal can
be degraded by fading. This is caused by signal
propagation over multiple paths (with different
values for the attenuation, delay and phase shift),
which produces interferences at the receiving
antenna.
There are several examples of such propagation
models that were implemented in wireless simula-
tors. Free Space Model is the simplest one, which
considers ideal conditions: a single radio signal
that attenuates with distance. Two Ray Ground
propagation model considers two rays between
the source and the destination: the direct line
and the ground reflection. It gives more accurate
estimates than the free space model, especially
for long distances. VNSim (Gorgorin, 2006)
implements this model and includes the signal
attenuation with distance in order to test different
ways to optimize the network traffic. Rayleigh
fading (Sklar, 1997) models multiple indirect paths
between the sender and the receiver, while the
Ricean fading model considers only one dominant
line of sight and multiple indirect signals. Reports
from real experiments have shown that vehicular
short-range wireless networks can be modeled
using the Nakagami fading model (Sklar, 1997),
which considers fast fading to characterize the
wireless channel.
Concerning the protocols, DSRC—Dedicated
Short-Range Communications Protocol (Taliwal,
2004) is a multi-channel wireless protocol, which
is based on the IEEE 802.11a Physical Layer and
the IEEE 802.11 MAC Layer. It operates over a 75
MHz licensed spectrum in the 5.9 GHz band al-
located by the Federal Communications Commis-
sion, and supports low latency vehicle-to-vehicle
and vehicle-to-infrastructure communications.
The motivation behind the development of DSRC
is mainly the need for a more tightly controlled
spectrum for maximized reliability. The use of
hand-held and hands-free devices on the 2.4 and 5
GHz bands along with the increase of WiFi could
cause intolerable and uncontrollable levels of
interference, which would significantly decrease
the reliability and effectiveness of vehicular safety
applications. But, even with a licensed band, some
issues arise, such as fair access to all applications,
including priority scheduling of traffic between
different application classes (safety over non-
safety). Unlike 802.11, multi-channel coordination
is a fundamental capability of DSRC.
Consistent effort has been put into matching
simulation results to real outdoor experiments.
Yin et al. (2004) observe significant degradation
of the network performance under the high node
mobility. The study evaluated the DSRC Physi-
cal Layer and the MAC Layer, in the context of a
traffic safety application, on top of a customized
simulator combining the CORSIM (2008) traffic
simulator and the Qualnet (2001) network simula-
tor. It showed a long term average throughput of
50-60%. Using the Nakagami m-distribution fad-
ing model implemented in ns-2, Killat et al. (2007)
estimated the probability of successful reception
at a given distance depending on vehicle density
and data load. Their idea is to combine discrete-
event simulation and mathematical modeling,
starting from the observation that the data trans-
mission that is actually interesting and relevant
for a VANET application represents only a small
percentage of the total data exchanged during a
traffic scenario. There is a lot of “background
traffic”, which can be much more efficiently
simulated with statistical methods, while keep-
ing the more accurate discrete-event simulation
techniques only for “interesting” data. Their work
is based on the concept of “hybrid simulation”
and offers much better performance compared
to purely discrete-event approaches. Although
actual outdoor experiments have been made,
the behavior of the wireless network (the DSRC
269

protocol) under stressed city-like scenarios with
high node density is still an open issue. This is
why simulation frameworks now focus on large
scale VANET simulations.
A lot of work has been done to study the
routing layer protocols in mobile ad-hoc net-
works, MANETs. They can be grouped in two
categories. The first category includes protocols
that use the IP addresses to locate nodes, such
as: DSDV—Destination-Sequence Distance
Vector Routing (Perkin, 1994), DSR—Dynamic
Source Routing (Johnson, 2001), AODV—Ad
hoc on demand distance vector (rfc3561, 2008),
and MPR—Mobile Relay Protocol (Nain, 2004).
Protocols in the second category are based on lo-
cation information. Some examples are: DREAM
(Basagni, 1998), GPSR—Greedy Perimeter State-
less Routing (Karp, 2000), and ABRP—Anchor
Based Routing Protocol (Huaizhi, 2005).
Classical MANET protocols have been proven
to perform poorly (Blum, 2004) due to specific
characteristics of vehicular ad-hoc networks,
namely: rapid topology changes, small network
diameter, and limited redundancy. This is why
other position-based forwarding solutions have
been proposed (Festag, 2004) for vehicular en-
vironments, which take into account some ele-
ments of the vehicle mobility model such as the
distance between communicating partners and
the available paths towards destination. This led
to a combined position-based and map-based ad
hoc routing.
Mobility Models
Mobility models emphasize local “behavior” of
individual vehicles by representing the velocity
and position of each vehicle at a given moment.
This type of simulation is especially helpful for
studying localized traffic interactions.
Vehicular traffic mobility models have been
studied for a long time. However, until recently,
the purpose of developing and studying accurate
mobility models has not been to use them for
270
Simulation of VANET Applications
VANET simulation. In this section, we review
vehicular mobility models, from the perspective
of a VANET simulator architect.
Generally, when choosing a mobility model
for a VANET simulator, the complexity and the
accuracy are the most important issues need to be
considered. The benefit of using a very complex
and accurate mobility model, together with the
performance penalty, should be carefully evalu-
ated, keeping in mind the purpose of the VANET
simulator and the type of applications for which
it is designed.
In the Random Waypoint mobility model
(Broch 1998), each mobile node has a position and a
velocity, and moves towards a random destination
in a predictable way that can be calculated as a
function of time. This model was used originally
in ns-2 (2008). It was designed for general-purpose
mobile networks and does not accurately represent
vehicular motion.
Jardosh et al. (2003) define a more complicated
model that corresponds to urban settings, in which
obstacles are introduced to constrain vehicles’
mobility and the wireless transmission. Two other
models have been defined by Potnis and Mahajan
(2006). The Stop Sign Model (SSM) introduces
stop signs at every intersection. Every vehicle
arriving at the intersection would wait for a fixed
period of time before proceeding to its destination.
The Traffic Sign Model (TSM) introduces traffic
lights at every intersection. A vehicle arriving at
an intersection waits for a random period. Both
models introduce queues at intersections and inter-
dependencies among vehicles on the same street.
A thorough analysis of these models is described
in (Mahajan et al., 2007). The authors studied the
dependency of the packet delivery ratio and delay
on the characteristics of the urban settings such
as street layout, traffic rules, etc. In some studies
(Baumann et al., 2008), specific models such as
the Manhattan model and GIS models, are used
to establish the street layouts.
Simulation of VANET Applications
In the model of Saha and Johnson (2004), each
vehicle starts at some randomly selected point
on a road and moves until a destination point
on another randomly selected road. Several new
elements are added to the model and to its use.
First, the path between the starting and destination
points is calculated by using the Dijkstra’s shortest
path algorithm. Second, the vehicle speed is also
randomly calculated using a uniform distribution
within a specified interval. Third, the authors use
the TIGER map database to create a more realistic
environment of the simulation study.
The mobility model of Choffnes and Busta-
mante (2005) is more complex: the motion of a
vehicle is influenced by the preceding vehicle (car
following model), and traffic control systems are
considered to introduce some actual traffic condi-
tions such as traffic congestion. In addition, the
model includes communication details, such as
the percentage of cars actively communicating,
and road maps details, such as characteristics
of street segments and number of lanes in each
direction.
The study of safety and traffic control ap-
plications requires an accurate model of driver
behavior. One example is the model developed
by Wiedemann (1974, 1991) and further studied
by Fellendorf and Vortisch (2001). This psycho-
physical model is used in the commercial traffic
simulator VISSIM (2008). The basic assumption is
that a driver can be in one of four modes, which are
classified according to the driver’s convenience:
free driving (superior convenience), approaching,
following, and braking (lowest convenience). In
the free driving mode there is no influence from
preceding vehicles on the same lane. In this situ-
ation, the driver will seek to obtain and maintain
a “desired” speed. The desired speed and the ac-
celeration depend on driver’s personality, and on
road’s characteristics. In the approaching mode,
there is a slower preceding vehicle that influences
the driver. In this situation, the driver decelerates
in order to obtain the same speed as the preced-
ing vehicle. The deceleration is a function of the
distance between the two vehicles, their speeds, as
well as other parameters. In the following mode,
there is a current and a preceding vehicle, but the
speeds of the two vehicles are practically equal.
In this situation, the driver will seek to keep the
speed constant. In the braking mode, there is a
slower preceding vehicle, very close in front. In
this mode, due to the immediate danger, the driver
will apply high deceleration rates.
This model was implemented in VNSim, to-
gether with a lane-changing model for multilane
roads, based on lane-usage rules valid for several
European countries. According to these rules, the
use of the first lane is required, unless it is oc-
cupied. It means that a driver will always try to
stay on the lower lanes, except when overtaking
another slower vehicle. Overtaking on the right
side is not allowed. Whenever a driver is in a dif-
ferent mode than free driving, she/he will always
check if the higher lane can provide a superior
mode. If this is the case, the driver will switch to
a higher lane. Similarly, whenever a driver is in a
different mode than braking, she/he will always
check if the lower lane provides at least equal
conditions. In this case, the driver will switch to
a lower lane.
Different driver profiles (aggressive, regular,
calm) can easily be modeled by using several
parameters: distance between vehicles, maximum
speed, desired speed, driver reaction time, criti-
cal distance for driver reaction, etc. Each driver
class is represented by a specific set of values for
these parameters. In order to further differentiate
the drivers, the model allows small deviations
from the specified values, which are randomly
computed for each individual driver. Fellendorf
and Vortisch (2001) showed that the model is ac-
curate, by comparing simulated traces with actual
measurement data taken from a German freeway
and from a US freeway. To further calibrate the
model, the authors of VNSim performed an ex-
tra validation, by measuring real-time traffic in
intersections and comparing the results to those
obtained by simulation (Gradinescu, 2007).
271

Estimating fuel consumption and pollutant
emissions is an increasingly important issue when
studying vehicular traffic. The model of Akcelik
and Besley (2003) considers the estimation of the
relation between fuel consumption and emissions,
on one side, and the speed and acceleration of the
vehicle, on the other. A simplified version of this
was used in VNSim to take into account only light
vehicles. Statistics and global measures have been
easily obtained.
Maps
A digital map is required in almost any VANET
application. Each vehicle that is part of the system
should have access to such a digital map. A very
popular option for digital maps used by VANET
simulators is TIGER—Topologically Integrated
Geographic Encoding and Referencing (2008),
which is freely-available, actual digital maps for
the USA. Most of the VANET simulators use
TIGER files as the source for their digital maps.
TIGER files contain detailed geographical infor-
mation about the roads in a region, from large
highways to small streets. For each road, the
TIGER files specify its end points, along with as
many intermediate points as needed, depending
on the road’s shape. Furthermore, for each road, a
“class” information is given (whether it is a small
street, a local road, a State Route, or an Interstate
Highway). Unfortunately, the TIGER database
lacks more detailed traffic-specific information,
like the number of lanes, or traffic control sys-
tems, like traffic lights, yield or stop signs. In
more realistic simulations, such information is
required. The authors of VNSim have chosen to
add the extra information, based on simple heu-
ristics and on the road class information included
in the TIGER files. For example, more lanes are
considered for higher class roads, yield or stop
signs are introduced for lower class roads, traffic
lights are used at intersections between equally
important roads, etc.
272
Simulation of VANET Applications
Implntation i
Approaches to Simulator
Development
The development of VANET simulators follows
two different approaches. One is to implement
the mobility model from scratch. In this situa-
tion, the integration between the mobility model
and the network communication model is ensured
in the design phase. The integration is implicitly
achieved since the developers start with this
integration in mind. On the other hand, the com-
plexity of realistic models for communication and
mobility makes this approach a very challenging
one. There is a huge amount of work required to
implement complex models.
A typical example of this approach is
GrooveSim (Mangharam, 2005), which is a
simulator for geographic routing in VANETs.
It has several modes of operation that allow the
evaluation of network protocols, extraction of
mobile wireless propagation models in real time,
the generation of test scenarios for simulation,
hybrid simulation with real and virtual vehicles,
and others. The most important limitation of
GrooveSim is the simplistic mobility model, which
does not consider car-following or lane-changing
models. The authors acknowledge this limitation
and mention that the focus of their work is to
provide basic models and evaluate their impact
on key performance metrics.
STRAW—STreet RAndom Waypoint (Choff-
nes, 2005) builds a mobility model that constrains
vehicles’ movement according to the map of a city,
different traffic profiles (for example vehicular
congestion) and specific control mechanisms.
It is implemented as an extension to a wireless
network simulator developed by Barr (2004)
at Cornell University. The mobility model of
STRAW is also simplistic, and does not consider
multi-lane roads.
Saha and Johnson (2004) based their simula-
tion on building a realistic mobility model com-
Simulation of VANET Applications
patible with the ns-2 network simulator. Their
development makes use of the TIGER database.
The mobility model which they implement is
very simple, without considering car-following
or lane-changing models. However, their simula-
tions are among the first ones which prove that
using a different mobility model than the well-
known random waypoint model will generate
significantly different results, thus proving that
the mobility model is an important component of
a VANET simulator
The other approach is to use an existing
mobility model implementation and integrate
it with a network simulator. This approach takes
advantage from existing vehicular traffic simula-
tors, which already implement complex, validated
mobility models. The challenge of this approach
is to achieve an online integration, in which the
movement of vehicles is directly influenced by
the VANET applications running on some of
the vehicles.
The easier way for integration is to generate
traffic patterns offline, and then use these results
as an input to a network simulator. VanetMobiSim
(Harri et al., 2006) is a generator of realistic ve-
hicular movement traces for network simulators.
It considers macro-mobility features, such as road
topology, number of lanes, speed limits, etc., and
micro-mobility characteristics, like crossroads
regulated by traffic signs or lane change capabil-
ity for vehicles. Similarly, the Generic Mobility
Simulation Framework—GMSF (Baumann et al.,
2008), which models and simulates node mobil-
ity, is designed to work together with a network
simulator. GMSF has some important distinctive
features: it implements new GIS-based models
with detailed road maps, and includes microscopic
features such as traffic lights management.
An example of using mobility traces is given
in (Yang et al., 2007). The authors propose an
evaluation framework for VANET applications.
It utilizes a hybrid emulation testbed TWINE,
which combines real implementation of high level
network layers (closer to the application) with the
simulation of low level layers (physical, MAC).
The framework offers several advantages to its
users: it allows the execution of real applications
in a simulated network environment; it integrates
realistic mobility traces and deployment data; it
uses evaluation metrics related to the applica-
tion.
A similar approach is described in (Rehuna-
than et al., 2007), where a microscopic traffic
simulator, MITSIMLab is used to generate vehicle
movement traces to be used in ns-2. This approach,
however, does not allow for an accurate evaluation
of a VANET application or protocol, because the
mobility pattern cannot be modified as a reaction
to the VANET application.
Using an existing traffic simulator, which
is based on already validated complex traffic
models, has the important advantage of avoiding
“re-inventing the wheel”. Among the most recent
attempts to implement a VANET simulator by
using this approach are MOVE (Karnadi et al.,
2007) and TraNS—Traffic and Network Simulator
(Piorkowski et al., 2008). Both use the open-source
microscopic traffic simulator SUMO (2008),
which implements complex validated vehicular
traffic mobility models. And both use two steps
for the VANET simulation. In the first step, traf-
fic traces are generated offline using the traffic
simulator. In the second step, a network simula-
tor processes the traffic traces and evaluates the
network communication between the nodes. The
authors of both simulators agree that the approach
does not allow the simulation of complex VANET
applications, in which the driver behavior would
be changed by the application itself. However,
both developer teams mention that this is a goal
for future work that will achieve a real integration
between the traffic mobility model and the network
communication model. Furthermore, the authors
of TraNS claim that their simulator is “the first
open source project that attempts to realize this
highly pursued coupling for application-centric
VANET evaluation” (Piorkowski et al., 2008).
273
 Simulation of VANET Applications

Sommer et al. (2008) describe the simulation simulator with a complex node mobility model,
framework Veins (Vehicles in Network Simula- which is responsible for accurately computing the
tion) that is based on the bidirectional coupling motion of all nodes, in each time quantum.
of OMNeT++, a network simulator capable of The graph in Figure 2 shows the evolution of
modeling realistic communication patterns of the simulator’s performance with the increase
VANET nodes, and of SUMO, the microscopic of the number of nodes. The results are based
traffic simulator. The bidirectional coupling is on simulations performed in a highway traffic
able to support the analysis of more realistic traf- scenario, with all the nodes running a neighbor-
fic models that reflect the impact of inter vehicle discovery and update protocol, with a 1 second
communication on the road traffic. period for the beacon messages.
VNSim is able to simulate motion of 1000
S vehicles, in a complex city-scenario in real-time
(1 second of simulation in 1 second of real time).
Accurate simulation of vehicles’ mobility and Experiments have been done by the authors with a
wireless transmission is time-costly. Also, for an squared region of 1km by 1km in Manhattan down-
application designer, the study of many applica- town, which has a high number of intersections
tion scenarios requires experiments with large and traffic lights (see Figure 3). VNSim was also
maps and with large numbers of vehicles. This used to perform the simulation of more complex
is why the scalability of VANET simulators is a scenarios, involving up to 10.000 vehicles.
main performance metric for which we present NCTuns (Wang, 2008), a highly-integrated
some guiding figures. simulator, cannot be used for large-scale simula-
Experiments performed with VNSim on a tions. Thus, for the simulation of a scenario with
1.6GHz uni-processor showed that it can support 100 vehicles, 22 minutes are required. For the
around 10.000 network events per second. This same scenario, if the number of vehicles doubles
value is lower than the throughput of the widely- then the amount of time required for the simula-
used network simulator ns-2, which can simulate tion grows approximately 4 times, to 84 minutes.
over 60.000 events per second, on a 2GHz uni- These results, obtained on a 1.8 GHz uniprocessor,
processor. This is explained by the of the network show that the simulation of large-scale vehicular

Figure 2 VNSim performance evolution with the Figure 3 Squared region in Manhattan down-
increase of the number of nodes town

500

450
sIMULATe A 300 seCONDs sCeNARIO
NUMbeR OF seCONDs NeCessARY TO

400

350

300

250

200

150

100

50

0
250 500 750 1000 1250 1500 1750 2000
NUMb eR OF NODes

274
Simulation of VANET Applications
networks (with tens of thousands of vehicles) can
be achieved.
Some performance results are presented in
(Choffnes, 2005) for the STRAW simulator. Here,
STRAW’s overhead has been evaluated for two
models: a simple one in which vehicle movement
is determined randomly at each intersection, and a
more complicated one in which the movements are
determined according to pre-computed shortest
paths. The evaluation was done by disabling the
wireless communication simulator. The experi-
ments were performed on a Pentium 4 at 2.4GHz
computer with HyperThreading enabled. When
the number of nodes considered in the experi-
ment increased from 100 to 1600, the runtime
increased four times for the simple model and
6.5 times for the more complicated model. The
highest overhead of 10% of the simulation time
has been registered for the complex model with
1600 vehicles.
One solution to improve simulators’ perfor-
mance is parallelization. Fujimoto et al (2003)
report experiments performed on packet level
network simulators. Parallelization is achieved by
federating several copies of the same simulator or
of different simulators. This approach was used to
produce two parallel simulators, PDNS—Parallel
/ Distributed Network Simulator (PDNS) based
on ns-2 and GTNetS developed at Georgia Tech.
The results showed a very good performance
measured in number of transmissions per second,
PTS—Packet Transmissions that can be simulated
per Second. More specific, for the best case of
parallel simulations, the number of PTS was 50
times higher than the number of PTS obtained in
the best sequential case.
MoVES, a Mobile Wireless Vehicular Envi-
ronment Simulator (Bononi et al., 2006) is also
parallel and distributed. Its idea is to distribute
subsets of the model entities to different proces-
sors that communicate via message passing.
The authors find that the parallel approach gives
good results when simulating massive wireless
vehicular networks.
Higher performance can also be achieved by
the use of hybrid simulators (Killat, 2007). The
authors present the results obtained for 2.500
to 3.000 vehicles in a scenario dedicated to the
analysis of the impact of the penetration rate of
radio-equipped vehicles on the average speed.
The speedup factor was 500 as compared to the
NS-2 simulation. The result is due to the use of
the hybrid simulation concept, which combines
the discrete-event simulation with analytical
traffic modeling.
As a general conclusion, developing a highly-
integrated simulator, with complex models for both
the vehicular mobility and the wireless commu-
nication has the advantage of higher accuracy but
has a severe impact on performance. Integrating
existing vehicular traffic simulators and network
simulators can lead to better performance but it
is not able to support more complex scenarios
involved in safety and traffic control applications.
Improvements can be achieved by parallelization,
which distributes the simulation task to different
computing resources, or by selecting a model that
reduces the computation load.
Simulation of rntati
city and highway application
SCENnario
This section presents several simulation use
cases together with results obtained in realistic
simulation experiments performed with the VN-
Sim simulator. The way these results could be
exploited by VANET developers and users are
also highlighted.
An important class of applications that can be
deployed over VANETs is concerned with traffic
operations and maintenance: dynamic route plan-
ning, weather conditions publishing and adaptive
signal control in intersections.
275

Traffic Coordination in Intersections
Traffic coordination in intersections is a very
challenging and not enough studied topic. As
an example of a VANET application, we pres-
ent an adaptive traffic light system based on
wireless communication between vehicles and
fixed controller nodes deployed in intersections
(Gradinescu et al., 2007). We describe the steps
required to deploy such an application in VNSim.
Furthermore, we present the results showing sig-
nificant improve of traffic fluency in intersections,
and the advantages over other architectures in
terms of cost and performance.
The goal of the VNSim project was to reduce
congestion and the number of accidents in inter-
sections by enabling wireless communication
between traffic lights and vehicles. In this way,
the traffic lights can get a real-time “image” of
the traffic, and make an informed decision on how
to set signal plans. As a secondary benefit, the
traffic lights can also detect dangerous situations
and alert drivers about the intersection status to
increase their awareness.
The following aspects need to be taken into
consideration when such an application is simu-
lated. First, there must be a data dissemination
system that supports the information flow from
vehicle to vehicle and to the traffic light. Based
on the TrafficView system (Dashtinezhad et. al
2004), data dissemination was designed so that
vehicles forward information about other vehicles
they know about (position, speed, direction) to
all the other vehicles in their range.
Figure 4. Traffic lights and vehicles communicate
to improve traffic efficiency
276
Simulation of VANET Applications
The simulation of this propagation scheme
in VNSim demonstrated that a lot of redundant
data is transmitted and a high data charge is
put on the network. Moreover, for large packet
size, the probability of the successful transmis-
sions decreases due to wireless interferences. A
solution is the use of aggregation mechanisms.
A possible method is the use of a probabilistic
scheme in which only some “leader” vehicle,
chosen randomly with a probability depending
on the traffic density, transmit the information.
The simulation experiments highlighted the great
decrease in the network load while preserving the
functionality of the solution.
The data dissemination algorithm supports
the implementation of the traffic light control
that establishes a signaling plan well suited for
the intersection. Considering the traffic light is a
node participating in the dissemination process, it
can gather traffic metric values (flows, delays) on
all the concurrent roads. Optimum cycle length
and green split phases are computed using traffic
engineering methods, such as Webster’s formula
(Gradinescu et. al 2007).
One of the simulation scenarios that were used
to analyze the efficiency of the algorithm was a
4-way intersection in Bucharest during rush hour.
As shown in Figure 5, the intersection recovered
sooner from congestion when using the adap-
tive strategy described above, compared to the
pre-timed system deployed in the intersection.
The total delay measured for all vehicles was
decreased by 28%.
Figure 5. Average control delay in intersection
Average Control Delay (s)
500
400
300
200
100
0
5
15
25
35
45
55
65
75
85
95
5
5
5
10
11
12
13
14
s imulation Time (min)
P re-tim ed A daptive - Ideal A daptive - Real
Simulation of VANET Applications
An additional component of the VNSim
simulator is the module for estimating the pol-
lutant emissions and fuel consumption, which
considers the particular behavior of each vehicle
speed and acceleration. This feature was used in
this application to demonstrate the benefits of the
adaptive solution due to increased traffic fluency:
6.5 % improvements in fuel consumption and CO2
emissions were measured.
Collaborative Route Computation
Finding the best route to destination in congested
roads traffic is a major challenge in the attempt
to improve traffic conditions. Collaborative route
planning is the second VANET application we
use to show how ad-hoc communications can
improve the driving experience.
The simulation of a collaborative, dynamic
routing application requires several features
from the simulation tool. First, in order to study
a congested roads network, a large number of
vehicles, moving on a large map, should be sup-
ported. Therefore, the simulator’s performance
and scalability are crucial. Furthermore, this kind
of application supposes that the mobility of the
vehicles is influenced by the application itself:
if drivers find out, by using the VANET route
planning application, that a road is less congested,
then they might change their initial path in order
to take advantage of this information. A simulator
which does not integrate the vehicular mobility
module with the application module will not be
able to represent dynamic route changes.
Diaconescu et al. (2007) used VNSim in order
to study the dynamic route computation problem.
The authors discuss two different approaches. The
first is to exclusively use the ad-hoc network for
all the information exchanges. The second ap-
proach is based on the existence of fixed nodes,
placed in intersections, which also participate to,
and facilitate the communication.
In the first approach, each driver first com-
putes a route towards its destination. This initial
computation is based on distance, on the roads’
types, and on the assumption that drivers tend to
prefer the main roads. After this initial computa-
tion, the driver starts the journey. Along the way,
periodically, queries about the traffic conditions
ahead are sent via the ad-hoc network. Depend-
ing on the replies (for example, the reception of
information about traffic jams ahead), re-calcula-
tions of the route occur.
The simulation experiments showed that global
improvements in travel times occur when only a
percentage of the vehicles use the algorithm, not all
of them. When all vehicles change their route, the
traffic jam “moves” to another road. Starting from
this observation, the authors suggested the use of
fixed-nodes placed in intersections, which are able
to communicate with vehicles and, at the same
time, are inter-connected through a separate high-
speed network. Of course, this solution requires
a more extensive infrastructure. In large cities, a
large part of this infrastructure might already be
present. The fixed-nodes gather information about
the actual traffic (number of vehicles traveling)
on each road and execute a collaborative route
planning with the other fixed nodes. The results
are sent to the vehicles in proximity informing
them about the best routes.
Highway Lane Reservation
An innovative VANET application was proposed
in (Ravi et al., 2007). The authors argue that
time guarantees of other means of transportation
(like train or plane) can be incorporated into the
highway system. The idea is to allow drivers to
reserve an “entry slot” onto dedicated lanes of a
highway by paying a premium price. Reservations
will be allowed for these lanes up to their carrying
capacity, so that the dedicated lanes will never
become jammed and the system can guarantee
the trip duration between any two highway points,
in the absence of accidents.
The general architecture of the system is shown
in Figure 6. The proposed system is complex and is
277

Figure 6. Lane reservation general architecture
made up of several subsystems. First, a reservation
system is required to allow users make advance
reservations of time-slots on the dedicated lanes.
Then, an enforcement system is needed to ensure
that only drivers with valid reservations travel on
the dedicated lanes. This could be achieved by
using RFID tags located on vehicles, and RFID
readers placed on the highway.
The lane entrance must be assisted by a
VANET. More specific, since the dedicated lane
will often be used at near-to-maximum capacity,
it is important that vehicles that enter or leave the
lane do not create disruptions in the traffic flow
on this lane. In order to allow smooth entering
and leaving, a VANET protocol can be designed.
For example, the car that wants to enter the lane
will send a request to a vehicle already on the
lane, asking for permission to enter.
In order to evaluate such a large-scale system,
a proper simulation tool is needed. It should be
scalable enough to be able to simulate a large-scale
highway traffic scenario. Furthermore, emulating
the application code run by each node would also
be required. The authors plan to investigate the
possibility of using VNSim in order to study and
validate a lane reservation system.
278
Simulation of VANET Applications
Concluion
This chapter discussed the simulation of VANET
applications. The focus was on the new trends
in communication protocols and traffic models,
and on new facilities incorporated in simulation
tools. The advances in the VANET technology and
protocols support the adoption and use of more
complex mobility models and of more flexible and
adaptable traffic controls. VANETs have specific
characteristics such as the rapid topology changes
or the changes in the vehicles mobility as reac-
tion to traffic changes. Obviously, these changes
must be captured by the simulation models, which
become more or less complicated and include
more elements that constrain the vehicle mobility:
maps, real traffic conditions (congestion), driver
behavior, fuel consumption, pollutant emissions,
etc. Recent research referenced in this chapter
targeted the validation of these models and the
exploitation of their facilities in different applica-
tions. Realistic models, such as those used in lane
reservation, were proposed and will eventually
constitute the subject of further research.
In order to more accurately evaluate the traffic
control applications, more powerful microscopic
simulator tools have been developed. The tools
integrate the network communication model and
the traffic / mobility model, and facilitate the
interfacing with the core application modules.
One drawback is the high time consumption re-
quired by simulation experiments besides other
aspects like usability and flexibility. Since micro-
scopic simulators demonstrated their utility in the
VANET applications design process, the increase
of tool performance, scalability, and flexibility
will be subjects to future research work.
The technological progress will stimulate
new applications aiming to improve the traffic
safety and control strategies. More vehicles will
be equipped with GPS receivers, communication
devices compatible with different protocols (for
example DSRC), radars for identifying neighbors’
positions, equipment for storing digital maps,
Simulation of VANET Applications
etc. These and other similar components will
facilitate the development of more complex safety
applications (like imminent collision warning,
obstacle detection, collision avoidance, etc.) and
traffic control applications (like traffic congestion
avoidance, traffic lights control, route computa-
tion, etc.). Obviously, the role of simulation in
the evaluation of different solutions will become
even more important.
REFERENCES
Akcelik, R., & Besley, M. (2003, December).
Operating cost, fuel consumption, and emission
models in aaSIDRA and aaMOTION. Paper
presented at the 25th Conference of Australian
Institute of Transport Research. Univ of South
Australia, Adelaide.
Barr, R. (2004). An efficient, unifying approach
to simulation using virtual machines. PhD thesis,
Cornell University, US.
Basagni, S., Chlamtac, I., Syrotiuk, V. R., &
Woodward., B. A. (1998). A distance routing effect
algorithm for mobility (DREAM). In Proceedings
of MobiCom ‘98 (pp. 76-84). Dallas, TX.
Baumann, R., Legendre, F., & Sommer, P. (2008).
Generic Mobility Simulation Framework (GMSF).
In Proceeding of the 1st ACM SIGMOBILE
workshop on Mobility models (pp 49-56). Hong
Kong, China
Bononi, L., Di Felice, M., Bertini, M., & Croci,
E. (2006). Parallel and Distributed Simulation of
Wireless Vehicular Ad Hoc Networks. In Proceed-
ings of the 9th ACM international symposium on
Modeling analysis and simulation of wireless and
mobile systems (pp 28-35). Terromolinos, Spain
Broch, J., Maltz, D. A., Johnson, D. B., Hu, Y. C., &
Jetcheva, J. G. (1998). A Performance Comparison
of Multi-HopWireless Ad Hoc Network Routing
Protocols. In Proceedings of MobiCom’98 (pp.
85-97). Dallas, TX.
Choffnes D. R., & Bustamante F. E. (2005).
An Integrated Mobility and Traffic Model for
Vehicular Wireless Networks. In Proceedings
of the 2nd ACM International Workshop on Ve-
hicular Ad Hoc Networks (pp. 69—78). Cologne,
Germany.
CORSIM (2008). Microscopic Traffic Simulation
Model. Retrieved March 19, 2008, from http://mc-
trans.ce.ufl.edu/featured/tsis/version5/corsim50.
htm
Dashtinezhad, S., Nadeem, T., Dorohonceanu,
B., Borcea, C., Kang, P., & Iftode L. (2004).
TrafficView: A driver assistant device for traffic
monitoring based on car-to-car communication,
In Proceedings of IEEE Semiannual Vehicular
Technology Conference. Milan, Italy.
Diaconescu, R., Gorgorin, C., Gradinescu, V.,
Cristea, V., & Iftode, L. (2007). Collaborative
Route Planning in Vehicular Ad-hoc Networks.
In Proceedings of the CSCS-16 (pp. 267-274).
Bucharest, Romania.
Eichler, S., Ostermaier, B., & Schroth, C. (2005)
Simulation of Car-to-Car Messaging: Analyzing
the Impact on Road Traffic. In Proceedings of the
13th IEEE International Symposium on Model-
ing, Analysis, and Simulation of Computer and
Telecommunication Systems (pp. 507-510). IEEE
Computer Society  Washington, DC, USA
Fellendorf, M., & Vortisch, P. (2001, January).
Validation of the Microscopic Traffic Flow Model
VISSIM in Different Real-World Situations. Paper
presented at the 80th Meeting of the Transporta-
tion Research Board. Washington, D.C.
Fujimoto, R. M., Perumalla, K., Park, A., Wu, H.,
Ammar, M. H., & Riley, G. F. (2003) Large-Scale
Network Simulation: How Big? How Fast? In
Proceedings of the 11TH IEEE/ACM International
Symposium on Modeling, Analysis and Simulation
of Computer Telecommunications Systems (pp.
116-123). IEEE Computer Society Washington,
DC, USA
279
 Simulation of VANET Applications

Gorgorin, C., Gradinescu, V., Diaconescu, R., and Networking Conference, WCNC 2007. IEEE
Cristea, V. & Iftode, L. (2006). An Integrated (pp. 2506-2511). Kowloon, Hong Kong.
Vehicular and Network Simulator for Vehicular
Karp, B., & Kung, H. T. (2000). GPSR: Greedy
Ad-hoc Networks. In Proceedings of the Eurosis
perimeter stateless routing for wireless networks.
European Simulation and Modeling Conference
In Proceedings of the 6th annual international
2006 (pp. 142-149). Toulouse, France.
conference on Mobile computing and network-
Gradinescu, V., Gorgorin, C., Diaconescu, R., ing (pp. 243-254). Boston, Massachusetts, United
Cristea, V., & Iftode, L. (2007). Adaptive Traf- States.
fic Lights Using Car-to-Car Communication. In
Katz, J. H. (1963). Simulation of a Traffic Network.
Proceedings IEEE 65th Vehicular Technology
Communications of the ACM, 6(8), 480-486
Conference (pp. 21-25). Dublin, Ireland.
Killat, M., Schmidt-Eisenlohr, F., Hartenstein,
Haerri, J., Filali, F., Bonnet, C., & Fiore, M. (2006).
H., Rössel, C., Vortisch, P., Assenmacher, S., &
VanetMobiSim: Generating Realistic Mobility
Busch, F. (2007). Enabling efficient and accurate
Patterns for VANETs. In Proceedings of the 3rd
large-scale simulations of VANETs for vehicular
international workshop on Vehicular ad hoc net-
traffic management. In Proceedings of the Fourth
works (pp 96—97). Los Angeles, CA, USA
ACM international Workshop on Vehicular Ad
Huaizhi, Li, & Singhal, M. (2005) A scalable Hoc Networks, VANET ‘07 (pp. 29-38). Montreal,
routing protocol for ad hoc networks. In Pro- Canada.
ceedings of the Vehicular Technology Confer-
Krajzewicz, D., Bonert, M., & Wagner, P. (2006).
ence 2005 Spring. 2005 IEEE 61st Volume 4 (pp.
The Open Source Traffic Simulation Package
2498—2503). Stockholm, Sweden
SUMO. Presented at RoboCup 2006 Infrastructure
Jardosh, A., Belding-Royer, Elizabeth M., Alm- Simulation Competition. Bremen, Germany.
eroth, K. C., & Suri, S. (2003). Towards realistic
Mahajan, A., Potnis, N., Gopalan, K., & Wang,
mobility models for mobile ad hoc networks. In
A. (2007). Modeling VANET Deployment in
MobiCom ’03: Proceedings of the 9th annual
Urban Settings. In Proceedings of the 10th ACM
international conference on Mobile computing
Symposium on Modeling, analysis, and simula-
and networking, (pp. 217—229). ACM Press,
tion of wireless and mobile systems (pp 151-158).
New York, USA.
Chania, Crete Island, Greece.
Jist/SWANS (2008). Java in Simulation Time /
Mangharam, R., Weller, D. S., Stancil, D. D., Ra-
Scalable Wireless Ad hoc Network Simulator.
jkumar & R., Parikh, J. S. (2005). GrooveSim: A
Retrieved March 19, 2008, from http://jist.ece.
Topography-Accurate Simulator for Geographical
cornell.edu/index.html
Routing in Vehicular Networks. In Proceedings
Johnson, D. B., Maltz, D. A. & Broch, J. (2001). of the 2nd ACM International Workshop on Ve-
DSR: The Dynamic Source Routing Protocol for hicular Ad Hoc Networks (pp. 59-68). Cologne,
Multi-Hop Wireless Ad Hoc Networks. In C. E. Germany
Perkins (Ed.), Ad Hoc Networking (pp. 139-172).
MOVE (2008). MObility model generator for
Addison-Wesley, 2001
VEhicular networks. Retrieved March 19, 2008,
Karnadi, F. K., Zhi, H. M., & Kun-chan, L. from http://www.csie.ncku.edu.tw/~klan/move/
(2007). Rapid Generation of Realistic Mobility index.htm
Models for VANET. Wireless Communications

280
Simulation of VANET Applications
ns-2 (2008). The ns-2 wiki. Retrieved March 18,
2008 at http://www.isi.edu/nsnam/ns/doc/
Petigara, N. D., & Balakrishnan, H. (2004).
Integrated routing and storage for messaging
applications in mobile ad hoc networks. Mobile
Networks and Applications, 9(6), 595-604.
Perkins, C., & Bhagwat, P. (1994). Highly Dy-
namic Destination-Sequence DistanceVector
Routing (DSDV) for Mobile Computers. ACM
SIGCOMM Computer Communication Review,
24(4), 234-244.
Piorkowski M., Raya M., Lugo A., Papadimitra-
tos P., Grossglauser M., & Hubaux J. P. (2007,
September). TraNS: Realistic Joint Traffic and
Network Simulator for VANETs. Short paper
presented at the Fourth ACM international Work-
shop on Vehicular Ad Hoc Networks, VANET ‘07.
Montreal, Canada.
Potnis, N., & Mahajan, A. (2006). Mobility Models
for Vehicular Ad Hoc Network Simulations. In
Proceedings of the 44th ACM Southeast regional
conference (pp 746-747). Melbourne, Florida.
Qualnet (2001). Version 3.1 User’s Manual. Re-
trieved March 19, 2008 from http://linux.ucla.
edu/~gavin/papers/SimulatorManual-3.1.pdf
Ravi, N., Smaldone, S., Iftode, L., & Gerla, M.
(2007). Lane Reservation for Highways (Position
Paper). Retrieved March 19, 2008 from www.
cs.rutgers.edu/~iftode/ILR-ITSC07.pdf
Rehunathan, D. B., Seet, B. C., & Luong, T. T.
(2007). Federating of MITSIMLab and ns-2 for
realistic vehicular network simulation. In Pro-
ceedings of the 4th international conference on
mobile technology, applications, and systems and
the 1st international symposium on Computer
human interaction in mobile technology (pp 62-
67). Singapore
rfc3561 (2008). Ad hoc On-Demand Distance Vec-
tor (AODV) Routing. Retrieved March 19, 2008
from http://www.ietf.org/rfc/rfc3561.txt
Saha, A., & Johnson, D. (2004). Modeling Mobil-
ity for Vehicular Ad-hoc Networks. In Proceed-
ings of the 1st ACM International Workshop on
Vehicular Ad Hoc Networks (VANET) (pp. 91-92).
Philadelphia PA, USA.
Sklar, B. (1997). Rayleigh Fading Channels in
Mobile Digital Communication Systems Part I:
Characterization. IEEE Communications Maga-
zine, 35(7), 90-100
Sommer, C., Yao, Z., German R., & Dressler, F.
(2008) On the Need for Bidirectional Coupling
of Road Traffic Microsimulation and Network
Simulation. In Proceedings of the Computer
Communications Workshops. INFOCOM (pp
1-6).Phoenix, AZ.
SUMO (2008). Simulation of Urban MObility.
Retrieved March 19, 2008 from http://sumo.
sourceforge.net/docs/bibliography.shtml
Taliwal, V., Jiang, D., Mangold, H., Chen, C., &
Sengupta, R. (2004). Empirical determination
of channel characteristics for DSRC vehicle-to-
vehicle communication. In Proc. 1st ACM Int.
Workshop on Vehicular Ad Hoc Networks (p. 88).
Philadelphia PA, USA.
TIGER (2008). Topologically Integrated Geo-
graphic Encoding and Referencing system. Re-
trieved March 19, 2008 from http://www.census.
gov/geo/www/tiger/
VISSIM (2008). VISSIM Micro-Simulation Soft-
ware. Retrieved March 19, 2008 from http://www.
english.ptv.de/cgi-bin/traffic/traf_vissim.pl
Wang, S. Y., Chou, C. L., Chiu, Y. H., Tseng, Y.
S., Hsu, M. S., Cheng, Y. W., Liu, W. L., & Ho,
T. W. (2007, September). NCTUns 4.0: An Inte-
grated Simulation Platform for Vehicular Traffic,
Communication, and Network Researches. Paper
presented at the 1st IEEE International Sympo-
sium on Wireless Vehicular Communications.
Baltimore, MD, USA
281

Wiedemann, R. (1974) Simulation des Stras-
senverkehrsflusses. Schriftenreihe des Instituts
für Verkehrswesen der Universität Karlsruhe,
Heft 8
Yang, Yi, Varshney, M., Mohan, S., & Bagrodia,
R. (2007). High-Fidelity Application-Centric
Evaluation Framework for Vehicular Networks.
In Proceedings of the fourth ACM international
282
Simulation of VANET Applications
workshop on Vehicular ad hoc networks (pp 49-
58). Montreal, Quebec, Canada
Yin, J., ElBatt, T., Yeung, G., Ryu, B., Habermas,
S., Krishnan, H., & Talty, T. (2004). Performance
evaluation of safety applications over DSRC ve-
hicular ad hoc networks. In Proceedings of the
1st ACM international Workshop on Vehicular Ad
Hoc Networks (pp. 1-9). Philadelphia PA, USA
 283

Chapter XV
In-Vehicle Network Architecture
for the Next-Generation
Vehicles
Syed Masud Mahmud
Wayne State University, USA

ABSTRACT

New types of communication networks will be necessary to meet various consumer and regulatory
demands as well as satisfy requirements of safety and fuel efficiency. Various functionalities of vehicles
will require various types of communication networks and networking protocols. For example, drive-
by-wire and active safety features will require fault tolerant networks with time-triggered protocols
to guarantee deterministic latencies. Multimedia systems will require high-bandwidth networks for
video transfer, and body electronics need low-bandwidth networks to keep the cost down. As the size
and complexity of the network grows, the ease of integration, maintenance and troubleshooting has
become a major challenge. To facilitate integration and troubleshooting of various nodes and networks,
it would be desirable that networks of future vehicles should be partitioned, and the partitions should
be interconnected by a hierarchical or multi-layer physical network. This book chapter describes a
number of ways using which the networks of future vehicles could be designed and implemented in a
cost-effective manner. The book chapter also shows how simulation models can be developed to evalu-
ate the performance of various types of in-vehicle network topologies and select the most appropriate
topology for given requirements and specifications.

Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.
 In-Vehicle Network Architecture for the Next-Generation Vehicles

INTRODUCTION body electronics. This is the reason why currently

Early vehicles used dedicated point-to-point con-
nections for inter-module communications. As
the number of modules and features increased
in vehicles, the wiring system became bulky,
complex, expensive and difficult to install and
maintain. As a result, over the years, evolution of
in-vehicle communications via a serial bus took
place. A serial bus can replace all the dedicated
point-to-point wiring between modules. Thus, it
significantly reduces wiring complexity as well
as weight of the vehicle making the vehicle more
fuel efficient. A serial bus system is also scalable,
meaning that more modules can be connected to
the bus at any time without requiring any changes
at other modules. Since different functions of a
vehicle need different data rates, such as power-
train needs higher data rate than body electronics,
the use of a single serial bus for the entire vehicle
may not be the best choice to design the in-vehicle
communication network. As high-speed electronic
components are more expensive than low-speed
electronic components, it would be desirable and
cost effective to partition the serial bus into two
buses: a high-speed serial bus and a low-speed
serial bus. The high-speed bus can be used for
power-train and the low-speed bus can be used for
most vehicles have at least two serial buses: one for
power-train and the other for body electronics, as
shown in Figure 1. Since there are some mutually
required data between the two partitions of the in-
vehicle network, the partitions are interconnected
by a gateway device as shown in Figure 1. The
two partitions need not use the same type of bus
though they can. Since the LIN (Local Intercon-
nection Network) bus is less expensive and LIN
protocol (Uplap, 2004) is less complex than the
corresponding CAN (Controller Area Network)
bus and CAN protocol (Motorola,1998), nowadays
many vehicles are considering using a low-speed
LIN bus for body electronics and a high-speed
CAN bus for power-train. The gateway device
which interconnects a CAN bus and a LIN bus
converts messages from one protocol to another
protocol when the messages need to go from one
partition to another partition.
As the government and consumer demands
are increasing for many features such as safety,
fuel efficiency, comfort, navigation, entertain-
ment, telematics, multimedia and many more,
future vehicles will have many more partitions.
Each partition will take care of a particular
type of application. All the partitions will need
to be interconnected for exchanging mutually

Figure 1. An in-vehicle networking system with two partitions

Partition-1: Power-Train

Anti-Lock Brakes Module Engine Module High-Speed

CAN bus

Transmission Module Accident Avoidance Module

Gateway
Partition-2: Body Electronics Module

Instrument Module Door Module

Low-Speed CAN
Climate Control Module Lights Module or LIN Bus

284
In-Vehicle Network Architecture for the Next-Generation Vehicles
required information as well as for diagnostic
purposes. Like early vehicles, dedicated point-
to-point connections among various partitions
will increase the complexity of connectivity, as
shown in Figure 2. Figure 2 shows an in-vehicle
networking system with four partitions. Each
has its own bus system. Each pair of partitions is
interconnected by a gateway device. For example,
Partition-1 and Partition-2 are connected by the
gateway device G12. Similarly, Partition-1 and
Partition-4 are connected by the gateway device
G14. The number of gateway devices increases
exponentially with the increase in the number of
partitions. For N partitions, the number of gate-
way devices required for dedicated connections
N!
among the various partitions is . Thus,
2 * ( N − 2)!
dedicated connections among the partitions will
be very complex and cost prohibitive.
As the number of partitions in a vehicle is
going to increase, the automakers must come up
with less complex and less expensive techniques
for interconnecting the partitions.
The main objective of this book chapter is
to show various cost-effective interconnection
mechanisms among the partitions of an in-vehicle
networking system. The advantages, disadvan-
tages and the feasibility of various interconnection
Figure 2. Dedicated connections among four partitions of an in-vehicle networking system.
Partition-1
G12
G14
G13
G34
Partition-3
mechanisms using the current technology are also
explained in this book chapter.
BACKGROUND
There is little doubt that the automobile is the
most revolutionary invention in the history of
transportation since the wheel. Over the last 100
years, automobiles have gone through numer-
ous phases of improvements. At the dawn of the
twentieth century, transportation of people and
goods in timely and cost effective manner was the
main motivation behind developing automobiles.
Safety, fuel efficiency, comfort, entertainment,
green house effect, etc. were not the major issues
for those early automobiles. Thus, the first genera-
tion automobiles were mostly mechanical devices,
and they had very few electrical and electronic
components. As a result, interconnecting those
few electrical and electronic components was not
a major problem. After that, various new features
were being added to automobiles out of necessity.
Initially, the necessity of safety and fuel efficiency
was the main driving force behind improvements
of automobiles.
Since the 1960’s, there have been significant
improvements in vehicle safety. The introduction
Partition-2
G23
G24
Partition-4
285
 In-Vehicle Network Architecture for the Next-Generation Vehicles
of safety features such as seat belts, air bags,
crash zone, lighting system and new vehicle
structures has dramatically reduced the rate of
crashes, injuries and fatalities. However, in spite
of these impressive improvements, according to
U.S. Department of Transportation there were
nearly 45,000 fatalities in transportation accidents
in the United States in 2004, of which 95 percent
involved highway motor vehicles (USDOT, 2006).
By continuing with the passive safety technolo-
gies, it is difficult to achieve significant further
gains in reducing crash costs. Thus, engineers
and researchers have started developing active
safety technologies to further reduce the costs
involved with vehicle crashes. As a result, more
electrical and electronic components are being
added to vehicles.
The energy crisis pushed the automotive
industries to make their vehicles more fuel ef-
ficient. Fuel efficiency requires efficient burning
of fuel inside engine cylinders which eventually
requires optimal air-fuel ratio, optimal spark
timing and optimal control of valves. As a result,
carburetors were replaced by computer controlled
electronic fuel injection systems; spark timing is
controlled by computers; mechanical valves are
being replaced by electrical valves; and many
more, which eventually added more electrical
and electronic components in vehicles.
As the electrical and electronic components
grew in vehicles, the dedicated point-to-point
interconnection among these components became
very bulky, complex and expensive. As a result,
the concept of multiplexing or in-vehicle network-
ing evolved since early 1980’s. In fact, one of the
first automotive papers that addressed the need
for vehicle multiplexing was published as early as
1976 (Bell, 1976). Multiplexing uses a serial bus,
and all electrical and electronic components are
connected to this bus via microcontrollers. The
microcontrollers take active role in exchanging in-
formation among various electrical and electronic
components of the vehicles. The microcontrollers
send information using messages. The messages
286
are created by transmitting microcontrollers and
decoded by receiving microcontrollers based on
a particular communication protocol used by
vehicles. In the 1980’s, automotive companies
were developing their own protocols and there
was no standardization in networking protocols
among various companies. As a result, jobs of
chip manufacturers, suppliers and mechanics
were becoming difficult because chip manufac-
turers had to make different chips for different
protocols, suppliers had to make different parts
to make them compatible with different protocols
and the mechanics had to use different diagnostic
computers for different protocols. Recently almost
all automotive companies both in US and Europe
have started using Controller Area Network
(CAN) protocol for in-vehicle networking. Thus,
nowadays almost every high end vehicle has at
least one CAN network.
The Society of Automotive Engineers (SAE)
has defined three basic categories of in-vehicle
networks based on network speed and functions:
Class A, Class B and Class C networks.
• Class A is a low-speed network with a bit
rate of 10 Kbits/sec or less. Class A network
is to be used for convenience features such
as entertainment, audio, trip computer, etc.
Class A functions require inexpensive and
low-speed communication, and typically
utilize generic UARTs (Universal Asyn-
chronous Receiver/Transmitter).
• Class B is a medium-speed network. Data
rate for Class B network is between 10
Kbits/sec and 125 Kbits/sec. The intended
application for Class B network is general
information transfer such as instrument
cluster, vehicle speed, legislated emissions
data, etc. In the US, the SAE adopted J1850
as the standard protocol for Class B net-
works. SAE J1850 has two basic versions.
The first is a 10.4 Kbits/sec VPW (Variable
Pulse Width) type which uses a single bus
wire. The second is a 41.6 Kbits/sec PWM
In-Vehicle Network Architecture for the Next-Generation Vehicles
(Pulse Width Modulation) type which uses
a two-wire differential bus.
• Class C is a high-speed network, and it is
to be used for real-time control applications
such as power-train, vehicle dynamics, anti-
lock brakes, etc. The data rate for Class C
network is 125 Kbits/sec to 1 Mbits/sec.
CAN protocol has been selected for Class
C applications.
Continuous regulatory demands for fuel ef-
ficiency, safety and emission control are forcing
the automakers to implement advanced safety and
engine management systems. These advanced
systems will require new types of networks and
networking protocols. The demands for improved
fuel efficiency require drive-by-wire systems that
eliminate camshafts, power-sapping belt drives
and pumps, and a great deal of unnecessary weight.
The goal of drive-by-wire is to replace as many
mechanical systems as possible by equivalent elec-
tronic and electrical systems. Electric brakes will
replace hydraulics, increasing safety, decreasing
weight, lowering operating cost and eliminating
the use of environmentally hazardous fluids. Steer-
ing columns will disappear and will be replaced
by electric steering, improving driver safety and
making it easy to provide left-hand and right-hand
drive models. Similarly, as the demand for other
new features such as telematics, entertainment,
multimedia, pre-crash warning, driver assistance
systems, remote diagnostic, software update, etc.
increases, the complexity of in-vehicle commu-
nication networks increases as well. Depending
upon the new functionality of vehicle electronic
modules, different sets of modules will require
different types of new networks. For example,
drive-by-wire and active collision avoidance
systems need fault tolerant networks with time-
triggered protocols such as TTP/C (TTTech,
2002), TTCAN (Fuehrer, 2001) and FlexRay
(FlexRay, 2004), to guarantee critical services
with deterministic latencies. Airbag systems
need a protocol that can react very quickly. For
airbag systems there are various protocols, such
as Safe-by-Wire (Lupini, 2004), Bosch Siemens
Temic (BST), DSI and Byteflight (Lupini, 2004),
which are being investigated. Multimedia systems
need high-bandwidth networks such as MOST and
IntelliBus (Lupini, 2004; Firlit, 2004), to transfer
video files. Wireless personal area networks with
voice activated control are necessary for conve-
nience such as turning on/off lights, wipers, radio,
heat and air conditioner without taking hands off
the steering wheel (Mahmud, 2006). Bluetooth
or Wi-Fi protocol could be used for wireless per-
sonal area network. Telematic systems need tech-
nologies such as Wi-Fi or Ultrawideband (UWB).
Thus, it is apparent that future vehicles will have
various types of networks with many partitions.
Interconnecting these partitions is necessary for
distributed control across various partitions and
ease of diagnosis. Since automotive industry is
very cost sensitive, interconnecting these parti-
tions will be a challenging task especially when
the number of partitions is large. Major studies
and in-depth analysis are necessary to determine
optimal topology for interconnecting all the in-
vehicle network partitions, to figure out optimal
bandwidth required from different types of net-
work partitions for keeping latencies under certain
bounds, and to select appropriate communication
protocols for different network partitions. Various
topologies for interconnecting different in-vehicle
network partitions are investigated and presented
in the next section of this book chapter.
INTERCONNECTION OF NETWORK
PARTITIONS
Issues, Controversies and Problems
Multiplexing or in-vehicle networking evolved
due to the need for reducing wiring harness,
ease of integration, ease of diagnosis and cost
savings. Most of the today’s vehicles have only
two network partitions as mentioned earlier in
287
 In-Vehicle Network Architecture for the Next-Generation Vehicles
this book chapter and also shown in Figure 1.
Interconnecting two partitions is a trivial problem.
We need only one gateway device to connect the
two partitions. The gateway device must support
protocols of both partitions. The gateway device
will do protocol conversions when messages need
to go from one partition to the other partition.
Nowadays, microcontrollers are available that
support various types of interfaces. For example,
PIC18F4685 microcontroller supports SPI, I2C,
RS-485, RS-232, LIN and CAN interfaces. Simi-
larly, PIC18F97J60 microcontroller supports SPI,
I2C, RS-485, RS-232, LIN and Ethernet interfaces.
Thus, a microcontroller that supports both inter-
faces of the two partitions, shown in Figure 1,
can be used as the gateway device between the
two partitions.
As the number of partitions grows, dedicated
point-to-point connections among various parti-
tions will require many gateway devices, and
beyond a certain point it would be cumbersome
and cost prohibitive. One of the challenging issues
will be the selection of an appropriate topology to
interconnect various partitions of the in-vehicle
network.
S 2. Figure 3 shows an additional unit called the
There are various existing network topologies such
as bus, ring, mesh, crossbar, star, etc. Bus-based
systems have many advantages over other inter-
connection systems (Das, 1985; Chen, 1999). Some
of the advantages of bus-based systems are low
cost, lower complexity, greater reliability, better
scalability, simpler communication protocols and
ease of diagnostics. Thus, in order to interconnect
all partitions, one logical approach would be to
use another bus such as a hierarchical bus.
Selection of a Topology
In this subsection, a number of multi-level
bus-based network topologies are discussed
and suggested for future in-vehicle networks.
288
A hierarchical bus-based network provides a
cost-effective interconnection solution technique
for partition-based systems (Mahmud, 1994).
Since interconnection using a hierarchical bus
requires less number of gateway devices, one
logical approach for interconnecting four parti-
tions of a vehicular network will be the use of a
second level bus or an Level 2 bus (L2 bus), as
shown in Figure 3 (Mahmud, 2005). Figure 3
shows that each partition has its own bus called
the Level 1 bus or L1 bus. Different partitions
have different types of L1 buses. For example,
one partition may have a CAN bus as its L1 bus,
and another partition may have a LIN bus as its
L1 bus. An L1 bus of a partition is connected to
the L2 bus by a gateway device. For example,
the L1 bus of Partition-1 is connected to the L2
bus using the gateway device G1, as shown in
Figure 3. Thus, the number of gateway devices
required to connect N partitions to the L2 bus
is N. Whereas for dedicated point-to-point con-
nections among various partitions, the number
N!
of required gateway devices is .
2 * ( N − 2)!
Hence, the complexity of connections shown in
Figure 3 is much less than that shown in Figure
Diagnostic Gateway. Dealer will connect its
diagnostic computer (diagnostic module) to the
vehicle via this diagnostic gateway. Thus, for N
partitions in a system, total number of connections
to the L2 bus including the diagnostic gateway is
(N + 1). For a dedicated point-to-point connection
system, there has to be a gateway device between
the diagnostic module and each partition. Thus
for a system with N partitions and a diagnostic
module, dedicated point-to-point connections will
( N + 1)!
require gateway devices. It is clear
2 * ( N − 1)!
that a two-level bus system, which contains an L1
bus for each partition and only one L2 bus for the
entire system, is a cost-effective way of intercon-
necting all the partitions. The bit rate of the L2
bus should be at least equal to the total bit rate
In-Vehicle Network Architecture for the Next-Generation Vehicles

necessary for inter-partition communications for Mi

li , j
all partitions. Normally not too much inter-parti- di = ƒ (1)
t
j =1 i , j
tion data needs to be exchanged among various
partitions. Thus, a Class B or even a Class A bus
may be good enough. However, the final selection Let f i be the fraction of Partition-i’s messages
decision must be made after an extensive analysis that will go to the L2 bus. Then the total traffic
and simulation of the system. density on the L2 bus is
Required Bandwidth for the L2 bus: An ap-
N ≈ Mi l ÷
proximate analysis of the required bandwidth for N
d L 2 = ƒ di f i = ƒ ∆ƒ
i, j
∆ ◊◊f i (2)
the L2 bus can be determined as follows. Assume t
i =1  j =1 i , j
i =1 
that the system has N partitions. Let di be the traf-
fic density (bits/sec) on the L1 bus of Partition-i.
Nowadays, most messages for a vehicular system As the traffic density on a bus increases, the
are periodic. We would like to make it clear that queuing delay for the messages to get the bus
periodic messages are not to be confused with increases as well. If u is the utilization of a par-
time-triggered messages. Time slots are reserved ticular bus, then the probability that a message
for time-triggered messages but not for periodic will get the bus during its first trial is (1 – u). The
messages. Periodic messages need to go through probability that the message will not get the bus
an arbitration process to get the bus. If all peri- during its first trial but will get the bus during
odic messages and their periods for a particular its second trial is u(1 – u). In general, it can be
partition are known, then the value of di can be shown that the probability that the message will
estimated. Let Mi be the number of messages in not get the bus during the first (i – 1) trials but it
Partition-i, and li,j be the length of Message-j of will get the bus during the ith trial is ui–1(1 – u).
Partition-i with period ti,j. The value of di can Hence, the average number of trials necessary for
∞
1
then be written as: the message to get the bus is ƒ iu i −1 (1 − u ) = .
i =1 1− u

Figure 3. An in-vehicle network with four L1 buses and one L2 bus

Partition-1 Partition-2

L1 Bus L1 Bus

G1 Diagnostic Gateway G2

L2 Bus
G3 G4

L1 Bus L1 Bus

Partition-3 Partition-4

289
 In-Vehicle Network Architecture for the Next-Generation Vehicles
The average message transmission delay through
the bus, including the queuing delay for the bus, is
lavg
. Where, lavg is the average length of a
(1 − u ) BW
message and BW is the bandwidth of the bus. For
example, if u = 30%, lavg = 120 bits and BW = 500
kbits/sec, then the average message transmission
delay including the queuing delay for the bus is
ulavg
343 μsec. The average queuing delay is
(1 − u ) BW
and it increases very rapidly with the increase in
bus utilization u. For example, for lavg = 120 bits
and BW = 500 kbits/sec, the average queuing
delay is 103 μsec for 30% bus utilization, and it is
360 μsec for 60% bus utilization.
In vehicular networks, normally the bus
utilization is kept around 30% or less to reduce
contention on the bus so that a message is not
significantly delayed to get the bus. Let uL2 be the
maximum allowable stead-state utilization of the
L2 bus. The required bandwidth for the L2 bus
can then be expressed as
N ≈ M i li , j ÷
dL2
ƒ ∆∆ƒ ◊◊f i
i =1  j =1 ti , j 
BWL 2 = = (3)
u L2 uL 2
Since, for a given system, the values of li,j and
ti,j are accurately known, and the value of f i can
also be accurately determined by investigating
all the messages of a partition, the value of the
required bandwidth, BWL2, for the L2 bus can be
determined for a given value of uL2. However, the
actual selection of the L2 bus needs to be made
after extensive simulation and analysis of the
entire system.
Box 1.
lavg − SL1 lavg − L 2
T= + TSG +
(1 − uSL1 )BWSL1 (1 − uL 2 )BWL 2
290
End-to-End Message Latency for a 2-Level Bus
System: In a 2-level bus system, an inter-partition
message will first go through the L1 bus of the
source partition, then it will go through the L2
bus, and finally it will go through the L1 bus of
the destination partition. Thus, various bus delays
and protocol conversion delays will contribute
to the total end-to-end message latency from
the source node to the destination node. For a
particular message, the end-to-end latency due to
network components, such as buses and gateway
devices, can be expressed as:
T = TSL1 + TSG +TL2 + TDG + TDL1 (4)
where, TSL1 is the delay due to the L1 bus of the
source partition; TSG is time required by the source
gateway device (the gateway device which con-
nects the source L1 bus to the L2 bus) to convert
messages from one protocol to another protocol;
TL2 is the delay due to the L2 bus; TDG is time
required by the destination gateway device (the
gateway device which connects the destination L1
bus to the L2 bus) to convert messages from one
protocol to another protocol; and TDL1 is the delay
due to the L1 bus of the destination partition. TSG
and TDG can also be called as protocol conversion
time. Substituting the values of bus delays in
Equation (4), the end-to-end message delay can
also be expressed as (see Box 1), where, lavg – SL1,
uSL1 and BWSL1 respectively indicate the average
message length, utilization and bandwidth of the
source L1 bus; lavg – L2 , uL2 and BWL2 respectively
indicate the average message length, utilization
and bandwidth of the L2 bus; and lavg – DL1 , uDL1
and BWDL1 respectively indicate the average
message length, utilization and bandwidth of the
lavg − DL1
+ TDG + (5)
(1 − uDL1 )BWDL1
In-Vehicle Network Architecture for the Next-Generation Vehicles
destination L1 bus. The message delay basically
has two components: the sum of the delays due to
all buses and the sum of the protocol conversion
times for the gateway devices. Thus, the message
delay for a two-level bus system can be expressed
as, (see Box 2).
Since the number of gateway devices in a two-
level bus system increases linearly, as opposed to
exponential increase for dedicated point-to-point
connections, a two-level bus system is scalable
meaning that the system is easily expandable.
Although the system shown in Figure 3 is scalable,
the number of connections on the L2 bus will be
too many if there are too many partitions in the
in-vehicle network. A bus with a given speed can
support up to a certain number of nodes and up to a
certain length due to various reasons as explained
in the next few sentences. As the number of nodes
increases, the capacitive loading effective on the
bus increases due to connections which eventu-
ally increases the rise-time and fall-time of the
electrical signals. As a result, the maximum bit
rate capacity of the bus decreases. When more
nodes are added to a bus, the total bandwidth
demand from all the nodes increases as well.
Thus, for a certain allowable bus utilization, the
bus will be able to support up to a certain number
of nodes. For example, if the bandwidth of a bus
is 500kbits/sec and the allowable bus utilization
is 30%, then total bandwidth demand from all
the nodes can’t exceed 150kbits/sec. If each node
needs a bus bandwidth of 10kbits/sec, then the bus
can support only up to 15 nodes. Lastly, when the
length of the bus increases, the capacitive loading
effective on the bus also increases. As a result,
the rise-time and fall-time of the signals increase,
especially for the end-to-end communication on
Box 2.
≈ lavg −bus
T= ƒ ∆
∆
over − two − L1− buses  (1 − ubus ) BWbus
and − one − L 2 − bus
the bus. Hence, the maximum bit/rate capacity
of the bus decreases.
Future vehicles many require many partitions,
one for each subsystem of the vehicle. For example,
the airbag system will need a partition for itself
because there could be as many as 64 nodes for
the airbag system (Lupini, 2004). In future, as the
number of partitions will grow, various inter-parti-
tion connection topologies should be investigated
in order to select an appropriate topology for
interconnecting partitions. The remaining parts
of this book chapter present various topologies
for interconnecting partitions.
Figure 4 shows another design where, for
the same number of partitions in the system,
the number of connections to the L2 bus is less
than that shown in Figure 3. Figure 4 shows that
six partitions (P1, P2, . . , P6) are connected to
the L2-bus using three gateway devices (G1, G2,
and G3). Each gateway device has three ports:
two for connecting to two partitions and one for
connecting to the L2-bus. As mentioned ear-
lier, nowadays commercial microcontrollers are
available that support many different interfaces.
Thus, for the gateway devices, G1, G2 and G3,
microcontrollers must be selected in such a way
that they support necessary interfaces for the L1
and L2 buses. The system shown in Figure 4 also
has a diagnostic gateway (DG) connected to the
L2 bus. For the design shown in Figure 4, total
number of connections to the L2 bus including
the diagnostic gateway isN/2+1, where N is the
number of partitions and the notation n means
ceiling of n. Moreover, for the same number of
partitions in the system, the data rate required
for the L2 bus of Figure 4 is less than that for the
L2 bus of Figure 3. The reason is that the data
÷
◊◊ + ƒ TPr otocolConversion (6)
 over − two − gateway
devices
291
 In-Vehicle Network Architecture for the Next-Generation Vehicles
transfer between the pairs of partitions (P1, P2),
(P3, P4) and (P5, P6) can go directly through the
gateway devices rather than through the L2 bus.
As a result, for the same L2 bus technology, the
design shown in Figure 4 can support more parti-
tions than that shown in Figure 3.
In order to design a system like the one shown in
Figure 4, a partition, say Pi, should be paired with
another partition, say Pj, such that inter-partition
traffic between the pair (Pi, Pj) will be more than
that between any other pair (Pi, Pk) for all j ≠ k.
In other words two partitions can be paired to-
gether if they have more functional dependencies.
Moreover, the two partitions which are going to
be paired together should be physically close to
each other. Otherwise, very long wires will be
necessary to connect them to the corresponding
gateway device which will reduce the benefits of
multiplexing.
In the recent past, drive-by-wire or x-by-wire
concepts have drawn significant attentions within
the automotive industries. As drive-by-wire may
become a standard sometimes in the future, vari-
ous safety critical messages may go through one
for more L1 buses. To guarantee on-time delivery
of safety critical messages, future in-vehicle net-
works should be designed such that the system
will tolerate bus faults. Otherwise, the safety of
vehicles and passengers will be compromised.
There are various existing time-triggered pro-
tocols for drive-by-wire applications. Some of
Figure 4. An in-vehicle network with six partitions (P1, P2, . . , P6), three gateways/switches (G1, G2
and G3) and one diagnostic gateway (DG)
P1 P2 P3
G1 G2
L2 Bus
292
these protocols are TTCAN, FlexRay and TTP/C.
FlexRay and TTP/C support a dual-bus system.
In these systems, safety critical messages are
sent through both buses at the same time. Thus,
safety critical messages can still be delivered if
only one bus is at fault at the time of delivery.
If the TTCAN protocol is ever going to be used
for drive-by-wire applications, then it must also
support two buses for tolerating single-bus faults.
Figure 5 shows another network architecture
where Partition P1 contains a network with a time-
triggered protocol for drive-by-wire applications.
P1 uses two buses to tolerate single-bus faults. As
a result, the gateway device G1 has four ports: two
for two L1 buses of P1, one for one L1 bus of P2
and another one for the L2 bus.
As mentioned earlier, pairing of partitions will
be feasible if the two partitions of a pair are close
to each other, and they have significant functional
dependencies. Otherwise, pairing will not provide
any additional benefits to vehicle multiplexing.
If pairing is not possible due to functional
or physical limitations, and if one L2 bus is not
enough to connect all partitions of a vehicular net-
work, then two or more interconnected L2 buses
can be used. The architecture shown in Figure 6
has two L2 buses. Each L2 bus connects a number
of partitions to form a cluster of partitions. One
L2 bus connects Partitions P1 through Pi to form
one cluster, and another L2 bus connects Parti-
tions Pi+1 through Pj to form another cluster. The
P4 P5 P6
G3
DG
In-Vehicle Network Architecture for the Next-Generation Vehicles

Figure 5. An in-vehicle network with six partitions, three gateways/switches and one diagnostic gateway.
Partition P1 has two L1 buses

P1 P2 P3 P4 P5 P6

G1 G2 G3

L2 Bus
DG

Figure 6. An in-vehicle network with two clusters of partitions. A gateway device G connects the two
clusters and a diagnostic gateway device DG

P1 P2 Pi Pi+1 Pi+2 Pj

G1 G2 Gi Gi+1 Gi+2 Gj
G
L2 Bus L2 Bus
DG
A Cluster of Partitions A Cluster of Partitions

two clusters are connected to a 3-port gateway
device G. The third port of G is connected to the
diagnostic gateway DG. In this type of systems,
inter-cluster messages will have more end-to-end
latencies than inter-partition messages within the
same cluster because the inter-cluster messages
will have to go through one more L2 bus and one
more gateway device G. Since clustering has to be
done in such a way that inter-cluster messages are
not going to control vehicle dynamics in real-time,
some additional latency for inter-cluster messages
could be tolerated by the system.
If the in-vehicle networking system is going to
have more than two clusters, then the additional
clusters can be connected in a cascade manner.
Figure 7, shows an in-vehicle networking system
with four clusters connected in a cascade manner.
The four L2 buses of four clusters are connected
in a cascade manner using three gateway devices:
GW1, GW2, and GW3. The end-to-end message
delay will be maximum when messages need to
go from Cluster 1 to Cluster 4 and vice versa.
The maximum delay can be expressed as, (see
Box 3).
The total number of gateway devices, including
the diagnostic gateway, in an in-vehicle network
with cascaded clusters is m + c, where m is the
number of partitions, and c is the number of
clusters in the system.
If the inter-cluster message delay between
Cluster 1 and Cluster 4 is too high and not accept-
able, then another way of interconnecting those

293
 In-Vehicle Network Architecture for the Next-Generation Vehicles

Box 3.

≈ lavg −bus ÷
TCascade = ƒ ∆
∆
over − two − L1− buses  (1 − ubus ) BWbus
◊◊ + ƒ− gateway TPr otocolConversion (7)
and − four − L 2 − buses
 devices
over − five

Box 4.

≈ lavg −bus ÷
THierarchical = ƒ ∆
∆
over − two − L1− buses ,  (1 − ubus ) BWbus
◊◊ + ƒ TPr otocolConversion
(8)
two − L 2 − buses
 over − four − gateway
devices
and − one − L 3− buses

clusters would be the use of a third level bus or an
L3 bus. Figure 8 shows a system with four clusters
interconnected by an L3 bus. The four clusters
are connected to the L3 bus using four gateway
devices: GW1, GW2, GW3 and GW4. The topology
shown in Figure 8 is in fact a hierarchical network
topology with three levels. For the network shown
in Figure 8, the inter-cluster message delay can
be expressed as, (see Box 4).
The total number of gateway devices, including
the diagnostic gateway, for a 3-level hierarchi-
cal in-vehicle network is m + c + 1, where m is
the number of partitions, and c is the number of
clusters in the system.
From Equations (7) and (8), it is seen that the
maximum end-to-end message delay for the hier-
archical network is less than that for the cascaded
network because for the cascaded network with
four clusters, in the worst case the messages have
to go through six buses, and five gateway devices
will have to run protocol conversion algorithms.
Whereas, for the hierarchical network with four

Figure 7. An in-vehicle network with four clusters connected in a cascade manner using gateway de-
vices: GW1, GW2, and GW3

Cluster 1 Cluster 2
P1 P2 Pi Pi+1 Pi+2 Pj

G1 G2 Gi Gi+1 Gi+2 Gj
GW1
L2 Bus L2 Bus
DG GW2
L2 Bus L2 Bus
GW3
Gk+1 Gk+2 Gm Gj+1 Gj+2 Gk

Pk+1 Pk+2 Pm Pj+1 Pj+2 Pk

Cluster 4 Cluster 3

294
In-Vehicle Network Architecture for the Next-Generation Vehicles

Figure 8. A hierarchical in-vehicle network with four clusters interconnected using an L3 bus
Cluster 1 Cluster 2
P1 P2 Pi Pi+1 Pi+2 Pj

G1 G2 Gi Gi+1 Gi+2 Gj

L2 Bus L2 Bus
GW1 DG GW2

L3 Bus
GW3 GW4
L2 Bus L2 Bus

Gk+1 Gk+2 Gm Gj+1 Gj+2 Gk

Pk+1 Pk+2 Pm Pj+1 Pj+2 Pk

Cluster 4 Cluster 3

Figure 9. An in-vehicle ring network with four will be a cost-effective choice than a hierarchical
clusters in-vehicle network.
If Clusters 1 and 4 are physically close to
Cluster 1 Cluster 2 each other, then another better way of forming
the network will be to connect the L2 buses of
GW1
L2 Bus L2 Bus Clusters 1 and 4 by a gateway device as shown
in Figure 9. As a result, a ring network is formed
GW4 DG GW2 at the level of L2 buses.
For the ring network, in the worst-case an
L2 Bus L2 Bus inter-cluster message will have to go through
GW3
two L1 buses, three L2 buses and four gateway
Cluster 4 Cluster 3 devices. Thus, the worst-case message delay can
be written as, (see Box 5).
From Equations (7), (8) and (9), it is seen that
the worst-case inter-cluster message delay for the
clusters, the messages have to go through five
ring network of Figure 9 is less than that for the
buses, and four gateway devices will have to run
cascaded network of Figure 7 but comparable
protocol conversion algorithms. For the hierarchi-
to that for the hierarchical network of Figure 8.
cal network, the price that we have to pay to get
However, the ring network will be less expensive
lower message delay compared to the cascaded
than the hierarchical network because it does not
network, is the cost of an extra gateway device
need the L3 bus. There is one minor disadvantage
and the cost of an extra bus (the L3 bus). Thus, if
of the ring network compared to the hierarchical
little bit longer inter-cluster message delay is not
network. In the hierarchical network, the messages
a major issue, then cascaded in-vehicle network

295
 In-Vehicle Network Architecture for the Next-Generation Vehicles
Box 5.
≈ lavg −bus
TRing = ƒ ∆
∆
over − two − L1− buses  (1 − ubus ) BWbus
and − three − L 2 − buses
from the diagnostic gateway to a network node
will have to go through three buses: the L3 bus, an
L2 bus and an L1 bus, and two gateway devices.
Whereas, in the ring network, for the worst-case,
the messages from the diagnostic gateway to a
network node will have to go through three buses:
two L2 buses and an L1 bus, and three gateway
devices. Since a slight increase in latency during
a diagnostic procedure is not a problem, a ring
network will be the choice among the all three
choices: cascade, hierarchical and ring, provided
the physical locations of the clusters enable a ring
network to be formed.
SMULATION MODELS FOR
ANALYSF VARIOUS
IN-VEHICLE NETWORK
TOPOLOGIES
Extensive simulation and analysis should be
done before a particular topology is selected for
a particular type of vehicle. The performance
of various partitions of the in-vehicle network
as well as the entire network can be determined
by developing simulation models. After execut-
ing the simulation models on computers, one
can determine the minimum requirements for
different partitions of the network architecture.
Some of these requirements will include mini-
mum processing powers needed from various
electronic modules, minimum bandwidth needed
from various buses, minimum computation power
and minimum size of buffers needed in various
gateways, etc. Computer simulations will also
allow us to select the optimal network topology
needed to meet the requirements of the next
296
÷
◊◊ + ƒ− gateway TPr otocolConversion (9)
 devices
over − four
generation vehicles. Types of simulations and
techniques of developing simulation models for
vehicular networks are described in the following
subsections of this book chapter.
Types of Simulations: Computer simulations
can be divided into two types (Molloy, 1989):
i) time-based simulation and ii) event-based
simulation. In a time-based simulation, the
program control loop is associated with time.
For each execution of the main control loop, the
simulation clock advances by a fixed time unit.
In the case of an event-based simulation, the
execution of the main loop represents a single
event. The simulation clock is simply advanced
by the amount of time since the last event. For
asynchronous systems, the event-based simula-
tion is computationally more expensive and ac-
curate than the time-based simulation. But, for
synchronous systems, the time-based simulation
is as good as the event-based simulation, and it is
computationally less expensive than event-based
simulation. CAN protocol is an event-based
protocol. Thus, an event-based simulation tool
will be the most appropriate tool for analyzing
various performance parameters of a CAN bus.
Buses that use time-triggered protocols, such as
TTCAN, and FlexRay, carry both time-triggered
and event-triggered messages. The time-based
simulation can be used to simulate time-triggered
messages and the event-based simulation can be
used to simulate event-triggered messages. Thus,
a combination of both time- and event-based
simulations is necessary to simulate a bus like
TTCAN or FlexRay.
Developing an Event-Based Simulation Tool:
Here we explain how to develop an event-based
simulation tool to meet the requirements of specific
In-Vehicle Network Architecture for the Next-Generation Vehicles
needs for the next generation vehicles. In a vehicle
system, an event is triggered by a sensor or by the
driver’s action such as pressing the brake or the
gas paddle. The event then goes to a processor
queue, as shown in Figure 10. If the processor is
busy or other events are waiting in the proces-
sor queue, then the event stays in the processor
queue. Otherwise, the processor starts working on
the event immediately. The event, queued in the
processor queue, will stay in the queue until the
previous events are taken care of by the proces-
sor. The processor spends some time (processing
time) on an event. After that, the processor tries
to send a message through the bus (say CAN bus)
connected to the processor. If the bus is available,
then the message is immediately sent to the bus.
Otherwise, the message is queued in the transmit
buffer of the node. The message will go through
the bus after the previous messages have been
transmitted. After a certain time (transmit time),
the message is queued to the receive buffer of the
destination node. If the processor at the receiving
node is free, then it will immediately process the
message. Otherwise, the processor will first take
care of other queued messages and then it will
process this new message. Figure 10 shows the
entire process of information flow from a sensor
to an actuator for a single-bus system.
Figure 10. Information flow from a sensor to an actuator
Processor Queue
Sensor
Receive Buffer
of CAN node
Processor
Components of the latency of an event: There
is a delay (latency) between the time an event
occurs at a sensor or due to a driver’s action, and
the time an actuator takes an action. The latency
has several components as shown in the follow-
ing equation.
latency = tpq + tps + ttb + tbus + tpropagation + trb + tpd
(10)
where,
• tpq = delay at the processor queue
• tps = processing time at the source node
• ttb = delay at the transmit buffer of the source
node
• tbus = transmit time through the bus
• tpropagation = propagation delay through the
bus
• trb = delay at the receive buffer of the desti-
nation node
• tpd = processing time at the destination
node
In the simulation model, the value of the delay,
tpq, at the queue of the source processor can be de-
termined by monitoring the arrival and exit times
of the event in the processor queue. The value of
Transmit Buffer
of CAN node
Processor
CAN Bus
Actuator
297
 In-Vehicle Network Architecture for the Next-Generation Vehicles
the processing time, tps, depends on the type of
the processor. This time can be expressed as
tps = k1s + k2sN (11)
Where k1s is a constant time, which represents the
constant overhead of the processor at the source
node in order to prepare a message. This constant
overhead may include:
• The time needed by the processor to get out
of the main program,
• The time needed by the processor to go to
the interrupt service routine (assuming that
the transmission of a message is interrupt
driven),
• The time needed by the processor to check
some status bits and set/clear some control
bits of the CAN controller,
• The time needed by the processor for some
other house keeping operations.
Thus, for a given processor, the value of k1s
can be calculated based on its clock frequency
and the number of instructions the processor
needs to execute the abovementioned operations.
The parameter k2s, shown in Equation (11), is
another constant, per byte of data in coding the
message, and N is the number of data bytes in
the message.
The value of ttb can be determined by monitor-
ing the arrival and exit times of the message in
the transmit buffer of the source node. The value
of tbus can be expressed as
L
tbus = (12)
S generated by a scheduler. The scheduler gener-
Where, L is the total number of bits in the mes-
sage, and S is the speed of the bus in bits/sec. For
example, if the message is a CAN message, then
L will include the start bit, arbitration bits, RTR
bit, control bits, data bits, stuff bits, CRC bits, all
delimiter bits and the end of frame bits.
298
The value of tpropagation should be negligible
compared to other components of the delay. It
depends on the speed of electrical signals through
the bus. In free space, electromagnetic signals
take approximately one nanosecond to move a
foot. Thus, the propagation delay through a bus
should be little over one nanosecond per foot. As a
result, the value of tpropagation should be in the order
of few tens of nanoseconds which is very small
compared to other components of Equation (10).
Thus, the propagation delay through the bus can
be ignored for performance analysis.
The value of trb can be determined by monitor-
ing the arrival and exit times of the message in
the receive buffer of the destination node. Like
tps, the value of tpd can be expressed as
tpd = k1d + k2dN (13)
Where k1d is a constant time, which represents the
constant overhead of the processor at the destina-
tion node in order to decode a message, and k2d
is the additional overhead, per byte of data, in
decoding the message.
For a system with multiple partitions, the
latency of an event can be expressed as, (see
Box 6).
The term tgateway indicates the delay at each
gateway device on the path of the signal. There
are three components of tgateway: queuing delays at
both the input and output queues of the gateway
and the time required for protocol conversion by
the gateway device.
Simulation Model: Figure 11 shows the block
diagram of an event-based simulation model. In
an event-based simulation model, an event is
ates this event based on some kind of realistic
stochastic process. We can develop a stochastic
model for a particular event by collecting a large
set of real data for that event. For example, the
stochastic model of the brake event (pressing the
brake paddle by the driver), can be developed by
In-Vehicle Network Architecture for the Next-Generation Vehicles
Box 6.
latency = t pq + t ps + ttb + ƒ (t
over − all − buses − on − the
path − of − the − signal
logging the driver’s brake activity over a long
period of time and collecting similar data from
many vehicles.
After generating an event, the scheduler keeps
that event in the event queue, as shown in Figure
11. When an event is dispatched from the event
queue, another event is generated. This new event
is then sent to the event queue. When a new event
enters into the event queue, it is sorted within the
event queue using the event-time as the sorting
key. The new event is then placed at an appropri-
ate point in the event queue. After dispatching an
event from the event queue, the event is then sent
to the event processing unit. The event processing
unit then simulates the event in exactly the same
way as a real system would process it.
The stochastic model of an event determines
the inter-arrival time of the event, based on the
behavior of the device that generates the event.
A synchronous device, such as a time-triggered
device, generates events after fixed interval of
time. Hence, for a time-triggered system, the
inter-arrival time is fixed. But, an asynchronous
device doesn’t generate events after fixed interval
of time. Most physical asynchronous systems
behave like a Poisson process. Thus, we can as-
sume that a non time-triggered device will behave
like a Poisson process. The stochastic model of
an event will have the appropriate values of the
Figure 11. Model for an event-based simulation
Queue for events
Event Scheduler
(an asynchronous
process)
bus + t propagation )+ ƒ t gateway + trb + t pd (14)
over − all − gateways − on
the − path − of − the − signal
parameters that define the characteristics of the
Poisson process of the event. As mentioned earlier,
these parameters of the stochastic model for an
event can be determined by collecting a large set
of field data for that particular event. The model
shown in Figure 11 can be used to simulate an
event-triggered system with an event-triggered
bus (say CAN bus).
Since both TTCAN and FlexRay support time-
and event-triggered messages, we can develop
the simulation models of these buses by building
another layer on the top of the event-based simula-
tion model. Figure 12 shows a simulation model
of a combined time- and event-based system.
Here, the event scheduler is a deterministic as
well as a stochastic process. In this case, a global
simulation clock is maintained to keep the time-
triggered events in sync. The frequency of this
global simulation clock (software clock) should
be at least equal to the highest frequency clock
used in the in-vehicle network architecture. This
requirement on the software clock will allow us
to determine the latency and other performance
parameters at a very high resolution (within a
clock cycle of the highest frequency clock of the
real system). At every global simulation clock
pulse, the event queue must be checked. If the
event-time of the event at the head of the event
queue is less than or equal to the current value of
Event Event
Dispatcher Processor
299
 In-Vehicle Network Architecture for the Next-Generation Vehicles
Figure 12. Model for an event- and time-based simulation
Event Scheduler Queue for events
(asynchronous and
synchronous
processes)
the global simulation clock, then the event will be
dispatched from the event queue. The dispatched
event will then be checked to determine whether
it is a deterministic event or a non-determinis-
tic event. If it is a deterministic event, then the
deterministic scheduling process will be used to
schedule the next deterministic event. Otherwise,
stochastic scheduling process will be used to
schedule the next non-deterministic event. The
scheduled event, whether deterministic or non-
deterministic, will then be sent to the event queue.
The event will be placed in the event queue after
sorting it with other events in the event queue.
The event-time will be used as the key to sort the
events in the event queue.
By developing a simulation model, one can de-
termine various performance parameters of the
in-vehicle network. These parameters will include,
but not limited to, the latency of different types of
message, busload, bandwidth, and throughput.
FUTURE TRENDS
There is no doubt that future trends are going to
be in the direction of drive-by-wire, active-safety,
inter-vehicle communications and vehicle-to-
infrastructure communications. Future vehicles
will access internet and exchange information
with other vehicles and road-side units for vari-
ous reasons such as collision avoidance, message
300
Global Simulation
Clock
Event Event
Dispatcher Processor
dissemination, software upgrade in electronic
modules, receiving data from content providers,
sharing files, etc. Some researchers have pro-
posed architecture for Next Generation Vehicle
Network and presented a dynamic discovery ser-
vice protocol for internet access (Baroody, 2005;
Baroody, 2006). If future in-vehicle networks
are not properly designed, then vehicles’ safety
could be compromised or its operation could be
degraded due to various types of cyber attacks
from external entities. The in-vehicle network
that is mainly responsible for controlling vehicle
dynamics, must be well protected from outside
cyber attacks. Good security software and fire-
walls must be installed in all gateway devices
that exchange vehicle’s dynamic information. The
gateway device that connects the telematic unit
to the internal network of the vehicle must have
a good firewall so that it can filter out all suspi-
cious information that is coming from outside.
After extensive research, if it is not possible to
come up with a good security solution to protect
vehicles’ safety-critical modules, then no external
traffic should be allowed to enter the safety-criti-
cal modules.
For collision avoidance applications, a vehicle
will disseminate its dynamic information. How-
ever, current in-vehicle networks are not secure.
This means that when receiving nodes are pick-
ing up CAN messages from a CAN bus, they
are not verifying whether or not the messages
In-Vehicle Network Architecture for the Next-Generation Vehicles
are authentic. The receiving nodes automatically
assume that the messages are valid as long as the
messages are free from various types of errors
such as CRC error, bit error, stuff error and frame
error. Thus, if a person with malicious intent
connects a device to the vehicle’s CAN bus and
sends messages with wrong information, other
nodes will still accept them as valid messages.
The problem will occur when these messages with
wrong information are picked up by a gateway
device and then routed to the vehicle’s telematic
unit. The telematic unit will then broadcast these
messages with wrong information to other ve-
hicles. As a result, there could be disruption of
traffic leading to accidents.
Since vehicle-to-vehicle communication is a
peer-to-peer communication, protection of driv-
ers’ or owners’ privacy from peers is going to
be another challenging task. Privacy protected
secure communication is not an easy task. Good
Cyber-Physical Systems at the network level are
necessary so that future vehicles will be able
to communicate with external entities without
getting involved in safety, security and privacy
problems.
The issues related to vehicular network secu-
rity are still at the research stage. The purpose of
this book chapter is not to provide any security
solutions but to explain the significance of this
topic and what needs to be done to safely oper-
ate future intelligent vehicles. Many researchers
have already started looking into these issues and
are trying to propose solutions (Hossain, 2007;
Rabadi, 2007; Rabadi, 2008).
CONCLUSON
In this paper, we discussed various types of to-
pologies and protocols that could be used in the
future in-vehicle networks. Due to the complexity
of the future in-vehicle networks, many partitions
of the network will be required. These partitions
must be interconnected together for functional
dependencies and better diagnostic purposes.
A number of network topologies has been pre-
sented and analyzed for cost, bandwidth and
message latencies. Since future vehicles will be
communicating with external entities for various
reasons, the book chapter also addressed the issues
of security, safety and privacy which should be
taken into consideration at the time of designing
the in-vehicle network components. Finally, some
ideas have been presented in developing simula-
tion models to analyze various types of networks
which will ultimately help in selecting a network
topology and various network components for a
given set of requirements and specifications.
REFERENCES
Baroody, R., Al-Holou, N., & Mohammad, U.
(2006). Reliable Dynamic Discovery Service-
Based JINI for the Next Generation Vehicle
Network. Proceedings of the 2006 SAE World
Congress, Detroit, Michigan, Paper Number
2006-01-1329.
Baroody, R., Al-Holou, N., & Rashid, A. (2005).
Next Generation Vehicle Network (NGVN): Per-
formance Analysis of Dynamic Discovery Service
Protocol utilized for Internet Access. Proceedings
of the 2005 SAE World Congress, Detroit, Michi-
gan, Paper Number 2005-05AE-237.
Bell, R. (1976). Multiplexing – Past, Present, and
Future. SAE Paper Number 760178, Society of
Automotive Engineers, Warrendale, PA, 1976.
Chen, C-H., & Lin, F-F. (1999). An Easy-to-Use
Approach for Practical Bus-Based System De-
sign. IEEE Transactions on Computer, 48(8),
780-793.
Das, C. R., & Bhuyan, L. (1985, October). Band-
width availability of multiple-bus multiproces-
sors. IEEE Transactions on Computers, 34(7),
918-926.
301
 In-Vehicle Network Architecture for the Next-Generation Vehicles

Firlit, T., Sandoval, J., & Ellerbrock, P. (2004). Congress, Detroit, Michigan, Paper Number
IntelliBusTM Protocol:Flexible and Cost-Effec- 2005-01-1531.
tive Automotive Bus. Techincal Report, Aeroflex
Mahmud, S. M., & Shanker, S. (2006, May). An
Colorado Springs.
In-Vehicle Secure Wireless Personal Area Net-
FlexRay Consortium (2004). FlexRay Commu- work (SWPAN). IEEE Transactions on Vehicular
nications System Protocol Specification. Version Technology, 55(3), 1051-1061.
2.0.
Molloy, M. K. (1989). Fundamentals of Per-
Fuehrer, T., Mueller, B., Hartwich, F., & Hugel, R. formance Modeling. Macmillan Publishing
(2001). Time-Triggered CAN (TTCAN). Proceed- Company.
ings of the 2001 SAE World Congress, Detroit,
Motorola (1998). Bosch Controller Area Network
Michigan, Paper Number 2001-01-0073.
Version 2.0 Protocol Standard.
Hossain, I., & Mahmud, S. M. (2007). Analysis
Rabadi, N. M., & Mahmud, S. M. (2008). Driv-
of Group Key Management Protocols for Secure
ers’ Anonymity with a Short Message Length for
Multicasting in Vehicular Software Distribution
Vehicle-to-Vehicle Communications Network.
Network. Proceedings of the Third IEEE Inter-
Proceedings of the 5th IEEE Consumer Com-
national Conference on Wireless and Mobile
munications and Networking Conference, Las
Computing, Networking and Communications
Vegas, NV, January 10-12, 2008.
(WiMob 2007), Crowne Plaza Hotel, White Plains,
New York, USA, October 8-10, 2007. Rabadi, N. M., & Mahmud, S. M. (2007). Privacy
Protection Among Drivers in Vehicle-to-Vehicle
Hossain, I., & Mahmud, S. M. (2007). Analysis of
Communication Networks. Proceedings of the
a Secure Software Upload Technique in Advanced
4th IEEE Consumer Communications and Net-
Vehicles using Wireless Links. Proceedings of
working Conference, Las Vegas, NV, January
the 2007 IEEE Intelligent Transportation Systems
11-13, 2007.
Conference, Seattle, WA, USA, Sept. 30 - Oct. 3,
2007, pp. 1010-1015. TTTech (2002). Time-Triggered Protocol TTP/C
High-Level Specification Document Specification,
Lupini, C. A. (2004). Vehicle Multiplex Com-
Document number D-032-S-10-028.
munication: Serial Data Networking Applied to
Vehicular Engineering. Warrendale, Pensylvania: Uplap, R., Okhade, M., Ghanekar, P., Panse,
SAE. ISBN 0-7680-1218-X. O., Narayan, S., & Sable, A. (2004). LIN Pro-
tocol –Technology Review and Demonstration
Mahmud. S. M. (1994, July). Performance Analy-
in Power Window Application. Proceedings of
sis of Multilevel Bus Networks for Hierarchical
the 2004 SAE World Congress, Paper Number
Multiprocessors. IEEE Transactions on Comput-
2004-28-0019.
ers, 43(7), 789-805.
USDOT (2006, December). Transportation Sta-
Mahmud, S. M., & Alles, S. (2005). In-Vehicle
tistics Annual Report, U.S. Department of Trans-
Network Architecture for the Next-Generation
portation, Bureau of Transportation Statistics.
Vehicles. Proceedings of the 2005 SAE World

302
 303

Compilation of References

802.11 (1999). Technical Specification. Retrieved from Abuhadrous, I., Nashashibi, F., Laurgeau, C., & Chin-
ref http://standards.ieee.org/getieee802/download. chole, M. (2004, June). Multi-Sensor Data Fusion
for Land Vehicle Localization Using MAPS TM. Paper
A*CAR. (2008). Automotive @ A*STAR. Brochure
presented at the IEEE Intelligent Vehicle Conference,
from Agency for Science, Technology and Research
Columbus, Ohio, USA.
(A*STAR).
Afkhamie, K. H., Katar, S., Yonge, L., & Newman, R. A.
AALCAR. (2008). Automotive CAN Network Diagnos-
(2005). Overview of the upcoming HomePlug AV Stan-
tics. Retrieved October 13, 2008, from http://www.aa1car.
dard. Power Line Communications and Its Applications,
com/library/can_systems.htm
International Symposium, 6-8 April, (pp. 400- 404).
Abbott, E., & Powell, D. (1999). Land-Vehicle Navigation
Ahmed, K., Moshe, E., Koutsopoulos, H., & Mishalani,
Using GPS. IEEE Proceedings of the Institute of Electri-
R. (1996). Models of freeway lane changing and gap ac-
cal and Electronics Engineer, 1(87), 145-162.
ceptance behavior. In Proc. ISTTT ’96 (pp. 501–515).
Abdessamie, R., & Gingras, D. (2008). A collaborative
Aijaz, A., Bochow, B., Dötzer, F., Festag, A., Gerlach,
navigation approach in intelligent vehicles. Paper pre-
M., Kroh, R., & Leinmüuller, T. (2006). Attacks on in-
sented at the Society of Automotive Engineering (SAE)
tervehicle communication systems - an analysis.In 3nd
Conference, paper no 1249 Detroit USA.
International Workshop on Intelligent Transportation,
Abdulhamid, H., Abdel-Raheem, E., & Tepe, K. E. (2007). Hamburg.
Channel estimation for 5.9 Ghz dedicated shortrange com-
Akcelik, R., & Besley, M. (2003, December). Operating
munications receiver in wireless access vehicular environ-
cost, fuel consumption, and emission models in aaSIDRA
ments. IET Communications Journal, 1(6), 1274-1279.
and aaMOTION. Paper presented at the 25th Conference
Abe, M., & Mokhiamar, O. (2007). An integration of of Australian Institute of Transport Research. Univ of
vehicle motion controls for full drive-by-wire vehicle. South Australia, Adelaide.
Proc. IMechE Part K: J. Multi-body Dynamics, 221,
Akyildiz, I. F., Su, W., Sankarasubramaniam, Y., &
117-127.
Cayirci, E. (2002). Wireless sensor networks: a survey.
Abidi, M. A., & Gonzalez, R. C. (1992). Data Fusion Computer Networks, 38 (4), 393–422.
In Robotics And Machine Intelligence. NY, USA: Aca-
Anda, J., LeBrun, J., Chuah, C.-N., Ghosal, D., & Zhang,
demic Press.
H. M. (2005). VGrid: Vehicular ad hoc networking and
Abramson, N. (1970). The ALOHA system – Another computing grid for intelligent traffic control. In Proc.
alternative for computer communications, AFIPS Con- IEEE VTC 2005-spring, 5, 2905–2909.
ference (pp. 295-298).

Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.

Anderson, B., & Moore, J. (1979). Optimal Filtering.
Englewood, NJ: Prentice Hall.
Andrieu, C., & Doucet, A. (2002). Particle filtering for
partially observed Gaussian state space models. Journal
Royal Statistical Society B., 4(64), 827–836.
Anon (2003, December). Cutting down on the chips.
Automotive Engineer, (pp. 48-51).
Anwar, S. (2005). Generalized Predictive Control of Yaw
Dynamics of a Hybrid Brake-By-Wire Equipped Vehicle.
International Journal of Mechatronics, 15(November),
1089-1108.
Anwar, S. (2006). Anti-Lock Braking Control of a Hybrid
Brake-By-Wire System. Journal of Automobile Engineer-
ing – IMechE Proceedings Part D., 220(8), 1101-1117.
Anwar, S., & Chen, L. (2007). An Analytical Redun-
dancy Based Fault Detection and Isolation Algorithm
for a Road-Wheel Control Subsystem in a Steer-By-Wire
System. IEEE Transactions on Vehicular Technology,
56(5), 2859-2869.
Ardeshiri, T. et al. (2006). Offset Eliminative Map
Matching Algorithm for Intersection Active Safety Ap-
plications. Proceedings Intelligent Vehicles Symposium
IV2006, Tokyo.
Arem, B. van, Driel, C. van., & Visser, R. (2006). The
Impact of Cooperative Adaptive Cruise Control on
Traffic-Flow Characteristics. IEEE Transactions On
Intelligent Transportation Systems, 7(4).
Armknecht, F., Festag, A., Westhoff, D., & Zeng, K.
(2007). Cross-layer privacy enhancement and non-
repudiation in vehicular communication. In WMAN
2007, Bern.
Armstrong, L. (2008). What is DSRC? Retrieved
September 2, 2008, from http://www.leearmstrong.
com/DSRC/DSRCHomeset.htm
Artimy, M., Robertson, W., & Phillips, W. (2005b).
Assignment of dynamic transmission range based on
estimation of vehicle density. In VANET’05, Cologne,
Germany, Sep., 2.
304
Compilation of References
Artimy, M., Robertson, W., & Phillips, W. (2006). Mini-
mum transmission range in Vehicular Ad Hoc Networks
over uninterrupted highways. In The 9th International
IEEE Conference on Intelligent Transportation Systems,
Toronto, Canada, Sep., 17-20.
Arulampalam, S., Maskell, S., Gordon N., & Clapp T.
(2002). A Tutorial on Particle Filters for Online Nonlin-
ear/Non-Gaussian Bayesian Tracking. IEEE Transactions
on Signal processing, 50(2), 174-188.
Ateniese, G., & Tsudik, G. (1999). Group signatures à-
la-carte. In SODA ‘99: Proceedings of the tenth annual
ACM-SIAM symposium on Discrete algorithms, (pp.
848-849), Philadelphia, PA, USA, Society for Industrial
and Applied Mathematics.
Audi CAN-bus. (2008). Trainer Information. Edition 1
Automotive Systems Research Group. (2008). Retrieved
from http://www.ttcan.com/ASRG
AUTOSAR. (2008). Retrieved September 2, 2008, from
http://www.autosar.org
Bai, F., Sadagopan, N., & Helmy, A. (2003, April).
Important: A framework to systematically analyze the
Impact of Mobility on Performance of Routing Protocols
for Ad-hoc Networks. IEEE INFOCOM, 2, 825-835.
Bajpai, G., Chang, B.C., & Lau, A. (2001, September).
Reconfiguration of Flight Control Systems for Actuator
Failures. IEEE AESS Systems Magazine, (pp. 29-33).
Bando, M., Hasebe, K., Nakayama, A., Shibata, A., &
Sugiyama, Y. (1995). Phys. Rev. E, 51(2), 1035–1042.
Bantli, H., Ring, E., & Goff, E. (1997). Requirements
for direct short range vehicle to roadside communica-
tions. Proceedings of the IEEE Aerospace Conference,
1, 41-61.
Baranowski, S., Lienard, M., & Degauque, P. (1991, Feb-
ruary). Beacon-vehicle link in the 1-10 GHz frequency
range. DRIVE Conference (pp. 194-217).
Baroody, R., Al-Holou, N., & Mohammad, U. (2006).
Reliable Dynamic Discovery Service-Based JINI for
the Next Generation Vehicle Network. Proceedings of
Compilation of References
the 2006 SAE World Congress, Detroit, Michigan, Paper
Number 2006-01-1329.
Baroody, R., Al-Holou, N., & Rashid, A. (2005). Next
Generation Vehicle Network (NGVN): Performance
Analysis of Dynamic Discovery Service Protocol uti-
lized for Internet Access. Proceedings of the 2005 SAE
World Congress, Detroit, Michigan, Paper Number
2005-05AE-237.
Barr, R. (2004). An efficient, unifying approach to
simulation using virtual machines. PhD thesis, Cornell
University, US.
Basagni, S., Chlamtac, I., Syrotiuk, V. R., & Woodward.,
B. A. (1998). A distance routing effect algorithm for
mobility (DREAM). In Proceedings of MobiCom ‘98
(pp. 76-84). Dallas, TX.
Baumann, R., Legendre, F., & Sommer, P. (2008). Generic
Mobility Simulation Framework (GMSF). In Proceed-
ing of the 1st ACM SIGMOBILE workshop on Mobility
models (pp 49-56). Hong Kong, China
Beattie, J. B. (1998). Self-localisation in the senario au-
tonomous wheelchair. Journal of Intelligent and Robotic
Systems, 22, 255-267.
Beikirch, H., & Voss, M. (2000). CAN-transceiver for
field-bus power line communication. ISPLC Symposium,
Limerick, Ireland, April, (pp. 257–264).
Bell, R. (1976). Multiplexing – Past, Present, and Fu-
ture. SAE Paper Number 760178, Society of Automotive
Engineers, Warrendale, PA, 1976.
Bellare, M., Micciancio, D., & Warinschi, B. (2003).
Foundations of group signatures: formal definition,
simplified requirements and a construction based on
trapdoor permutations. In Eli Biham, (Ed.), Advances
in cryptology - EUROCRYPT 2003, vol. 2656 of Lecture
Notes in Computer Science, (pp. 614-629), Warsaw,
Poland, Springer-Verlag.
Bellare, M., Shi, H., & Zhang, C. (2004). Foundations
of group signatures: The case of dynamic groups.
In CT-RSA’05, Lecture Notes in Computer Science,
Springer-Verlag.
Ben-Akiva, M., Davol, A., Toledo, T., Koutsopoulos, H.
N., Burghout, W., Andréasson, I., Johansson, T., & Lun-
din, C. (2002). Calibration and evaluation of mitsimlab in
stockholm. In Procedings of 81st Transportation Research
Board Meeting, Washington, DC, USA, Jan. 13-17.
Berefelt, F., Boberg, B., Nygards, J., Stromback, P., &
Wirkander, S.-L. (2004). Collaborative GPS/INS Naviga-
tion in Urban Environment. Swedish Defence Research
Agency ION-NTM FOI, (pp. 1114-1125). Sweden.
Bertsekas, D., & Gallager, R. (1992). Data networks.
Englewood Cliffs, NJ: Prentice Hall.
Bevas, R., & Bevas, V. (1997). Routing in ad hoc networks
using a spine. Proceedings in the Sixth International
Conference in Computer and Communication Networks,
(pp.34–39).
Bevly, D. M. (2004). Global positioning system (GPS):
A low-cost velocity sensor for correcting inertial sensor
errors on ground vehicles. J. Dyn. Syst. Meas. Contr.,
126(2), 255–264.
Biswas, P., & Ye Y. (2004, April). Semidefinite Program-
ming for Ad Hoc Wireless Sensor Network Localization.
Presented at the IPSN-04 Conference on Information
Processing in Sensor Networks, Berkeley, CA, USA,
(pp. 46-54).
Biswas, S., Tatchikou, R., & Dion, F. (2006). Vehicle-to-
vehicle wireless communication protocols for enhanc-
ing highway traffic safety. In IEEE Communications
magazine, 44 , 74-82).
Blough, D. M., Leoncini, M., Resta, G., & Santi, P. (2006).
The k-neighbors approach to interference bounded and
symmetric topology control in ad hoc networks. IEEE
Transactions on Mobile Computing, 5(9), 1267-1282.
Bluetooth (2008). Retrieved from http://www.bluetooth.
org/
Bluetooth. (2008). Bluetooth standard. Retrieved October
13, 2008, from http://www.bluetooth.com
Blum, J., & Eskandarian, A. (2004) The threat of intel-
ligent collisions. IT Professional, 6, 24-29.
305

Blum, J., Eskandarian, A., & Hoffman, L. (2004). Chal-
lenges of intervehicle ad hoc networks. IEEE Transactions
on Intelligent Transprotation Systems, 5(4), 347-351.
BMBF. Network on Wheels. Retrieved June 8, 2008,
from: http://www.network-on-wheels.de
Böhm, A. (2007). State of the art on network layer
aspects for inter-vehicle communication (Tech. Rep.
IDE 0748).
Bonanno, G., Fantoni, R., Fichera, A., Fornetti, G.,
Moriconi, C., Poggi, C., et al. (2003). The sensing sub-
system of RAS.
Boneh, D., & Shacham, H. (2004). Group signatures with
verifier-local revocation. In CCS ‘04, (pp. 168–177), New
York, NY, USA, ACM Press.
Boneh, D., Boyen, X., & Shacham, H. (2004). Short group
signatures, In Crypto ‘04, LNCS 3152 (pp. 41-55).
Bonnifait, P., Bouron, P., Meizel, D., & Crubille, P.
(2003). Dynamic Localization of Car-like Vehicles using
Data Fusion of Redundant ABS Sensors. The Journal of
Navigation, 3(56), 429–441. UK.
Bononi, L., Di Felice, M., Bertini, M., & Croci, E.
(2006). Parallel and Distributed Simulation of Wireless
Vehicular Ad Hoc Networks. In Proceedings of the 9th
ACM international symposium on Modeling analysis and
simulation of wireless and mobile systems (pp 28-35).
Terromolinos, Spain
Borgonovo, F., Capone, A., Cesana, M., & Fratta, L.
(2004, July) ADHOC MAC: A new MAC architecture
for ad hoc networks providing efficient and reliable
point-to-point and broadcast services. ACM Wireless
Networks, 10, 359-366.
Boris & Arkady (1999). Link state routing protocol with
cluster based flooding for mobile ad-hoc computer net-
works. Proceedings of the Workshop on Computer Science
and Information Technologies CSIT’99, (pp.28–36).
Bose, A., & Ioannou, P. A. (2003). Analysis of traffic flow
with mixed manual and semi-automated vehicles. IEEE
Trans. Intell. Transport. Syst., 4(4), 173–188.
306
Compilation of References
Brasche, G., Rokitansky, C.-H., & Wietfeld, C. (1994).
Communication Architecture and Performance Analysis
of Protocols for RTT Infrastructure Networks and Vehicle-
Roadside Communications. Proceedings of the IEEE 44th
Vehicular Technology Conference, 1, 384-390.
Brickell, E., Camenisch, J., & Chen, L. (2004). Direct
anonymous attestation. In CCS ‘04, (pp. 132-145), New
York, NY, USA, 2004. ACM Press.
Broch, J., Maltz, D. A., Johnson, D. B., Hu, Y. C., &
Jetcheva, J. G. (1998). A Performance Comparison of
Multi-HopWireless Ad Hoc Network Routing Protocols.
In Proceedings of MobiCom’98 (pp. 85-97). Dallas,
TX.
Broch, J., Maltz, D., Johnson, D., Hu, Y.-C., & Jetcheva J.
(1998). A performance comparison of multi-hop wireless
ad hoc networks routing protocols. In Proceedings of the
4th ACM/IEEE International Conference on Mobile Com-
puting and Networking (MOBICOM), (pp. 85-97).
Broughton, J., & Walter, L. (2007). Trends in fatal car
accidents: Analysis of CCIS Data (Published project
report PPR 172). TRL Limiter.
Brown, M., Hankerson, D., Lopez, J., & Menezes, A.
(2001). Software implementation of the NIST elliptic
curves over prime fields. In CT-RSA 2001, (pp. 250-265),
London, UK. Springer-Verlag.
Brown, R. G., & Hwang, P. C. (1997). Introduction to
Random Signals and Applied Kalman Filtering. New
York, NY, USA: John Wiley and Sons.
Burkhart, M., Rickenbach, P. von, Wattenhofer, R., &
Zollinger, A. (2004). Does topology control reduce inter-
ference. In MobiHoc’04, Roppongi, Japan, May 24-26.
C2C. (2008). European based Car-to-Car communica-
tion Consortium. Retrieved October 13, 2008, from
http://www.car-to-car.org
C2C-CC. Retrieved June 8, 2008, from: http://www.car-
to-car.org
Calandriello, G., Papadimitratos, P., Hubaux, J.-P., & Lioy,
A. (2007). Efficient and robust pseudonymous authen-
tication in VANET. In Proceedings of the Fourth ACM
Compilation of References
international Workshop on Vehicular Ad Hoc Networks
(Montreal, Quebec, Canada, September 10th, 2007).
VANET ‘07. ACM, New York, NY, (pp. 19-28).
California PATH. Retrieved September 2, 2008, from
http://www.path.berkeley.edu
CALM. (2008). CALM Continuous Communications for
Vehicles. Retrieved September 2, 2008, from http://www.
calm.hu/
CAN. (2001). Retrieved from official web site http://www.
can-cia.org/
Cannon, M. E., Nayak, R., Lachapelle, G., Salychev, O.
S., &Voronov, V. V. (2001) Low-cost INS/GPS Integra-
tion: Concepts and Testings, Journal of Navigation,
54(1), 119-134.
Car 2 Car Communication Consortium. (2008). Retrieved
September 2, 2008, from http://www.car-to-car.org
Caron, F., Davy, M., Duflos, E., & Vanheeghe, P. (2007).
Particle Filtering for Multisensor Data Fusion with
Switching Observation Models: Application to Land
Vehicle Positioning. IEEE Transactions on Signal Pro-
cessing, 6(55), 2703-2719.
Caron, F., Duflos, E., Pomorski, D., & Vanheeghe, P.
(2006). GPS/IMU Data Fusion using Multisensor Kalman
Filtering: Introduction of Contextual Aspects. Informa-
tion Fusion, 7(2), 221-230. Holland: Elsevier.
Caron, F., Smets, P., Duflos, E., & Vanheeghe, P. (2005,
July). Multisensor data fusion in the frame of the TBM
on reals. Application to land vehicle positioning. Paper
presented at the International Conference on Information
Fusion (FUSION’05), Philadelphia, PA, USA.
CarTALK 2000. Communication Architecture. Retrieved
June 8, 2008, from: http://www.cartalk2000.net
Chan, Y-T, Tsui, W-Y., So, H-C. & Ching, P-C. (2006),
Time-of-Arrival Based Localization Under NLOS Con-
ditions. IEEE Transactions on Vehicular Technology,
1(55), 17-24.
Chang, J.-M., & Maxemchuk, N. F. (1984). Reliable
Broadcast Protocols. ACM Transactions on Computer
Systems, 2, 251-273.
Chang, S. C. (2007). Adoption of state feedback to control
dynamics of a vehicle with a steer-by-wire system. Proc.
IMechE Part D: J. Automobile Engineering, 221, 1-12.
Chaum, D. (1985). Security without identification: Trans-
action systems to make big brother obsolete. Commun.
ACM, 28(10), 1030-1044.
Chaum, D., & van Heyst, E. (1991). Group signatures.
In EUROCRYPT, (pp 257-265).
Chen, C., Jia, Z., & Varaiya, P. (2001). Causes and cures
of highway congestion. IEEE Control Syst. Mag., 21(6),
26–32.
Chen, C., Petty, K., Skabardonis, A., Varaiya, P., & Jia,
Z. (2001). Freeway performance measurement system:
Mining loop detector data. Transp. Res. Rec., 1748,
96–102.
Chen, C-H., & Lin, F-F. (1999). An Easy-to-Use Ap-
proach for Practical Bus-Based System Design. IEEE
Transactions on Computer, 48(8), 780-793.
Chen, Q., Schmidt-Eisenlohr, F., Jiang, D., Torrent-
Moreno, M., Delgrossi, L., & Hartenstein, H. (2007).
Overhaul of IEEE 802.11 modeling and simulation in
ns-2. In Proceedings of the 10th ACM Symposium on
Modeling, analysis, and simulation of wireless and
mobile systems (Chania, Crete Island, Greece, October
22-26, (pp 159-168).
Chen, Y., & Chen, L. (2005). Using Bluetooth wireless
technology in vehicles. IEEE International Conference
Vehicular Electronics and Safety, 14-16 October, (pp.
344- 347).
Chennikara-Varghese, J., Chen, W., Altintas, O., & Cai,
S. (2006, July). Survey of routing protocols for inter-ve-
hicle communications. IEEE Workshop on Mobile and
Ubiquitous Systems Conference (pp. 1-5).
Ching-Chuan, Hsiao-Kuang, W., & Mario, G. (1997). Rout-
ing in clustered multihop, mobile wireless networks with
fading channel. IEEE Singapore International Conference
on Networks, (pp. 197–211).
Choffnes D. R., & Bustamante F. E. (2005). An Inte-
grated Mobility and Traffic Model for Vehicular Wireless
307

Networks. In Proceedings of the 2nd ACM International
Workshop on Vehicular Ad Hoc Networks (pp. 69-78).
Cologne, Germany.
Choffnes, D. R., & Bustamante, F. E. (2005). An inte-
grated mobility and traffic model for vehicular wireless
networks. In Proc. VANET ’05 (pp. 69–78). New York,
NY, USA: ACM.
Chowdhury, D., Santen, L., & Schadschneider, A. (2000).
Statistical physics of vehicular traffic and some related
systems. Physics Reports, 329, 199.
Chu, L., Liu, H. X., Recker, W., & Zhang, H. M. (2004).
Performance evaluation of adaptive rampmetering
algorithms using microscopic traffic simulation model.
J. Transp. Engrg., 130(3), 330–338.
Chuah, T. C., & Sharif, B. S. (2001). Robust adaptive
spread-spectrum receiver with neural-net preprocess-
ing in non-gaussian noise. IEEE Transaction on Neural
Networks, 3(12), 546-558.
CI-6558, P. (n.d.). Acceleration sensor. (http://hug.
phys.huji.ac.il/PHYS HUG/MAABADA/MABADA
A/Pasco/ci/ci-6558/012-06361b/012-06361b.pdf, last
accessed: April 2008).
CiA. (2008). CAN in Automation. Retrieved October 13,
2008, from http://www.can-cia.org/
CiA. (2008). CAN-based in-vehicle networks. Retrieved
September 2, 2008, from http://www.can-cia.de/index.
php?id=228
CISPR25. (2002). Limits and methods of measurement
of radio disturbance characteristics for the protection
of receivers used on board vehicles. Technical report,
CISPR-25, 2002.
Controller Area Network. (2008). Retrieved September
2, 2008, from http://www.can-cia.org/
Cook, J. A., Kolmanovsky, I. V., McNamara, D., & Prasad,
K. V. (2007). Control, Computing, and Communications:
Technology for the twenty-first century Model T. Pro-
ceedings of the IEEE Special Issue Advanced Automobile
Technologies, 95(2).
308
Compilation of References
CORSIM (2008). Microscopic Traffic Simulation Model.
Retrieved March 19, 2008, from http://mctrans.ce.ufl.
edu/featured/tsis/version5/corsim50.htm
Cottingham, D. N., Wassell, I. J., & Harle, R. K. (2007).
Performance of IEEE 802.11a vehicular context. 65th IEEE
Vehicular Technology Conference, (pp. 854-858).
Crawford, M., & Workman, J. (1980). Predicting free-
space radiated emissions from electronic equipment
using TEM cell and open-field site measurements. Proc.
IEEE Int. Symp. on Electromagn. Compat., (pp. 80-85).
Baltimore, MD.
Crawford, S., & Cannon, M. E. (2005). Performance
Evaluation of Sensor Combinations for Mobile Platoon
Control; Positioning, Location, and Navigation (PLAN)
Group, (Tech. Rep.) University of Calgary, Calgary
Alberta, Canada
Crisan, D., & Doucet, A. (2002). A survey of theoreti-
cal results on particle filtering for practitioners,” IEEE
Transaction on. Signal Processing, 3(50), 736-746.
Crowther, W., Rettberg, R., Walden, D., Ornstein, S., &
Heart, F. (1973). A system for broadcast communication:
Reservation-ALOHA. Hawaii International Conference
on System Sciences.
Cui, Y. J., & Ge, S. S. (2002). Autonomous vehicle posi-
tioning with GPS in Urban Canyon environments, IEEE
Transactions in Robotics and Automation, 19.
Cvetkovski, B., & Gavrilovska, L. (1998). A simulation
of a mobile Highway traffic. Proceedings of the IEEE
Vehicular Technology Conference, 2, 1429-1433.
CVIS Project Partners. (2008). Retrieved June 2008,
from http://www.cvisproject.org
Da Chen, Z., Kung, H.-T., & Vlah, D. (2001). Ad Hoc Relay
Wireless Networks over Moving Vehicles on Highways.
Proceedings in International Conference on Mobile
Computing and Networking, (pp. 247-250).
Dammeyer, J. (2008). Wireless Controller Area Network.
Retrieved September 2, 2008, from http://www.autoarti-
sans.com/documents/canrf_prod_announcement.pdf
Compilation of References
Daniels, J. (2005). Practicalities. Automotive Engineer,
30(10), 33.
Das, C. R., & Bhuyan, L. (1985, October). Bandwidth
availability of multiple-bus multiprocessors. IEEE
Transactions on Computers, 34(7), 918-926.
Dashtinezhad, S., Nadeem, T., Dorohonceanu, B., Borcea,
C., Kang, P., & Iftode L. (2004). TrafficView: A driver
assistant device for traffic monitoring based on car-to-
car communication, In Proceedings of IEEE Semiannual
Vehicular Technology Conference. Milan, Italy.
Davis, B. (2001). Making the right connections. Profes-
sional Engineering, February 28, (pp. 36-39).
Davis, L. C. (2004). Effect of adaptive cruise control sys-
tems on traffic flow. Physical Review E, 69(066110).
Davis, L. C. (2007). Effect of adaptive cruise control
systems on mixed traffic flow near an on-ramp. Physica
A, 379(274–290).
Dedicated Short Range Communications. (2008). Re-
trieved September 2, 2008, from http://grouper.ieee.
org/groups/scc32/dsrc/
Defense Advanced Research Projects Agency (2008).
Retrieved June 2008, from http://www.darpa.mil/grand-
challenge/index.asp
Degardin, V., Laly, P., Liénard, M., & Degauque, P.
(2006). Impulsive Noise on In-Vehicle Power Lines:
Characterization and Impact on Communication Perfor-
mance. IEEE International Power line Communications
and Its Applications Conference, ISPLC 2006, 26-29
March, (pp. 222-226).
Degardin, V., Laly, P., Liénard, M., & Degauque, P.
(2007). Performances of the HomePlug PHY layer in
the context of in-vehicle powerline communications.
IEEE International Power line Communications and
Its Applications Conference, ISPLC 2007, 26-28 March,
(pp. 93-97).
Deng, S., Hubing, T., & Beetner, D. (2007). Characterizing
the electric field coupling from IC heatsink structures
to external cables using TEM cell measurements. IEEE
Trans. on Electromagn. Compat., 49(4) 785-791.
Department of Transport. (2008). Retrieved June 2008,
from http://www.itslessons.its.dot.gov/
Dia, H., & Purchase, H. (1999). Modeling the impacts of
advanced traveler information systems using intelligent
agents. Road and Transport Research, 8.
Diaconescu, R., Gorgorin, C., Gradinescu, V., Cristea,
V., & Iftode, L. (2007). Collaborative Route Planning
in Vehicular Ad-hoc Networks. In Proceedings of the
CSCS-16 (pp. 267-274). Bucharest, Romania.
Dickey, S. R., Huang, C., & Guan, X. (2007). Field
Measurements of Vehicle to Roadside Communication
Performance. 66th IEEE Vehicular Technology Confer-
ence, (pp. 2179-2183).
DICKEY-John Products. (n.d.). Radar II. (http://www.
dickey-john.com/products/public-works/ground-speed-
sensors/radar-ii, last accessed: April 2008)
Dobias, F., & Grabow, W., (1994). Adaptive array antenna
for 5.8 Ghz vehicle to roadside communication. 44th IEEE
Vehicular Technology Conference, 3, 1512-1516.
Doherty, L., Ghaoui, L. E., & Pister, K. (2001). Convex
position estimation in wireless sensor networks. Paper
presented at the INFOCOMM Conference., (3), 1655-
1663. Anchorage, Alaska, USA
Doherty, L., Pister, K. S. J., & El Ghaoui, L. (2001, April).
Convex Position Estimation in Wireless Sensor Networks.
Paper presented at the INFOCOM 2001. 20TH Annual Joint
Conference of the IEEE Computer and Communications
Societies. Proceedings (pp. 1655-1663).
Doucet, A., & Wang, X. (2005). Monte Carlo methods
for signal processing: A review in the statistical signal
processing context. IEEE Signal Processing Magazine,
6(22), 152-170.
Drane, C., & Rizos, C. (1997). Positionng systems in
intelligent transportation systems. Artech House, Bos-
ton, Ma., USA.
Dridi, S., Gouissem, B., Hasnaoui, S., & Rezig, H. (2006).
Coupling Latency Time to the Throughput Performance
Analysis on Wireless CAN Networks. In Prof. of the
International Multi-Conference on Computing in the
Global Information Technology (ICCGI’06).
309

DSRC (2006). Dedicated Short Range Communications.
Retrieved September, 2008, from http://grouper.ieee.
org/groups/scc32/dsrc/index.html.
DSRC. (2008). Dedicated Short Range Communications.
Retrieved October 13, 2008, from http://grouper.ieee.
org/groups/scc32/dsrc/
Du, L., & Ukkusuri, S. (2008). Optimization models to
characterize the broadcast capacity of vehicular ad hoc
networks. Transportation Resaerch Part C.
Du, L., & Ukkusuri, S. (2008). The relative mobility
of vehicles improves the performance of information
flow in vehicle ad hoc networks. Networks and Spatial
Economics.
Du, L., Ukkusuri, S., & Kalyanaraman, S. (2007).
Geometric connectivity of vehicular ad hoc networks:
Analytical characterization. In The Fourth ACM Work-
shop on Vehicular Ad Hoc Networks (VANET), Montreal,
Canada, Sep. 10.
Dumitrescu, V., & Guo, J. (2005, June). Context assisted
routing protocols for inter-vehicle wireless communi-
cation. IEEE Symposium on Intelligent Vehicles (pp.
594-600).
Durant-Whyte, H. (1994). Where am I? Industrial Robot,
21(2), 11-16. MCB University Press
EASIS. (2007). Retrieved September 2, 2008, from
http://www.easis-online.org
Eichler, S. (2007, October). Performance Evaluation of
the IEEE 802.11p WAVE Communication Standard. IEEE
Vehicular Technology Conference (pp. 2199-2203).
Eichler, S., Ostermaier, B., & Schroth, C. (2005) Simula-
tion of Car-to-Car Messaging: Analyzing the Impact on
Road Traffic. In Proceedings of the 13th IEEE Interna-
tional Symposium on Modeling, Analysis, and Simulation
of Computer and Telecommunication Systems (pp. 507-
510). IEEE Computer Society  Washington, DC, USA
El Zarki, M., Mehrotra, S., Tsudik, G., & Venkatasubra-
manian, N. (2002). Security issues in a future vehicular
network. In Proceedings of EuroWireless 2002
310
Compilation of References
ElBatt, T., Goel, S. K., Holland, G., Krishnan, H., &
Parikh, J. (2006). Safety: Cooperative collision warning
using dedicated short range wireless communications.
In Proc. VANET ’06 (pp. 30-39). New York, NY, USA:
ACM.
ElBatt, T., Goel, S., Holland, G., & Krishnan, H. (2006).
Cooperative collision warning using dedicated short
range communications. In Proceedings of the Third ACM
International Workshop on Vehicular Ad Hoc Networks
(Los Angeles, CA, USA, September 29th, 2007). VANET
‘06. ACM, New York, NY, (pp. 1-9).
Elliott, J. D., Langlois, D., & Croft, E. A. (2001, June).
Sensor Uncertainty Management for an Encapsulated
Logical Device Architecture: Part I. American Control
Conference, Arlington, USA, Proceedings (pp. 4282-
4387).
Erbas, F., Garcia, J.-E., & Jobmann, K. (2004). Position-
based QoS routing in mobile ad-hoc networks: problem
statement and a novel approach. IEEE Performance,
Computing, and Communications Conference (pp.
619-623).
ERTICO-ITS Europe. PReVENT Project. Retrieved June
8, 2008, from: http://www.prevent-ip.org
eSafety. (2008). Retrieved September 2, 2008, from
http://ec.europa.eu/information_society/activities/es-
afety/index_en.htm
Estrin, D., Govindan, R., Heidemann, J., & Kumar, S.
(1999). Next century challenges: scalable coordination
in sensor networks. In MobiCom ’99 (pp. 263–270). New
York, NY, USA: ACM.
European Commission. (2008). European Road Safety
Day on 13 October 2008. Retrieved September 2, 2008,
from http://ec.europa.eu/transport/roadsafety/road_safe-
ty_days/index_2008_en.htm
Eymann, T. (2006). The EASIS Security Architecture
Approach. 1st C2C-CC Security Workshop. Berlin,
November 2006.
Farrell, J., & Barth, M. (1999). The Global Positioning
System and Inertial Navigation. NY, USA: McGraw-
Hill.
Compilation of References
Federal Highway Administration. (2005). Traffic conges-
tion and reliability: Trends and advanced strategies for
congestion mitigation. (http://ops.fhwa.dot.gov/conges-
tion report/, last accessed: April 2008)
Fellendorf, M., & Vortisch, P. (2001, January). Valida-
tion of the Microscopic Traffic Flow Model VISSIM in
Different Real-World Situations. Paper presented at the
80th Meeting of the Transportation Research Board.
Washington, D.C.
Fennel, H., & Bunzel, S. (2006). Achievements and ex-
ploitation of the AUTOSAR development partnership,
Copyright © 2006 CTEA. Retrieved from official web
site http:www.autosar.org
Festa, D. C., Longo, G., Mazzulla, G., & Musolino, G.
(2001). Experimental analysis of different simulation
models for motorway traffic flow. Proceedings of the
IEEE Intelligent Transportation Systems Conference,
(pp. 675-680).
Firlit, T., Sandoval, J., & Ellerbrock, P. (2004). Intelli-
BusTM Protocol:Flexible and Cost-Effective Automotive
Bus. Techincal Report, Aeroflex Colorado Springs.
Fleming, W. J. (2001). Overview of automotive sensors.
IEEE Sens. J., 1(4), 296–308.
FlexRay Consortium (2004). FlexRay Communications
System Protocol Specification. Version 2.0.
FlexRay Consortium. (2005). FlexRay Communications
System, Physical Layer, EMC Measurement Specifica-
tion Version 2.1.
FlexRay. (2008). FlexRay In-vehicle Network. Retrieved
October 13, 2008, from http://www.interfacebus.com
FlexRay. (2008). Retrieved September 2, 2008, from
http://www.flexray.com/
Ford Sync. (2008). Today’s Drivers Demand Staying Con-
nected. Retrieved September 2, 2008, from http://media.
ford.com/article_display.cfm?article_id=25169
Fowler, D. (2003). Down to the wire. The Engineer,
(pp. 29-30).
Freudiger, J., Raya, M., Feleghhazi, M., Papadimitratos,
P., & Hubaux, J.-P. (2007). Mix Zones for Location
Privacy in Vehicular Networks. In Proceedings of the
First International Workshop on Wireless Networking for
Intelligent Transportation Systems (WiN-ITS 2007)
Fuehrer, T., Mueller, B., Hartwich, F., & Hugel, R. (2001).
Time-Triggered CAN (TTCAN). Proceedings of the 2001
SAE World Congress, Detroit, Michigan, Paper Number
2001-01-0073.
Fujimoto, R. M., Perumalla, K., Park, A., Wu, H., Am-
mar, M. H., & Riley, G. F. (2003) Large-Scale Network
Simulation: How Big? How Fast? In Proceedings of the
11TH IEEE/ACM International Symposium on Modeling,
Analysis and Simulation of Computer Telecommunica-
tions Systems (pp. 116-123). IEEE Computer Society
Washington, DC, USA
Fussen, M., Wattenhofer, R., & Zollinger, A. (2004).
Sensor Networks: interference reduction and possible
application. Diploma Thesis, ETH Zurich, Dept. of
Computer Science.
Füßler, H., Mauve, M., Hartenstein, H., Käsemann, M.,
& Vollmer, D. (2003). MobiCom Poster: Location-Based
Routing for Vehicular Ad-Hoc Networks. Proceedings in
ACM SIGMOBILE Mobile Computing and Communica-
tion Review, 7(1), 47-49.
Gadda, C. D., Laws, S. M., & Gerdes, J. C. (2007). Gen-
erating Diagnostic Residuals for Steer-by-Wire Vehicles.
IEEE Trans. on Control Sys. Tech., 15(3), 529-540.
Gao, S., & Chabini, I. (2006). Optimal routing policy
problems in stochastic time-dependent networks. Trans-
portation Research Part B, 40.
Gartner, N. H., Pooran, F. J., & Andrews, C. M. (2001).
Implementation of the OPAC adaptive control strategy
in a traffic signal network. In Proc. IEEE ITSC ’01 (pp.
195–200).
Gast, M. S. (2002). 802.11 Wireless Networks, the defini-
tive guide. O’Reilly Publishing.
311

Gaubert, N., Beauregard, M., Michaud, F., & de Lafon-
taine, J. (2003, October). Emulation of Collaborative
Driving Systems Using Mobile Robots. Presented at the
IEEE International Conference on Systems, Man and
Cybernetics, 1, 856-861. Washington DC, USA.
Gavette, S., & other HomePlug members HomePlug
Powerline Alliance (2006). HomePlug AV Technology
Overview. Technical report. Sharp Laboratories of Amer-
ica. Retrieved from download.microsoft.com/download.
Gouret, W., Nouvel, F., & EL Zein, G. (2006). Additional
Network Using Automotive Powerline Communication.
IEEE International Power line Communications and Its
Applications Conference, ISPLC 2006, 26-29 March,
(pp 1087-1092).
GEARS. (2008). Google GEARS. Retrieved October 13,
2008, from http://gears.google.com
Gehring, O., & Fritz, H. (1997). Practical results of a
longitudinal control concept for truck platooning with
vehicle to vehicle communication. IEEE Conference on
Intelligent Transportation Systems (pp. 117-122).
Gelb, A. (1974). Applied Optimal Estimation. Cambridge,
MA: The MIT Press.
Gentile, C. (2005, May). Sensor Location through Linear
Programming with Triangle Inequality Constraints. Pa-
per presented at the ICC-IEEE International Conference
on Communications, Proceedings, 5, 3192-3196.
Gerlach, M. (2005). VaneSe - an approach to VANET se-
curity. In V2VCOM 2005, San Diego, California, USA.
Gerlach, M., Festag, A., Leinmüller, T., Goldacker, C.,
& Harsch, C. (2007). Security architecture for vehicular
communication. In 2nd International Workshop on Intel-
ligent Transportation, Hamburg.
Gershenson, C. (2004). Self-organizing traffic lights.
Complex Systems, 16, 29.
Gershenson, C. (2005). Self-Organizing Traffic Lights.
Complex Systems, 16(1), 29-53.
Gharavi H., Prasad K. V., & Ioannou P.A. (Eds.), (2007).
Special issue Advanced Automobile Technologies. Pro-
ceedings IEEE , 95(2).
312
Compilation of References
Gillièron, P. Y., & Konnen, J. (2003). Enhanced Navi-
gation System for Road Telematics. In Proceedings the
Third Swiss Transport Research Conference, (STR 03),
Ascona/Ticino.
Golle, P., Greene, D., & Staddon, J. (2004). Detecting and
correcting malicious data in VANETs. In VANET ‘04,
(pp. 29-37). New York, NY, USA: ACM Press.
Gorgorin, C., Gradinescu, V., Diaconescu, R., Cristea, V.
& Iftode, L. (2006). An Integrated Vehicular and Network
Simulator for Vehicular Ad-hoc Networks. In Proceed-
ings of the Eurosis European Simulation and Modeling
Conference 2006 (pp. 142-149). Toulouse, France.
Gouret, W., Nouvel, F., & EL Zein, G. (2007). High Data
Rate Network Using Automotive Power Line. IEEE Inter-
national Power line Communications and Its Applications
Conference, ISPLC 2007, 26-28 March. (pp. 1-4).
Grabow, W., & Schrei, A. (1992, October). A study
on diversity techniques in a 5.8 GHz microwave link
for traffic management systems. IEEE Symposium on
Personal, Indoor and Mobile Radio Communications
(pp. 556-560).
Gradinescu, V., Gorgorin, C., Diaconescu, R., Cristea,
V., & Iftode, L. (2007). Adaptive Traffic Lights Using
Car-to-Car Communication. In Proceedings IEEE 65th
Vehicular Technology Conference (pp. 21-25). Dublin,
Ireland.
Grewal, M. S., Weill, L. R., & Andrews, A. P. (2001).
Global Positioning Systems, Inertial Navigation, and
Integration. New York, NY, USA: John Wiley & Sons.
Guopei, L., & Gingras, D. (2007, April). Adaptive and
Reconfigurable Data Fusion Architectures in Vehicle
Positioning Navigation Systems. Paper presented at the
Society of Automotive Engineering (SAE) conference.
Paper no. 1108, Detroit, USA.
Haas, Z., & Deng, (2002, June). Dual busy tone multiple
access (DBTMA)-a multiple access control scheme for
ad hoc networks. IEEE Transactions on communica-
tions, 50, 975-985.
Compilation of References
Haerri, J., Filali, F., Bonnet, C., & Fiore, M. (2006).
VanetMobiSim: Generating Realistic Mobility Patterns
for VANETs. In Proceedings of the 3rd international
workshop on Vehicular ad hoc networks (pp 96—97).
Los Angeles, CA, USA
Hagan, M. T., & Menhaj, M. (1994). Training feedforward
networks with the Marquardt algorithm. IEEE Transac-
tions on Neural Networks, 6(5), 989-993.
Haggag, S. Alstrom, D., Cetinkunt, S., & Egelja, A.
(2005). Modeling, Control, and Validation of an Electro-
Hydraulic Steer-by-Wire System for Articulated Vehicle
Applications. ASME/IEEE Trans. on Mechatronics,
10(6), 688-692.
Hall, R. W., & Tsao, H.-S. J. (1997). Capacity of auto-
mated highway systems: merging efficiency. In Proc.
ACC ’97, 3, 2046–2050.
Halle, S., Laumonier, J., & Chaib-Draa, B. (2004). A
decentralized approach to collaborative driving coordina-
tion. Proceedings of the 7th International IEEE Confer-
ence on Intelligent Transportation Systems, Washington,
DC (pp. 453-458).
Hara, S., & Prasad, R. (1997). Overview of multicarrier
CDMA. IEEE Communications Magazine, December,
(pp. 126-133).
Hasan, M. S., & Anwar, S. (2008). Sliding Mode Observer
Based Predictive Fault Diagnosis of a Steer-By-Wire Sys-
tem. Proceedings of the 17th Int’l Federation of Automatic
Control (IFAC) World Congress, Seoul, S. Korea.
Hashimoto, M., Kawashima, H., Nakagami, T., & Oba
F. (2001, October). Sensor Fault detection and Identi-
fication in Dead-Reckoning System of Mobile Robot:
Interacting Multiple Model Approach. Paper presented
at the IEEE/RSJ International Conference on Intelligent
Robot and Systems, Mani, Hawaii, USA.
Hebden, R. G., Edwards, C., & Spurgeon, S. K. (2004).
Automotive steering control in a split-mu maneuver us-
ing an observer based sliding mode controller. Vehicle
System Dynamics, 41(3), 181-202.
Heddebaut, M., Rioult, J., Ghys, J. P., Gransart, Ch., &
Ambellouis, S. (2005). Broadband vehicle-to-vehicle
communication using an extended autonomous cruise
control sensor. Measurement Science and Technology,
16, 1363-1373.
Heffernan, D., & Leen, G. (2008). ICT based re-
search at Limerick contributes to automotive ‘drive-
by-wire’ technology. Retrieved September 2, 2008,
from http://www.irishscientist.ie/2002/contents.
asp?contentxml=02p237b.xml&contentxsl=is02pages.
xsl
Heide, F. M. A. D., Schindelhauer, C., Volbert, K., &
Gruenewald, M. (2002). Energy, congestion and dilation
in radio networks. Proceedings of the 14th Annual ACM
Symposium on Parallel Algorithms and Architectures
(SPAA), Winnipeg, Manitoba, Canada, Aug. 11-13.
Heiskala, J., & Terry, J. (2002). OFDM wireless LANS: A
Theorethical and Practical guide. Sams Publishing.
Helbing, D. (2001). Traffic and related self-driven many-
particle systems. Rev. Mod. Phys., 73(4), 1067–1141.
Hernandez, W. (2005). Robust multivariable estimation
of the relevant information coming from a wheel speed
sensor and an accelerometer embedded in a car under
performance tests. Sensors, 5, 488–508.
Hidas, P. (2005). Modelling vehicle interactions in mi-
croscopic simulation of merging and weaving. Transport.
Res. Part C, 13(1), 37–62.
Higgins, A., & Koucky, S. (2002). Mercedes pumps ‘Fly
By Wire” brakes into new roadster. Machine Design,
(p. 26).
Hightower, J., & Barriello, G. (2001). Location systems for
ubiquitous computing. IEEE Computer, 34(8), 57–66.
Home Plug V1.0 (2008). HomePlug Powerline Alliance.
Retrieved from official web site www.homeplug.org.
Hoogendoorn, S. P., & Bovy, P. H.-L. (2001). State-of-the-
art of Vehicular Traffic Flow Modelling. Special Issue
on Road Traffic Modelling and Control of the Journal of
System and Control Engineering, 215(4), 283-303.
313

Horowitz, R., & Varaiya, P. (2000). Control design of an
automated highway system, Proceedings of the IEEE,
88(7), 913-925.
Hoseinnezhad, R. (2006). Position Sensing in Brake-By-
Wire Calipers Using Resolvers. IEEE Transactions on
Vehicular Technology, 55(3), 924-932.
Hossain, I., & Mahmud, S. M. (2007). Analysis of a Secure
Software Upload Technique in Advanced Vehicles using
Wireless Links. Proceedings of the 2007 IEEE Intelligent
Transportation Systems Conference, Seattle, WA, USA,
Sept. 30 - Oct. 3, 2007, pp. 1010-1015.
Hossain, I., & Mahmud, S. M. (2007). Analysis of Group
Key Management Protocols for Secure Multicasting in
Vehicular Software Distribution Network. Proceedings
of the Third IEEE International Conference on Wireless
and Mobile Computing, Networking and Communica-
tions (WiMob 2007), Crowne Plaza Hotel, White Plains,
New York, USA, October 8-10, 2007.
Hou, T., & Li, V. (2006). Transmission range control in
multihop packet radio network. IEEE transactions on
wireless communications, 5(1), 38-44.
Huaizhi, Li, & Singhal, M. (2005) A scalable routing
protocol for ad hoc networks. In Proceedings of the Ve-
hicular Technology Conference 2005 Spring. 2005 IEEE
61st Volume 4 (pp. 2498—2503). Stockholm, Sweden
Hubaux, J. P., Capkun, S., & Luo, J. (2004). The Secu-
rity and Privacy of Smart Vehicles. IEEE Security and
Privacy Magazine 2(3) (2004), 49-55.
Huck, T., Schirmer, J., & Dostert, K. (2005). Tutorial
about the implementation of a vehicular high speed
communication system. IEEE ISPLC.
i2010 Intelligent Car Initiative. (2008). Retrieved Sep-
tember 2, 2008, from http://ec.europa.eu/information_so-
ciety/activities/intelligentcar/index_en.htm
IDB Forum. (2008). Retrieved from web site http://www.
idbforum.org.
IDB1394, (2008). 1394 Automotive Presentation, Max
Bassler, Shenzen, China, 10 April 2008. Retrieved from
web site http://www.idbforum.org
314
Compilation of References
IEC. (2008). CISPR 12—Vehicles, boats and internal
combustion engines—Radio disturbance character-
istics—Limits and methods of measurement for the
protection of off-board receivers.
IEEE 1363a (2004). IEEE standard specifications for
public-key cryptography- amendment 1: Additional
techniques.
IEEE 1901 (2008). Retrieved from official web http://
grouper.ieee.org/groups/1901/Scope_Purpose.html
IEEE 802.11p. (2008). Retrieved October 13, 2008, from
http://en.wikipedia.org/wiki/IEEE_802.11p
IEEE 802.11p. (2008). Retrieved September 2, 2008, from
http://en.wikipedia.org/wiki/IEEE_802.11p
IEEE. (2006). IEEE Trial-Use Standard for Wireless Ac-
cess in Vehicular Environments (WAVE) - Multi-channel
Operation. IEEE Std 1609.4-2006.
IEEE. DSRC Home. Retrieved June 8, 2008 from: http://
grouper.ieee.org/groups/scc32/dsrc/index.html
IEEE1609.2 (2006). IEEE trial-use standard for wireless
access in vehicular environments - security services for
applications and management messages.
IEEE802.11. (2008). IEEE 802.11 working group.
Retrieved October 13, 2008, from http://grouper.ieee.
org/groups/802/11/
International Electrotechnical Commission – IEC. (2002).
CISPR 25—Radio disturbance characteristics for the
protection of receivers used on board vehicles, boats,
and on devices—Limits and methods of measurement.
International Organization for Standardization—ISO
11452-2. (2005). Road vehicles — Vehicle test methods
for electrical disturbances from narrowband radiated
electromagnetic energy — Part 2: Off-vehicle radiation
sources.
Internet ITS Consortium. (2008). Retrieved September
2, 2008, from http://www.internetits.org/
Iperf, (2008). Iperf tool and documentation, Retrieved
from official web http://sourceforge.net/projects/iperf.
Compilation of References

Isermann, R. (2005). Fault-Diagnosis Systems: An In- Jardosh, A., Belding-Royer, Elizabeth M., Almeroth,
troduction-From Fault Detection To Fault Tolerance. K. C., & Suri, S. (2003). Towards realistic mobility
Springer Verlag, NY, USA. models for mobile ad hoc networks. In MobiCom ’03:
Proceedings of the 9th annual international conference
Isermann, R., Schwarz, R., & Stolzl, S. (2002, October).
on Mobile computing and networking, (pp. 217—229).
Fault tolerant drive-by-wire systems. IEEE Control
ACM Press, New York, USA.
Systems Magazine, (pp. 64-81).
JasPar. (2008). Retrieved September 2, 2008, from
ISO 10605. (2008). Road vehicles — Test methods for
https://www.jaspar.jp/english/index_e.php
electrical disturbances from electrostatic discharge.
Jekeli, C. (2001). Inertial Navigation Systems with
ISO 11452-3. (2001). Road vehicles — Component test
Geodetic Applications, De Gruyter, (pp. 308-318). New
methods for electrical disturbances from narrowband
York, USA
radiated electromagnetic energy — Part 3: Transverse
electromagnetic mode (TEM) cell. Jiang, D., & Delgrossi, L. (2008). IEEE 802.11p: To-
wards an International Standard for wireless Access
ISO 7637-2. (2004). Road vehicles — Electrical distur-
in Vehicular Environments. Retrieved September 2,
bances from conduction and coupling — Part 2: Electrical
2008, from http://ieeexplore.ieee.org/stamp/stamp.
transient conduction along supply lines only.
jsp?arnumber=04526014
ISO 7637-3. (2007). Road vehicles — Electrical distur-
Jin, W., & Recker, W. (2006). Instantaneous information
bance by conduction and coupling — Part 3: Vehicles
propagation in a traffic stream through inter-vehicle
with nominal 12 V or 24 V supply voltage — Electrical
communication. Transportation Research Part B, 40(3),
transient transmission by capacitive and inductive
230-250.
coupling via lines other than supply lines.
Jist/SWANS (2008). Java in Simulation Time / Scalable
ISO TC 204 Working Group 16. (2007). CALM Continu-
Wireless Ad hoc Network Simulator. Retrieved March
ous Communications for Vehicles. Retrieved June 8, 2008
19, 2008, from http://jist.ece.cornell.edu/index.html
from: http://www.calm.hu
Johansson, K. H., Törngren, M., & Nielsen, L. (2005).
J2056/2 Survey. (1994). J2056/2 survey of known proto-
Vehicle applications of controller area network. In D.
cols. In SAE Handbook. Warrendale, PA: Soc. Automotive
Hristu-Varsakelis & W. S. Levine (Ed.), Handbook of Net-
Eng. (SAE), vol. 2.
worked and Embedded Control Systems, Birkhäuser.
Jaillet, P., & Wagner, M. (2006). Online routing problems:
Johansson, K., Törngren, M., & Nielsen, L. (2005), Hand-
Value of advanced information as improved competitive
book of Networked and Embedded Control Systems, D.
ratios. Transportation Science, 40(2).
Hristu-Varsakelis and W. S. Levine, Eds. Boston, MA:
Jain, K., Padhye, J., Padmanabhan, V., & Qiu, L. (2003). Birkhäuser, 2005.
Impact of interference on multi-hop wireless network
Johnson, D. B., & Maltz, D. A. (1996). Dynamic source
performance. In MobiCom 2003, San Diego, California,
routing in ad hoc wireless networks. In Imielinski &
USA, Sep. 14-19.
Korth (Eds.), Mobile computing, 353. Kluwer Academic
Jakobsson, M., Wang, X., & Wetzel, S. (2004). Stealth Publishers.
attacks in vehicular technologies. In VTC-Fall 2004.
Johnson, D. B., Maltz, D. A. & Broch, J. (2001). DSR:
Jaman, G. (2007). Structural Monitoring Using Wireless The Dynamic Source Routing Protocol for Multi-Hop
Sensors and Controller Area Network. IEEE 5th Annual Wireless Ad Hoc Networks. In C. E. Perkins (Ed.), Ad
Conference on Communication Networks and Services Hoc Networking (pp. 139-172). Addison-Wesley, 2001
Research (CNSR’07).

315

Julier, S. (1997). Process Models for the Navigation of
High-Speed Land Vehicles. Wadham College, Oxford,
UK (p. 184).
Julier, S. J., & Uhlmann, J. K. (1997). A New Exten-
sion of the Kalman Filter to Nonlinear Systems. Paper
presented at the 11th International Symposium on
Aerospace/Defense Sensing, Simulation and Controls.
Orlando, FL, USA
Jurdak, R., Lopes, C., & Baldi, P. (2004). A survey,
classification and comparative analysis of medium
access control protocols for wireless networks. IEEE
Communication Surveys and Tutorials, 6.
Jurgen, R. K. (1998). Navigation and Intelligent Trans-
portation Systems. Progress In Technology, Automotive
Electronics Series. SAE International, Detroit, USA.
Kaiser, Kenneth L. (2005). Electromagnetic Compat-
ibility Handbook. CRC Press.
Kalman, R. E. (1960). A new approach to linear filtering
and prediction problems. Transactions of the American
Society of Mechanical Engineering (ASME). (82) (Series
D), (pp. 35-45).
Kanaris, A., Kosmatopoulos, E. B., & Loannou, P. A.
(2001). Strategies and spacing requirements for lane
changing and merging in automated highway systems.
IEEE Trans. Veh. Technol., 50(6), 1568–1581.
Karn, P. (1990, September). MACA – A new channel
access method for packet radio, ARRL/CRRL Amateur Ra-
dio Computer Networking Conference (pp. 134-140).
Karnadi, F. K., Zhi, H. M., & Kun-chan, L. (2007). Rapid
Generation of Realistic Mobility Models for VANET.
Wireless Communications and Networking Confer-
ence, WCNC 2007. IEEE (pp. 2506-2511). Kowloon,
Hong Kong.
Karp, B., & Kung, H. T. (2000). GPSR: Greedy perimeter
stateless routing for wireless networks. In Proceedings
of the 6th annual international conference on Mobile
computing and networking (pp. 243-254). Boston, Mas-
sachusetts, United States.
316
Compilation of References
Karp, B., & Kung, H.-T. (2000). Greedy Perimeter State-
less Routing for Wireless Networks. Proceedings of the
6th ACM International Conference on Mobile Computing
and Networking, (pp. 243-254).
Kato, S., & Tsugawa, S. (2001). Cooperative driving of
autonomous vehicles based on localization, inter-vehicle
communications and vision systems. JSAE Rev., 22(4),
503–509.
Katz, J. H. (1963). Simulation of a Traffic Network.
Communications of the ACM, 6(8), 480-486
Kelly, A., & Stentz, A. (1998). Rough Terrain Autonomous
Mobility - Part 1: A Theoretical Analysis of Require-
ments. Autonomous Robots, 5.
Kendall, J. (2005). Current Technology. Commercial
Motor, 201(5117), 62-64.
Kerner, B. S. (2004). Three-phase traffic theory and
highway capacity. Physica A, 333, 379–440.
Kesting, A., Treiber, M., & Helbing, D. (2007). General
lane-changing model MOBIL for car-following models.
Transp. Res. Rec., 1999, 86–94.
Kesting, A., Treiber, M., Schönhof, M., & Helbing, D.
(2008). Adaptive cruise control design for active con-
gestion avoidance. Transport. Res. Part C, doi:10.1016/
j.trc.2007.12.004.
Killat, M., Schmidt-Eisenlohr, F., Hartenstein, H., Rössel,
C., Vortisch, P., Assenmacher, S., & Busch, F. (2007).
Enabling efficient and accurate large-scale simulations
of VANETs for vehicular traffic management. In Pro-
ceedings of the Fourth ACM international Workshop
on Vehicular Ad Hoc Networks, VANET ‘07 (pp. 29-38).
Montreal, Canada.
Kim, S.-B., Lee, S.-Y., Choi J., & et al. (2003). Multi-Sen-
sor data fusion for future telematics applications. Journal
Astronomical Space Sciences, 20(4), 359–364.
Kleinrock, L. (1964). Communication nets: Stochastic
message flow and delay. New York, New York: Mc-
Graw-Hill.
Compilation of References
Kleinrock, L., & Tobagi, F. (1975, December). Packet
switching in radio channels: Part I – Carrier sense
multiple-access modes and their throughput-delay
characteristics. IEEE Transactions on Communications,
23, 1400-1416.
Knight, D. T. (1997). Rapid development of tightly-cou-
pled GPS/INS systems. IEEE Aerospace and Electronic
Systems (AES) Magazine, 12(2), 14-18.
Ko, Y., Shankarkumar, V., & Vaidya, N. (2000). Medium
access control protocols using directional antennas in ad
hoc networks. IEEE INFOCOM (pp. 13-21).
Kobayashi, F., & Arai, F. (1998). Sensor fusion system
using recurrent fuzzy inference. Journal of Intelligent
and Robotic Systems, (23), 201-216.
Koblitz, N., & Menezes, A. (2005). Pairing-based
cryptography at high security levels. Cryptology ePrint
Archive, Report 2005/076.
Kopetz, H., et al. (2002). Specification of the TTP/A Pro-
tocol. (Technical Report). Vienna, Austria, University.
Technology.
Kosch, T. (2004). Technical Concept and Prerequisites
of Car-to-Car Communication. Retrieved October 13,
2008, from http://www.car-to-car.org/fileadmin/doku-
mente/pdf/C2C-CC_presentation_3_BMW.pdf
Kosh, T., Schwingenschlögl, C., & Ai, L. (2002). Informa-
tion Dissemination in Multihop Inter-vehicle Networks
– Adapting the Ad-hoc On-demand Distance Vector
Routing Protocol (AODV). Proceedings in the 5th IEEE
International Conference on Intelligent Transportation
Systems, (pp. 685-690).
Krajzewicz, D., Bonert, M., & Wagner, P. (2006).
The Open Source Traffic Simulation Package SUMO.
Presented at RoboCup 2006 Infrastructure Simulation
Competition. Bremen, Germany.
Krishna, V., Chatterjee, & Pradhan (1997). A cluster-
based approach for routing in dynamic networks. ACM
SIGCOMM, Computer Communication Review, (pp.
49–65).
Kronander, J. (2004, December). Robust Automotive
Positioning: Integration of GPS and Relative Motion
Sensors. PhD Dissertation, Linköping, Sweden.
Kuban, P. A. (2007). A Controller Area Network Gateway
to ZigBee- A Proposition of an Architecture to Extend
CAN. VDM Verlag Dr. Mueller e.K.
Kukshya, V., Krishnan, H., & Kellum, C. (2005, June).
Design of a System Solution for Relative Positioning of
Vehicles Using Vehicle-to-Vehicle Radio Communica-
tions During GPS Outages. Paper presented at the IEEE
Vehicular technology Conference, 2, 1313-1317. Dallas,
Tx, USA.
Kvaser. (2008). CAN Kingdom. Retrieved October 13,
2008, from http://www.kvaser.com/index.htm
Kwak, J. S., & Jae, H. L. (2004). Infrared transmission
for intervehicle ranging and vehicle-to-roadside com-
munications systems using spread spectrum technique.
IEEE Transactions on Intelligent Transportation Systems,
5(1), 12-19.
Laberteaux, K., & Hu, Y.-C. (2006). Strong VANET se-
curity on a budget. In ESCAR 2006, Berlin, Germany.
Lammer, S., Kori, H., Peters, K., & Helbing, D. (2006).
Decentralised control of material or traffic flows in net-
works using phase-synchronisation. Physica A, 363(1),
39–47.
Laneurit, J., Blanc, C., Chapuis, R., & Trassoudaine
L. (2003, June). Multisensorial data fusion for global
vehicle and obstacles absolute positioning. Paper pre-
sented at the IEEE Intelligent Vehicle Conference, San
Francisco, USA.
Laugier, C., Vasquez, D., Yguel M., Fraichard Th. &
Aycard O., (2008). Geometric and Bayesian Models for
Safe Navigation in Dynamic Environments. Intelligent
Service Robotics 1(1).
LEA. (2008). LEA Organisation. Retrieved from official
web site http://www.leacom.fr.
LeCunn, Y., & Bottou, L. (2001). Gradient-Based Learn-
ing Applied to Document Recognition. IEEE Press, NY,
USA.
317

Lee, A. (2003). Driven by wire. The Engineer, 5(8),
28-29.
Lee, D., Attias, R., Puri, A., Sengupta, R., Tripakis, S.,
& Varaiya, P. (2001). A wireless token ring protocol
for intelligent transportation systems. Proceedings of
IEEE Intelligent Transportation Systems Conference
(pp. 1152-1157).
Lee, E. A. (2008). Cyber-Physical Systems: Design
Challenges. 11th IEEE International Symposium on
Object Oriented Real-Time Distributed Computing (pp.
363–369).
Lee, J., & Ha, I. (2001). Real-Time Motion Capture for a
Human Body using Accelerometers. Robotica, 19, 601–
610. Cambridge, UK: Cambridge University Press.
Lee, K. W. (2007). Autonomous vehicle navigation in
neighbourhood environments: A map assisted and a
feature-based approach. Unpublished doctoral disserta-
tion, Nanyang Technological University.
Lee, K.W., Wijesoma, S., & Ibanez-Guzman, J. (2007).
A constrained SLAM approach to robust and accurate
localisation of autonomous ground vehicles. Robotics
Autonomous Systems, 55.
Lee, M. K., Latchman, H. A., Newman, R. E., Katar,
S., & Yonge, L. (2002). Field performance comparison
of IEEE 802.11b and HomePlug 1.0 . IEEE Conference
Conference Local Computer Networks LCN 2002. 27th
Annual (pp. 598- 599).
Leen, G., & Hefferman, D. (2001). Vehicles without
wires. Computing & Control Engineering Journal,
12(5), 205-21.
Lettieri, P., & Srivastava, M. B. (1998, April). Adap-
tive frame length control for improving wireless link
throughput, range, and energy efficiency. IEEE INFO-
COM, 2, 564-571.
Lezniak, T. W., Lewis, R. W., & McMillen, R. A. (1977).
A Dead Reckoning/Map Correlation System for Auto-
matic Vehicle Tracking. IEEE Transactions on Vehicular
Technology, 1(VT-26), 47-60.
318
Compilation of References
Li, F., & Wang, Y. (2007). Routing in vehicular ad hoc
networks: A survey. IEEE Vehicular Technology Maga-
zine, 2(2), 12-22.
Li, J., Jannotti, J., De Couto, D., Karger, D., & Morris, R.
(2000). A Scalable Location Service for Geographic Ad
Hoc Routing. Proceedings of the 6th ACM International
Conference on Mobile Computing and Networking, (pp.
120-130).
Li, L., & Wang, F.-Y. (2006). Cooperative Driving at Blind
Crossings Using Intervehicle Communication. IEEE
Transactions on Vehicular Technology, 55, 1712-1724.
Li, L., Song, J., Wang, F. Y., Niehsen, W., & Zheng, N. N.
(2005). IVS 05: New developments and research trends
for intelligent vehicles. IEEE Intell. Syst., 20(4), 10–14.
Li, M., Poovendran, R., Sampigethaya, K., & Huang,
L. (2005). CARAVAN: Providing location privacy for
VANET. In ESCAR 2005, Cologne, Germany.
Li, N., Hou, C., & Sha, L. (2003). Design and analysis of
an mst-based topology control algorithm. In Proc. of the
22nd Annual Joint Conference of the IEEE Computer and
Communications Societies (INFOCOM), San Franciso,
CA, USA, Mar. 30-Apr. 3.
Li, W., & Leung, H. (2003). Constrained Unscented
Kalman Filter Based Fusion of GPS/INS/Digital Map
for Vehicle Localization. Paper presented at the IEEE
International Conference on Intelligent Transportation
Systems, Shangai, China.
Liang, C., & Peng, H. (2000). String stability analysis of
adaptive cruise controlled vehicles. JSME Intl. J., Series
C, 43(3), 671–677.
Lienard, M., Carrion, M., Degardin, V., & Degauque, P.
(2008). Modeling and Analysis on In-vehicle power line
communication channels. IEEE Transaction Vehicular
TEchnologie, 57(2), 670-679.
Limpibunterng, T., & Fujioka, T. (2002). A new design
approach for steer-by-wire system by dual port system.
Vehicle System Dynamics, Suppl. 37, 197-208.
Compilation of References
Limpibunterng, T., & Fujioka, T. (2004). Bilateral driver
model for steer-by-wire controller design. Vehicle System
Dynamics, Suppl. 41, 381-390.
LIN Consortium. (2003). LIN Specification Package. Re-
vision 2.0.Retrieved from http://www.lin~subbus.org/.
Lin, X., Sun, X., Ho, P., & Shen, X. (2007). GSIS: A
Secure and Privacy-Preserving Protocol for Vehicular
Communications. In IEEE Transactions on Vehicular
Technology 56, 3442-3456.
Lin, Y. J., Haniph, A., Latchman, H., Neuman, R. (2003,
April). A comparative Performance study of Wireless and
Power Line Network. IEEE Communications Magazine,
(pp. 54-63).
Lin, Y., Latchman, H., Katar, S., & Lee, M. K. (2003, May).
Theoretical and Field Performance Comparison between
802.11 Wireless and Powerline HomePlug 1.0 Protocols.
IEEE Communications Magazine with Focus Theme on
Powerline Local Area Networks, (pp. 54-63).
LIN. (2008). Local Interconnection Network (LIN). Re-
trieved October 13, 2008, from http://www.interfacebus.
com/Design_Connector_LIN_Bus.html
Liu, Y., & Ozguner, U. (2003). Effect of inter-vehicle
communication on rear-end collision avoidance. Intel-
ligent Vehicles Symposium (pp. 168-173).
LiveMesh. (2008). Microsoft Live Mesh. Retrieved Octo-
ber 13, 2008, from http://news.cnet.com/8301-10784_3-
9926578-7.html
Lo, H., & Szeto, W. (2004). Modeling advanced traveler
information services: static versus dynamic paradigms.
Transportation Research Part B: Methodological, 38(6),
495-515.
Local Interconnect Network. (2008). Retrieved Septem-
ber 2, 2008, from http://www.lin-subbus.org/
Lochert, C., Füßler, H., Hartenstein, H., Hermann, D., Tian,
J., & Mauve, M. (2003). A Routing Strategy for Vehicular
Ad Hoc Networks in City Environments. Proceedings of
the IEEE Intelligent Vehicles Symposium, (pp. 156-161).
Lu, R., Lin, X., Zhu, H., Ho, P., & Shen, X. (2008). ECPP:
Efficient Conditional Privacy Preservation Protocol for
Secure Vehicular Communications. In The 27th IEEE
International Conference on Computer Communications
(INFOCOM 2008), Phoenix, Arizona.
Lu, X. Y., Tan, H. S., Shladover, S. E., & Hedrick, J. K.
(2004). Automated vehicle merging maneuver implemen-
tation for AHS. Vehicle Syst. Dyn., 41(2), 85–107.
Lu, X-Y., & Hedrick, J. K. (2005). Heavy-duty vehicle
modelling and longitudinal control. Vehicle System
Dynamics, 43(9), 653–669.
Lupini, C. A. (2004). Vehicle Multiplex Communica-
tion: Serial Data Networking Applied to Vehicular
Engineering. Warrendale, Pensylvania: SAE. ISBN
0-7680-1218-X.
Lygeros, J., Godbole, D. N., & Sastry, S. (1996). Multia-
gent hybrid system design using game theory and optimal
control. In Proc. IEEE CDC ’96 (pp. 1190–1195).
Lygeros, J., Godbole, D. N., & Sastry, S. (1998). Verified
hybrid controller for automated vehicles. IEEE Trans.
Autom. Control, 43(4), 522–539.
Madhavan, R., Messina, E. R., & Albus, J. S. (2007).
Intelligent Vehicle Systems: A 4D/RCS Approach. New
York: Nova Science Publishers.
Magrabi, S. M. & Gibbens, P. W. (2000, June). Decen-
tralized Fault Detection and Diagnosis in Navigation
Systems for Unmanned Aerial Vehicles. Paper presented at
the IEEE Position Location and Navigation Symposium,
San Diego,USA.
Mahajan, A., Potnis, N., Gopalan, K., & Wang, A. (2007).
Modeling VANET Deployment in Urban Settings. In
Proceedings of the 10th ACM Symposium on Modeling,
analysis, and simulation of wireless and mobile systems
(pp 151-158). Chania, Crete Island, Greece.
Mahmud, S. M., & Alles, S. (2005). In-Vehicle Network
Architecture for the Next-Generation Vehicles. Proceed-
ings of the 2005 SAE World Congress, Detroit, Michigan,
Paper Number 2005-01-1531.
319

Mahmud, S. M., & Shanker, S. (2006, May). An In-Ve-
hicle Secure Wireless Personal Area Network (SWPAN).
IEEE Transactions on Vehicular Technology, 55(3),
1051-1061.
Mahmud. S. M. (1994, July). Performance Analysis
of Multilevel Bus Networks for Hierarchical Multi-
processors. IEEE Transactions on Computers, 43(7),
789-805.
Mangharam, R., Weller, D. S., Stancil, D. D., Rajkumar
& R., Parikh, J. S. (2005). GrooveSim: A Topography-Ac-
curate Simulator for Geographical Routing in Vehicular
Networks. In Proceedings of the 2nd ACM International
Workshop on Vehicular Ad Hoc Networks (pp. 59-68).
Cologne, Germany
Mao, X., Wada, M., Hashimoto, H., Saito, M. & Suzuki,
I. (2003, November). EM/Unscented Smoothing Based
Parameter Learning for Nonlinear Models for GPS Po-
sitioning. Paper presented at the 10th World Congress on
Intelligent Transport Systems, Madrid, Spain.
Margolin, B. (1997). Analog, mixed-signal ICs steering
drive-by-wire toward reality. Computer Design, 36(12),
31-34.
Matheus, K. (2002). Car-to-Car Communication – Mar-
ket Introduction and Success factors. Retrieved October
13, 2008, from http://www.car-to-car.org/fileadmin/do-
kumente/pdf/C2C-CC_presentation_4_VW_Audi_Car-
meq.pdf
Matric. (2008). The Controller Area Network (CAN)
has gone wireless! Retrieved September 2, 2008, from
http://www.matric.com/resources/Canbridge_bro-
chureV2.pdf
Matthaiou, M., Laurenson, D. I., Cheng-Xiang, W. (2008).
PHY 30-3 – Capacity Study of vehicle-to-roadside
MIMO channels with a line-of-sight component. IEEE
Wireless Communications and Networking Conference,
(pp. 775-779).
Maxemchuk, N. F., & Shur, D. (2001). An Internet mul-
ticast system for the stock market. ACM Transactions
on Computer Systems, 19, 384-412.
320
Compilation of References
Maxemchuk, N. F., Tientrakool, P., & Willke, T. L. (2007).
Reliable Neighborcast. IEEE Transactions on Vehicular
Technology, 56, 3278-3288.
May, A. D. (1990). Traffic flow fundamentals. Englewood
Cliffs, NJ: Prentice-Hall, Inc., A Division of Simon and
Schuster.
Merwe, A. D. R., Freitas, N., & Wan, E. (2000, August).
The Unscented Particle Filter, Tech report CUED/F-
INFENG/TR-380. Cambridge University Engineering
Department, Cambridge, England.
Metcalfe, R., & Boggs, R. (1976, July). Ethernet: Dis-
tributed packet switching for local computer networks,
Communication of the ACM, 19, 395-404.
Michaelides, R. (2008). Wireless CAN Interface. Re-
trieved September 2, 2008, from http://www.rmcan.
com/index.php?id=17&L=1
Minderhoud, M. M. (2002). Simulation model of motor-
ways with next generation vehicles, technical specifica-
tion.
Molloy, M. K. (1989). Fundamentals of Performance
Modeling. Macmillan Publishing Company.
Morris, R., Jannotti, J., Kaashock, F., Li, J., & Decouto,
D. (2000). A Scalable Ad Hoc Wireless Network System.
Proceedings of the 9th workshop on ACM SIGOPS Euro-
pean Workshop: beyond the PC: New challenges for the
operating system, (pp. 61-65).
Moscibroda, T., & Wattenhofer, R. (2005). Minimizing
interference in ad hoc and sensor networks. In DIALM-
POMC’05, Cologne, Germany, Sep., 2.
MOST (2006). MOST Core Compliance Test Specification
V2.00. Technical Report. Retrieved, from http://www.
mostcooperation.com/publications.
MOST Cooperation (2008). MOST Specification Revi-
sion. Retrieved from http://www.mostnet.de.
MOST. (2008). Media Oriented System Transport
(MOST). Retrieved October 13, 2008, from http://www.
interfacebus.com/MOST_Bus_Interface.html
Compilation of References
Motorola (1998). Bosch Controller Area Network Version
2.0 Protocol Standard.
Mouskos, K. C., Greenfeld, J., & Pignataro, L. J. (1996).
Toward a multi-modal advanced traveler information
system. NJIT Research, 4.
MOVE (2008). MObility model generator for VEhicular
networks. Retrieved March 19, 2008, from http://www.
csie.ncku.edu.tw/~klan/move/index.htm
Mussa, R., Upchurch, J. (2000). Simulator evaluation of
incident detection using VRC system. Proceeding of the
IEEE Intelligent Vehicles Symposium, (pp. 442-450).
Nagel, K., & Schreckenberg, M. (1992). A cellular
automaton model for freeway traffic. J. Phys. I France,
2(12), 2221–2229.
Nakagami, M. (1960). The m-distribution, a general
formula of intensity distribution of the rapid fading. In
W. G. Hoffman (Ed.), Statistical methods in radio wave
propagation, (pp. 3-36), Oxford, UK. Pergamon.
Nakamura, E. F., Loureiro, A. A. F., & Frery, A. C.
(2007). Information fusion for wireless sensor networks:
Methods, models, and classifications. ACM Comput.
Surv., 39(3), A9/1–A9/55.
National Science Foundation (2006). NSF Workshop on
Cyber-Physical Systems. Retrieved September 12, 2008,
from http://varma.ece.cmu.edu/cps/
Navet (2008). The Automotive Embedded Systems Hand-
book. Industrial Information Technology series, CRC
Press / Taylor and Francis, ISBN 978-0849380266, to
appear November 2008.
Navet, N., Song, Y-Q., Simonot-Lion, F., & Wilvert, C.
(2005, June). Trends in Automotive Communication Sys-
tems. Proceedings of the IEEE, special issue on Industrial
Communications Systems., 93(6), 1204-1223.
Nejad, K., & Li, X. (2005). Low- interference topology
control for wireless ad-hoc networks. Ad Hoc Sensor
Wireless Networks, 1, 41-64.
Network on Wheels (NoW). (2008). Retrieved September
2, 2008, from http://www.network-on-wheels.de/
Ni, J., & Chandler, S. (1994). Connectivity properties
of a random radio network. IEE Proc. Commun., 141(4),
289-296.
Nirupama, B. (2005). Wireless Sensor Networks, A Sys-
tems Perspective, 2005 Artech House, Boston, USA
NITRD. National Workshop on High-Confidence Au-
tomotive Cyber-Physical Systems (2008). Retrieved
September 12, 2008, from http://varma.ece.cmu.edu/
Auto-CPS/
Nolte, T. (2003). Reducing Pessimism and Increasing
Flexibility in the Controller Area Network. Mälardalen
University Licentiate Thesis, May 2003 Department of
Computer Science and Engineering Mälardalen Univer-
sity Västeras, Sweden, Copyright Thomas Nolte, Printed
by Arkitektkopia, Västeras, Sweden Distribution .
Nolte, T. (2006). Share-Driven Scheduling of Embedded
Networks. Mälardalen University, Dissertation, May
2006. Department of Computer Science and Engineer-
ing Mälardalen University Västeras, Sweden, Printed by
Arkitektkopia, Västeras, Sweden Distribution.
Nouvel, F., El Zein, G., & Citerne, J. (1994, June). Code
division multiple access for an automotive area network
over power-lines. Proceedings of IEEE Vehicular Tech-
nology Conf., (pp. 525–529).
ns-2 (2008). The ns-2 wiki. Retrieved March 18, 2008 at
http://www.isi.edu/nsnam/ns/doc/
O’Gorman, E. (2002). Using group communication to
support inter-vehicle coordination. Unpublished master’s
thesis, University of Dublin, Ireland.
Oh, S-W., Chae, H-C., Yun, S-C., & Han, C-S. (2004).
The design of a controller for the steer-by-wire system.
JSME International Journal, Series C, 47(3), 896-907.
OpenSSL. Retrieved September, 2008, from http://www.
openssl.org.
Papadimitratos, P., Buttyan, L., Hubaux, J.-P., Kargl, F.,
Kung, A., & Raya, M. (2007). Architecture for secure and
private vehicular communications. In ITST’07, Sophia
Antipolis, France.
321

Papadimitratos, P., Calandriello, G., Hubaux, J.-P., &
Lioy, A. (2008). Impact of Vehicular Communication
Security on Transportation Safety. In IEEE MOVE 2008,
Phoenix, (AZ), US, April 17, 2008.
Papadimitratos, P., Gligor, V., & Hubaux, J.-P. (2006).
Securing Vehicular Communications - Assumptions,
Requirements, and Principles. In ESCAR 2006, Berlin,
Germany.
Papadimitratos, P., Kung, A., Hubaux, J.-P., & Kargl, F.
(2006). Privacy and Identity Management for Vehicular
Communication Systems: A Position Paper. In Workshop
on Standards for Privacy in User-Centric Identity Man-
agement, Zurich, Switzerland, July 2006.
Papageorgiou, M., & Kotsialos, A. (2002). Freeway ramp
metering: an overview. IEEE Trans. Intell. Transport.
Syst., 3(4), 271–281.
Papageorgiou, M., Diakaki, C., Dinopoulou, V., Kotsialos,
A., & Wang, Y. (2003). Review of road traffic control
strategies. Proc. of the IEEE, 91(12), 2043–2067.
Papp, Z., Brown, C. L., & Bartels, C. (2008). World
Modeling for Cooperative Intelligent Vehicles. Pro-
ceedings 2008 IEEE Intelligent Vehicles Symposium,
IV2008, Eindhoven.
Parker R., & Shahrokh V. (2006, September). Vehicle
Localization in Vehicular Networks. Paper presented at
the IEEE 64th Vehicular Technology Conference VTC-
2006, Melbourne, Australia.
Parker, D. S., & Stedman, T. C. (1994). Energy efficient
transportation for Florida. (http://www.p2pays.org/
ref/12/11355.pdf, last accessed: Nov. 2007)
Parno, B., & Perrig, A. (2005). Challenges in securing
vehicular networks. In HotNets-IV, College Park, MA,
USA.
Patton, R., Frank, P., & Clark, R. (1989). Fault Diagnosis
in Dynamic System: Theory and Applications. Prentice
Hall, USA.
Paul, C. R. (2006). Introduction to Electromagnetic
Compatibility, 2nd. ed. Wiley-Interscience.
322
Compilation of References
Pedersen, M. M., & Ruhoff, P. T. (2002). Entry ramps
in the Nagel-Schreckenberg model. Phys. Rev. E, 65(5),
056705.
Perkins, C., & Bhagwat, P. (1994). Highly Dynamic
Destination-Sequence DistanceVector Routing (DSDV)
for Mobile Computers. ACM SIGCOMM Computer Com-
munication Review, 24(4), 234-244.
Perkins, C., Belding-Royer, E., & Das, S. (2003). Ad hoc
On-Demand Distance Vector (AODV) Routing. http://
www.ietf.org/rfc/rfc3561.txt. Request for Comments
(Work in Progress).
Petigara, N. D., & Balakrishnan, H. (2004). Integrated
routing and storage for messaging applications in mobile
ad hoc networks. Mobile Networks and Applications,
9(6), 595-604.
Pfeiffer, G., & Schrei, A. (1994, June). Investigation on
propagation characteristics of a short range radio link
with statistical distributed reflectors. IEEE Vehicular
Technology Conference, 3, 1728-1732.
Piorkowski M., Raya M., Lugo A., Papadimitratos P.,
Grossglauser M., & Hubaux J. P. (2007, September).
TraNS: Realistic Joint Traffic and Network Simulator
for VANETs. Short paper presented at the Fourth ACM
international Workshop on Vehicular Ad Hoc Networks,
VANET ‘07. Montreal, Canada.
Potnis, N., & Mahajan, A. (2006). Mobility Models for
Vehicular Ad Hoc Network Simulations. In Proceedings
of the 44th ACM Southeast regional conference (pp 746-
747). Melbourne, Florida.
Pouliezos, A. D., & Stavrakakis, G. S. (1994). Real Time
Fault Monitoring of Industrial Processes. NY, USA:
Kluwer Academic Publishers.
PReVENT. (2008). Retrieved September 2, 2008, from
http://www.prevent-ip.org/
PwC Automotive Institute. (2008). Analyst Note, August
1, 2008.
Qi, Y., Kobayashi, H., & Suda, H. (2006). On Time-
of-arrival Positioning in a Multipath Environment.
IEEE Transactions on Vehicular Technology, 5(55),
1516-1526.
Compilation of References
Qualnet (2001). Version 3.1 User’s Manual. Retrieved
March 19, 2008 from http://linux.ucla.edu/~gavin/pa-
pers/SimulatorManual-3.1.pdf
Rabadi, N. M., & Mahmud, S. M. (2007). Privacy
Protection Among Drivers in Vehicle-to-Vehicle Com-
munication Networks. Proceedings of the 4th IEEE
Consumer Communications and Networking Conference,
Las Vegas, NV, January 11-13, 2007.
Rabadi, N. M., & Mahmud, S. M. (2008). Drivers’
Anonymity with a Short Message Length for Vehicle-to-
Vehicle Communications Network. Proceedings of the
5th IEEE Consumer Communications and Networking
Conference, Las Vegas, NV, January 10-12, 2008.
Raghupathy, Bevas, & Vaduvur (1998). Spine routing
in ad hoc networks. ACM/Baltzer Cluster Computing
Journal, 1(2), 237–248.
Rajaraman, R. (2002). Topology control and routing in
ad hoc network: A survey. ACM SIGACT News, 33(2),
60-73.
Ramanathan, R., & Rosales-Hain, R. (2000). Topology
control of multihop wireless networks using transmit
power adjustment. In Proc. of the 19th Annual Joint
Conference of the IEEE Computer and Communications
Societies (INFOCOM), Tel-Aviv, Israel, Mar. 26-30.
Ravi, N., Smaldone, S., Iftode, L., & Gerla, M. (2007).
Lane Reservation for Highways (Position Paper). Re-
trieved March 19, 2008 from www.cs.rutgers.edu/~iftode/
ILR-ITSC07.pdf
Raya, M., & Hubaux, J.-P. (2005). The security of ve-
hicular ad hoc networks. In SASN ‘05, (pp. 11-21), New
York, NY, USA: ACM Press.
Raya, M., Hubaux, J.P. (2005). The Security of Vehicle
Ad Hoc Networks. Proceedings of the ACM Workshop on
Security in Ad hoc and Sensor Networks (SASN’05), pp.
11-21, Alexandria, Virginia, USA, November 2005.
Raya, M., Papadimitratos, P., & Hubaux, J.-P. (2006).
Securing vehicular communications. IEEE Wireless
Communications magazine, 13(5).
Redmill, K. A., Kitajima, T., & Özgüner, U. (2001, June).
DGPS/INS Integrated Positioning for Control of Auto-
mated Vehicle. Paper presented at the IEEE Intelligent
Transportation Systems, Oakland, USA.
Redmill, K. A., Martin, J. I., & Özgüner, U. (2000,
October). Sensor and data fusion design and evaluation
with a virtual environment simulator. Paper presented
at the IEEE Intelligent Vehicles Symposium. Dearborn,
Mi, USA, Proceedings (pp. 668-674).
Reed, P. (2005). We test the tips: What really saves
gas? and how much? (http://www.edmunds.com/owner-
ship/driving/articles/106842/article.html, last accessed:
Nov. 2007).
Regt, d. H. (2007). Back To Basics & Future Trends:
Automotive Networking. December 7. 2007. Retrieved
September 8, 2008, from http://www.automotivedesign-
line.com/howto/204702733
Rehunathan, D. B., Seet, B. C., & Luong, T. T. (2007).
Federating of MITSIMLab and ns-2 for realistic vehicular
network simulation. In Proceedings of the 4th interna-
tional conference on mobile technology, applications,
and systems and the 1st international symposium on
Computer human interaction in mobile technology (pp
62-67). Singapore
Reichardt, D., Miglietta, M., Moretti, L., Morsink, P., &
Schulz, W. (2002). CarTALK 2000: Safe and comfort-
able driving based upon inter-vehicle-communication.
Proceedings of the IEEE Intelligent Vehicles Symposium
(pp. 545-550).
rfc3561 (2008). Ad hoc On-Demand Distance Vector
(AODV) Routing. Retrieved March 19, 2008 from http://
www.ietf.org/rfc/rfc3561.txt
Rickenbach, P., Schmid, S., Wattenhofer, R., & Zollinger,
A. (2005). A robust interference model for wireless ad
hoc networks. In Proc. 5th IEEE International Work-
shop on Algorithms for Wireless, Mobile, Ad-Hoc and
Sensor Networks (WMAN), Denver, Colorado, USA,
April 4-8.
Road Vehicles. (2004). Controller Area Network
(CAN)—Part 4: Time-Triggered Communication, ISO
11 898-4.
323
 Compilation of References

Roberts, L. (1975, April). ALOHA packet systems with SAFESPOT. (2008). Retrieved September 2, 2008, from
and without slots and capture, Computer Communica- http://www.safespot-eu.org/pages/page.php
tions Review, 5.
Saha, A. K., & Johnson, D. B. (2004). Modeling mobility
Robinson, R. (2006). The VII Consortium and the for vehicular ad-hoc networks. In Proc. VANET ’04 (pp.
Cooperative Agreement with USDOT. Retrieved Sep- 91–92). New York, NY, USA: ACM.
tember 2, 2008, from http://www.itsmichigan.org/An-
Saha, A., & Johnson, D. (2004). Modeling Mobility for
nualConf2006/presentations/Ralph-ITS-M%20Meeting-
Vehicular Ad-hoc Networks. In Proceedings of the 1st
%20VIIC%20Overview_a.ppt
ACM International Workshop on Vehicular Ad Hoc Net-
Rokitansky, C.-H., & Wietfeld, C. (1993). Comparison of works (VANET) (pp. 91-92). Philadelphia PA, USA.
Adaptive Medium Access Control Schemes for Beacon-
Saito, H. (2006). Performance analysis of combined
Vehicle Communications. Proceedings of the IEEE-IEE
vehicular communication. IEICE Transactions on Com-
Vehicle Navigation and Information Systems Conference,
munications, E89B(5),. 1486-1494.
(pp. 295-299).
Sameshima, H., Ogasa, M., & Yamamoto, T. (2004).
Rokitansky, C.-H., & Wietfeld, C. (1995). Methods and
On-Board Characteristics of Lithium Ion Batteries for
Tools for Performance Evaluation and Validation of Ve-
improving energy regenerative efficiency. Quarterly
hicle-Roadside Communication Proposed for Standardiza-
report of RTRI, 45(2), 45-52.
tion. Proceedings in Vehicular Technology Conference,
2, 964-970. Santos, R. A., Edwards, A., Edwards, R. M., & Seed, N.
L. (2005). Performance evaluation of routing protocols in
Ross, P. (2003, February). Top Ten Techno-Cool Cars.
vehicular ad-hoc networks. International Journal Ad Hoc
IEEE Spectrum, (pp. 30-35).
and Ubiquitous Computing, 1(1/2), 80-91.
Rossi, C., Tilli, A., & Tonielli, A. (2000). Robust Control
Sarvi, M., & Kuwahara, M. (2007). Microsimulation
of a Throttle Body for Drive by Wire Operation of Auto-
of freeway ramp merging processes under congested
motive Engines. IEEE Transactions on Control System
traffic conditions. IEEE Trans. Intell. Transport. Syst.,
Technology, 8(6), 993-1002.
8(3), 470–479.
Roumeliotis, S., & Bekey, G. (2000, April). Bayesian
Savarese, C., Rabaey, J. M., & Beutel, J. (2001, May).
estimation and Kalman filtering: A unified framework for
Location in Distributed Ad-Hoc Wireless Networks.
mobile robot localization. Paper presented at the IEEE
Paper presented at the IEEE Conference on Acoustics,
International Conference on Robotics & Automation,
Speech, and Signal Processing, Salt Lake City, Utah,
San Francisco, CA, USA.
USA, Proceedings (pp. 2037-2040).
Royer, E., & Toh, C. (1999). A Review of current routing
Savvides, A., Park, H., & Srivastava, M. B. (2002, Sep-
protocols for ad hoc wireless networks. IEEE Personal
tember). The Bits and Flops of the N-hop Multilatera-
Communications, (pp. 46-55).
tion Primitive For Node Localization Problems. Paper
Rubin, A. (2002). Implementing automotive protocols presented at the ACM conference on Wireless Sensor
for communications over noisy battery power lines. Networks and Applications. WSNA’02, Atlanta, Georgia,
Proceedings of the IEEE Conference on Conv. Elect USA Proceedings (pp. 112-121).
Electron. Eng. Israel, December, (p. 306).
Schäonhof, M., Kesting, A., Treiber, M., & Helbing, D.
SAFESPOT Project Partners. (2008). Retrieved June (2006). Coupled vehicle and information flows: Mes-
2008, from http://www.safespot-eu.org sage transport on a dynamic vehicle network. Physica
A-Statistical Mechanics and ITS Applications, 363(1),
73-81.

324
Compilation of References
Schiffer, A. (2000). Statistical channel and noise model-
ling of vehicular dc-lines for data communication. Pro-
ceedings of IEEE Vehicular Technology Conf.,Tokyo.
Schlingelhof, M., Betaille, D., Bonnifait, P., & Demas-
eure, K. (2008). Advanced positioning technologies for
co-operative systems. IET Intelligent Transportation
Systems, 2(2).
Segawa, M., Kimura, S., Kada, T., & Nakano, S. (2004).
A study of the relationship between vehicle behavior
and the steering wheel torque on steer by wire vehicles.
Vehicle System Dynamics, Suppl. 41, 202-211.
Seiler, P., Pant, A., & Hedrick, K. (2004). Disturbance
propagation in vehicle strings. IEEE Transactions on
Automatic Control, 49, 1835–1842.
Setlur, P. J., Wagner, R., Dawson, D. M., & Braganza,
D. (2006). A Trajectory Tracking Steer-by-Wire Control
System for Ground Vehicles. IEEE Transactions on
Vehicular Technology, 55(1), 76-85.
SEVECOM. (2008). Retrieved September 2, 2008, from
http://www.sevecom.org/
Shaheen, S., Heffernan, D., & Leen, G. (2003). A compari-
son of emerging time-triggered protocols for automotive
X-by-wire control networks. Proc. Instn Mech. Engrs
Part D: J. Automobile Engineering, 217, 13-22.
Shang, Y., Rumi, W., Zhang, Y., & Fromherz, M. P.
J. (2003, June). Localization from Mere Connectivity.
Paper presented at the ACM conference on Mobile Ad
Hoc Networking and Computing. Annapolis, Maryland,
USA, Proceedings (pp. 201-212).
Shavitt, Y., & Sha, A. (2005). Optimal routing in gossip
networks. IEEE Transactions on Vehicular Technology,
54(4).
Sichitiu, M. L., & Kihl, M. (2008). Inter-Vehicle Com-
munication Systems: A Survey. IEEE Communications
Surveys & Tutorials, 10(2), 88-105.
Singh, J. P., Bambos, N., Srinivasan, B., & Clawin, D.
(2002, December). Wireless LAN performance under
varied stress conditions in vehicular traffic scenarios.
IEEE Vehicular Technology Conference, 2, 743-747.
Sklar, B. (1997). Rayleigh Fading Channels in Mobile
Digital Communication Systems Part I: Characterization.
IEEE Communications Magazine, 35(7), 90-100
Smets, P., & Ristic, B. (2004, April). Kalman filters and
joint tracking and classification in the transferable belief
model framework. Paper presented at the 7th International
Conference on Information Fusion, (pp. IF04–0046),
Stockholm, Sweden.
Sommer, C., Yao, Z., German R., & Dressler, F. (2008)
On the Need for Bidirectional Coupling of Road Traffic
Microsimulation and Network Simulation. In Proceed-
ings of the Computer Communications Workshops.
INFOCOM (pp 1-6).Phoenix, AZ.
Song, B., & and Hedrick, J. K. (2004). Design and ex-
perimental implementation of longitudinal control for
automated transit buses. Proceedings of the American
Control Conference (pp. 2751-2756).
Spdicom (2008). Spdicom Inc., Retrieved from official
web Official web site http://www.spidcom.com.
Sreenivasiah, I., Chang, D. C., & Ma, M. T. (1981).
Emission characteristics of electrically small radiating
sources from tests inside a TEM cell. IEEE Trans. on
Electromagn. Compat., (23), 113-121.
Srini, V. P. (2006, December). A vision for supporting
Autonomous Navigation in Urban Environments, IEEE
Computer.
Srinivasan, K. K., & Mahmassani, H. S. (2003). Ana-
lyzing heterogeneity and unobserved structural effects
in route-switching behavior under ATIS: a dynamic
kernel logit formulation. Transportation Research Part
B: Methodological, 37(9), 793-814.
Stanton, N., & Marsden, P. (1997). Drive-by-wire systems:
some reflections on the trend to automate the driver role.
Proceeding of Inst. Mech. Engrs. Part D, 211, 267-276.
St-Pierre, M., & Gingras, D. (2004, March). Neural
network based data fusion for vehicle positioning in
land navigation system. Paper presented at the Society
of Automotive Engineering (SAE) conference. Paper no.
345, Detroit, USA.
325

St-Pierre, M., & Gingras, D. (2005, June). Comparison
between the unscented Kalman filter and the extended
Kalman filter for the position estimation module of
an integrated navigation information system. Paper
presented at the Institute of Electrical and Electronics
Engineering (IEEE) Intelligent Vehicles Conference,
Parma, Italy.
SUMO (2008). Simulation of Urban MObility. Retrieved
March 19, 2008 from http://sumo.sourceforge.net/docs/
bibliography.shtml
Suresh, B., & Sivan, K. (2004, April). Performance
evaluation of multi-sensor data-fusion systems in launch
vehicles. Sadhana , 29(2), 175–188. India
Swaroop, D. (1995). String stability of interconnected
systems: An application to platooning in automated
highway systems. Unpublished doctoral dissertation,
University of California, Berkeley, CA.
Syverson, P. F., & Stubblebine, S. G. (1999). Group
principals and the formalization of anonymity. In World
Congress on Formal Methods, (pp. 814-833)
Taliwal, V., Jiang, D., Mangold, H., Chen, C., & Sengupta,
R. (2004). Empirical determination of channel charac-
teristics for DSRC vehicle-to-vehicle communication.
In Proc. 1st ACM Int. Workshop on Vehicular Ad Hoc
Networks (p. 88). Philadelphia PA, USA.
Tian, J., Han, L., & Rothennel, K, (2003, October).
Spatially aware packet routing for mobile Ad Hoc inter-
vehicle radio networks. IEEE Intelligent Transportation
Systems Conference, 2, 1546- 1551.
TIGER (2008). Topologically Integrated Geographic
Encoding and Referencing system. Retrieved March 19,
2008 from http://www.census.gov/geo/www/tiger/
Tippet, J. C. & Chang, D. C. (1976). Radiation charac-
teristics of electrically small devices in a TEM trans-
mission cell, IEEE Trans. on Electromagn. Compat.,
18(4), 134-140.
Toh, C.-K. (2002). Ad Hoc Mobile Wireless Networks:
Protocols and Systems. Prentice-Hall International.
326
Compilation of References
Tokuda, K., Akiyama, M., & Fujii, H. (2000). DOLPHIN
for inter-vehicle communications system. Proceedings
of the IEEE Intelligent Vehicles Symposium (pp. 504-
509).
Toledo-Moreo, R. et al. (2007). High-Integrity IMM-
EKF-Based Road Vehicle Navigation With Low-Cost
GPS/SBAS/INS. IEEE Transactions on Intelligent
Transportation Systems, 8(3).
Torrent-Moreno, M., Corroy, S., Schmidt-Eisenlohr, F.,
& Hartenstein, H. (2006). IEEE 802.11-based one-hop
broadcast communications: understanding transmission
success and failure under different radio propagation
environments. In MSWiM ‘06, (pp. 68-77), New York,
NY, USA, ACM Press.
Torrent-Moreno, M., Jiang, D., & Hartenstein, H. (2004).
Broadcast reception rates and effects of priority access in
802.11-based vehicular ad-hoc networks. In VANET’04,
(pp. 10-18), New York, NY, USA. ACM Press.
Torrent-Moreno, M., Santi, P., & Hartenstein, H. (2006).
Distributed fair transmit power adjustment for vehicular
ad hoc networks. In SECON ‘06, (pp. 479-488).
Traffic Safety Facts. (2004). NHTSA´s National Center
for Statistics Analysis.
Transportation Statistics Annual Report. (2006). U.S.
Department of Transportation Bureau of Transportation
Statistics.
Treiber, M., & Helbing, D. (2002). Microsimulations of
freeway traffic including control measures. (http://arxiv.
org/abs/cond-mat/0210096)
Treiber, M., Hennecke, A., & Helbing, D. (2000). Con-
gested traffic states in empirical observations and micro-
scopic simulations. Phys. Rev. E, 62(2), 1805–1824.
Treiber, M., Kesting, A., & Helbing, D. (2006). Delays,
inaccuracies and anticipation in microscopic traffic
models. Physica A, 360(1), 71–88.
Tsai, H., Saraydar, C., Talty, T., Ames, M., Macdonald,
A., & Tonguz, O. (2007). ZigBee-based Intra-car Wireless
Sensor Network, Communications. IEEE International
Conference ICC’07, 24-28 June, (pp. 3965-3971).
Compilation of References
Tsugawa, S., & Kato, S. (2003). Evaluation of incident
information transmission on highways over inter-vehicle
communications. In Intelligent Vehicles Symposium,
2003. Proceedings. IEEE, Columbus, Ohio, USA, Jun.
9-11.
Tsukamoto, K., Fujii, M., Itami, M., & Itoh K. (2001,
May). Study on acquisition and tracking of vehicle
location using IVC and RVC dynamic networks. Paper
presented at the IEEE Intelligent Vehicle Symposium,
(pp. 467-470), Tokyo, Japan.
TTCAN. (2008). Time Triggered Controller Area Network
(TTCAN). Retrieved October 13, 2008, from http://www.
ttcan.com/
TTTech (2002). Time-Triggered Protocol TTP/C High-
Level Specification Document Specification, Document
number D-032-S-10-028.
U.S. DOT. Intelligent Vehicle Initiative. Retrieved July
4, 2008, from http://www.its.dot.gov/ivi/ivi.htm
Ukkusuri, S., & Du, L. (2008). Geometric connectivity of
vehicular ad hoc networks: Analytical characterization.
Transportation Research Part C, 16(5), 615-634.
Uldrich, J. (2008). The Future of the Automobile. March
10, 2008. Retrieved September 8, 2008, from http://www.
jumpthecurve.net/index.php/recent_posts/the_future_
of_the_automobile/
Uplap, R., Okhade, M., Ghanekar, P., Panse, O., Narayan,
S., & Sable, A. (2004). LIN Protocol –Technology Re-
view and Demonstration in Power Window Application.
Proceedings of the 2004 SAE World Congress, Paper
Number 2004-28-0019.
USB. (2008). Wireless USB. Retrieved October 13, 2008,
from http://www.usb.org/wusb/
USDOT (2006, December). Transportation Statistics
Annual Report, U.S. Department of Transportation,
Bureau of Transportation Statistics.
USDOT. (2008). Retrieved September 2, 2008, from
http://www.its.dot.gov/its_overview.htm
UWB. (2008). Ultra-wideband Technolgy. Retrieved
October 13, 2008, from http://www.intel.com/technol-
ogy/comms/uwb/
Vahidi, A., & Eskandarian, A. (2003). Research advances
in intelligent collision avoidance and adaptive cruise
control. IEEE Transactions on Intelligent Transportation
Systems, 4(3), 143-153.
Valéo (2006). Valéo Inc., Electrical and Electronic Dis-
tribution Systems: Focus on Power Line Communication.
Retrieved from official web http://www.valeo.com/ .
Varaiya, P. (1993). Smart cars on smart roads: Problems of
control. IEEE Trans. Autom. Control, 38(2), 195–207.
Vasilash, G. S. (2005). Hardware-in-the-loop: Assur-
ing performance & quality by combining the real &
the virtual testing. Automotive Design & Production.
Gardner Publications, Inc. Retrieved October 14, 2008,
from http://findarticles.com/p/articles/mi_m0KJI/is_/
ai_n13784131
Vehicle Safety Communications Consortium. (2008).
Retrieved September 2, 2008, from http://www-nrd.
nhtsa.dot.gov/pdf/nrd-12/CAMP3/pages/VSCC.htm
Vejrazka, F. (2007, April). Galileo and the Other Satellite
Navigation Systems. Paper presented at the IEEE 17th
Radioelektronika International Conference.
Victoria, E. (n.d.). How you can save on fuel costs.
(http://www.epa.vic.gov.au/air/savefuel/savefuel.asp,
last accessed: Nov. 2007)
VIIC and VII Program Overview. (2005). Retrieved
September 2, 2008, from http://www.leearmstrong.
com / DSRC%20Home/St a nd a rd s%20P rog r a ms/
N o r t h% 2 0A m e r i c a n / P r e v i o u s % 2 0 M e e t i n g s /
March%2006/VIIC%20Overview.ppt#1
VISSIM (2008). VISSIM Micro-Simulation Software.
Retrieved March 19, 2008 from http://www.english.ptv.
de/cgi-bin/traffic/traf_vissim.pl
Vlacic, L., Parent, M., & Harashima, F. (2001). Intelligent
vehicle technologies: Theory and applications. London:
Butterworth-Heinemann.
327

VSCC. VSC Project Overview. Retrieved June 8, 2008,
from http://www-nrd.nhtsa.dot.gov/pdf/nrd-12/CAMP3/
pages/VSCC.htm
Wagstaff, I. (1999, April). Braking new ground. Automo-
tive Engineer, (pp. 61-62).
Waller, S., & Ziliaskopoulos, A. (2002). On the online
shortest path problem with limited arc cost dependen-
cies. Networks, 40(4).
Wan, E. A., & Merwe, R. V. D. (2000). The Unscented
Kalman Filter for Nonlinear Estimation. Paper presented
at the Symposium on Adaptive Systems for Signal Pro-
cessing, Communication and Control (AS-SPCC-2000),
Lake Louise, Alberta, Canada.
Wang, B., Sen, I., & Matolak, D. (2007, October). Per-
formance evaluation of 802.16e in vehicle to vehicle
channels. IEEE Vehicular Technology Conference (pp.
1406-1410).
Wang, H., Xu, J., Ibañez-Guzmán, J, Jarvis, R., Goh, T.,
& Chang, C. W. (2000). Real Time Obstacle Detection
for AGV Navigation Using Multi-baseline Stereo. In
D. Rus & S. Singh (Eds.), International Symposium on
Experimental Robotics Lecture Notes in Control and
Information Sciences, 271. Springer.
Wang, S. Y., Chou, C. L., Chiu, Y. H., Tseng, Y. S., Hsu,
M. S., Cheng, Y. W., Liu, W. L., & Ho, T. W. (2007,
September). NCTUns 4.0: An Integrated Simulation
Platform for Vehicular Traffic, Communication, and
Network Researches. Paper presented at the 1st IEEE
International Symposium on Wireless Vehicular Com-
munications. Baltimore, MD, USA
Wang, X. (2007). Modeling the process of information
relay through inter-vehicle communication. Transporta-
tion Research Part B, 41(6), 684-700.
Wang, Z., Kulik, L., & Ramamohanarao, K. (2007).
Proactive traffic merging strategies for sensor-enabled
cars. In Proc. VANET ’07 (pp. 39–48). New York, NY,
USA: ACM.
WAVE. (2008). IEEE Wireless Access in Vehicular
Environment (WAVE), IEEE 1609. Retrieved October
328
Compilation of References
13, 2008, from http://standards.ieee.org/announcements/
pr_P1609.0_new.html
Wiedemann, R. (1974) Simulation des Strassenverkeh-
rsflusses. Schriftenreihe des Instituts für Verkehrswesen
der Universität Karlsruhe, Heft 8
Wietfeld, C., & Rokitansky, C.-H. (1994). Performance
of Vehicle-Roadside Communication Systems support-
ing Multiple RTI-Applications. Proceedings in Vehicle
Navigation & Information Systems Conference, (pp.
133-138).
Wietfeld, C., & Rokitansky, C.-H. (1995). Markov Chain
Analysis of Alternative Medium Access Control Protocol
for Vehicle-Roadside Communications. Proceedings in
Vehicular Technology Conference, 2, 958-963.
Willke, T. L., & Maxemchuk, N. F. (2005). Coordinated
interaction using reliable broadcast in mobile wireless
networks. Proceedings of the 4th International IFIP-TC6
Networking Conference, NETWORKING 2005, Waterloo,
Canada (pp. 1168-1179).
Willke, T. L., Tientrakool, P., & Maxemchuk, N. F. (2008).
A Survey of Inter-Vehicle Communication Protocols and
Their Applications. IEEE Communications Surveys and
Tutorials. To appear.
Wolf, W. (2007). The Good News and the Bad News.
Computer, 40(11), 104-105.
Wu, Q., Gao, Z., & Wan, D. (2002, April). An adaptive
information fusion method to vehicle. Paper presented at
the IEEE Position location and navigation symposium.
Proceedings (pp. 248-253).
Xu, Q., Mak, T., Ko, J., & Sengupta, R. (2004). Vehicle-
to-vehicle safety messaging in DSRC. In VANET ‘04, (pp.
19-28), New York, NY, USA. ACM Press.
Xue, F., & Kumar, P. (2004). The number of neighbors
needed for connectivity of wireless network. Wireless
Networks, 10, 169-181.
Yang, X., Liu, J., Zhao, F., & Vaidya, N. H. (2004). A ve-
hicle-to-vehicle communication protocol for cooperative
collision warning. In First Annual International Confer-
ence on Mobile and Ubiqiotous Systems: Networking
and Services.
Compilation of References
Yang, Yi, Varshney, M., Mohan, S., & Bagrodia, R.
(2007). High-Fidelity Application-Centric Evaluation
Framework for Vehicular Networks. In Proceedings of
the fourth ACM international workshop on Vehicular ad
hoc networks (pp 49-58). Montreal, Quebec, Canada
Yih P., & Gerdes, J.C., Modification of Vehicle Handling
Characteristics via Steer-by-Wire. IEEE Trans. on Control
Sys. Tech., 13(6), 965-976.
Yin, J., ElBatt, T., Yeung, G., Ryu, B., Habermas, S.,
Krishnan, H., & Talty, T. (2004). Performance evalua-
tion of safety applications over DSRC vehicular ad hoc
networks. In Proceedings of the 1st ACM international
Workshop on Vehicular Ad Hoc Networks (pp. 1-9).
Philadelphia PA, USA
Yong, M. S. (2008). A*CAR. A*STAR Program Pro-
posal.
Zandbergen, A., & Van der Ree, A. (1992). Experiments
on vehicle to roadside microwave communication. IEEE
MTT-S International Microwave Symposium Digest, 2,
617-620.
Zarki, M. E., Mehrotra, S., Tsudik, G., & Venkata-
subramanianm, N. (2002). Security Issues in Future
Vehicular Network. Proceedings of European Wireless
Conference 2002.
Zeng, K. (2006). Pseudonymous pki for ubiquitous
computing. In Atzeni A. & Lioy, A. (Ed.) EuroPKI 2006,
(pp.207-222), Turin, Italy, Springer-Verlag.
Zhang, Y., Kosmatopoulos, E. B., Ioannou, P. A., &
Chen, C. C. (1999). Autonomous intelligent cruise con-
trol using front and back information for tight vehicle
following maneuvers. IEEE Transactions on Vehicular
Technology, 48, 319-328.
Zhao, Y. (1997). Vehicle Location and Navigation Sys-
tems. Artech House, Norwood, USA
Zheng, B., & Anwar, S. (2008). Fault Tolerant Control of
the Road Wheel Subsystem in a Steer-By-Wire System.
International Journal of Vehicular Technology, Article
ID 859571, doi:10.1155/2008/859571, 1-8.
Zheng, B., Huang, M., Daugherty, B., & Anwar, S. (2006).
Integrated Steer-By-Wire control and diagnostic system.
Proceedings of the American Control Conference, Min-
neapolis, Minnesota.
ZigBee (2000). ZigBee Specification. Version 1.0, De-
cember. Retrieved from http://www.zigbee.org;
Zou, X., Ramamurthy, B., & Magliveras, S. (2002).
Routing Techniques in Wireless Ad Hoc Networks
Classification and Comparison. Proceedings of the Sixth
World Multiconference on Systemics, Cybernetics and
Informatics, SCI.
329
330

About the Contributors

Huaqun Guo obtained BEng and MEng from Tianjin University in 1989 and 1991 respectively, and
obtained M.Eng and PhD from the National University of Singapore (NUS) in 2001 and 2007 respectively.
She was a senior engineer, Kent Ridge Digital Labs (KRDL), Singapore from 2000 to May 2001. She was
a senior research staff, NUS from June 2001 to March 2003. She was an Ordinary Member of Singapore
Computer Society from 2000 to 2005. She is an IEEE member from 2007, a committee member of IEEE
SG WIE AG 2008, and a member of SingAREN (Singapore Advance Research and Education Network)
from 2000. Dr. Guo is currently a research fellow at the Institute for Infocomm Research (I2R), Agency
for Science Technology and Research (A*STAR) in Singapore. She has published number of papers in
the international journals and conferences. Her research areas include multicast, vehicular network, P2P
computing, security, networked middleware, and multimedia communication system.

***

Omar Álvarez-Cardenas received the MSc degree in telematics at the University of Colima in
1999. He is finishing the PhD in telecommunications at the Jose Antonio Echeverria Institute from La
Havana. Since 1998 he has been teaching networking at the School of Telematics where he is a reseach-
professor in telecommunications. His research focuses on network simulations, QoS in metropolitan
ethernet networks, mobile computing and bandwidth efficiency.

Sohel Anwar is an assistant professor in the Department of Mechanical Engineering at School of

Engineering and Technology, Purdue University Indianapolis, Indianapolis, Indiana, USA. He received
his PhD in mechanical engineering in 1995 from the University of Arizona, Tucson, Arizona, USA. Dr.
Anwar worked for Caterpillar, Inc. from 1995 to 1999 where he was involved in developing novel X-by-
wire control algorithms for wheel loaders. In 1999, he joined Ford Motor Company / Visteon Corporation
where he focused on developing advanced fault tolerant control algorithms for drive by wire systems.
Dr. Anwar holds 12 US patents and several foreign patents. He has published over 40 papers in refereed
journals and conference proceedings. He is an associate editor for IEEE Transactions on Vehicular
Technology. His research interests include fault tolerant drive by wire systems, hybrid electric vehicle
control, automotive control systems, and mechatronics. He may be contacted at: soanwar@iupui.edu.

Raúl Aquino-Santos graduated from the University of Colima with a BE in electrical engineering,
received his MS degree in telecommunications from the Centre for Scientific Research and Higher Edu-
cation in Ensenada, Mexico in 1990. He holds a PhD from the Department of Electrical and Electronic

Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.
About the Contributors

Engineering of the University of Sheffield, England. Since 2005, he has been with the College of Tele-
matics, at the University of Colima, where he is currently a research-professor in telecommunications
networks. His current research interests include wireless and sensor networks.

Giorgio Calandriello holds a MSc degree in informatics and system engineering and a PhD in
computer engineering. Currently he’s research assistant at Politecnico di Torino in the TORSEC group,
mainly working on security of vehicular communication. His research interests include wireless systems
security, denial-of-service attacks and vehicular networks security and safety.

Valentin Cristea is a professor of the Computer Science and Engineering Department of University
Politehnica of Bucharest, Romania. The main fields of expertise include distributed systems, vehicular
networks, high performance computing, and collaborative environments. He has a teaching experience
of over 30 years in these domains and published more than 25 books, 130 specialist articles, and 40
scientific reports. Cristea coordinated several projects in mobile and vehicular computing, simulation of
large scale distributed systems, and VANETs simulation. He co-authored several papers on designing
and simulating VANET applications. He is a member of IEEE and ACM Society.

Raluca Diaconescu is a software engineer working for IBM in Bucharest, Romania. She obtained
an MSc degree in 2008 and a BSc degree in 2006 from “Politehnica” University Bucharest, computer
science faculty. She studied VANET applications for her graduation thesis and co-authored several
research papers in the field of VANETs. Her research topics include VANET simulators and collabora-
tive route planning.

Lili Du was born in Luoyang, China, in 1975. She received the BS degree in mechanical engineer-
ing from Xi’an Jiaotong University, Xi’an, China, in 1998, and the MS in industrial engineering from
Tsing Hua University, Beijing, China, in 2003. And then she received the MS in operation research and
statistics and PhD degree in decision sciences and engineering systems from Rensselaer Polytechnic
Institute (RPI), Troy, NY, in 2008. In September 2008, she joined NEXTRANS in Purdue University
as a research associate. Her main research interests are vehicular ad hoc networks (VANETs), on-line
routing, network design under disasters condition, and transportation network modeling.

Arthur Edwards-Block received his masters degree in Education from the University of Hous-
ton in 1985. He has been a researcher-professor at the University of Colima since 1985, where he has
served in various capacities. He has been with the School of Telematics since 1998. His primary areas
of research are computer assisted language learning (CALL), distance learning, collaborative learning,
multimodal leaning and mobile learning. The primary focus of his research is presently in the area of
mobile collaborative learning.

Miguel A. Garcia-Ruiz graduated in computer systems engineering and obtained his MSc in Computer
Science from the University of Colima, Mexico. He received his PhD in Computer Science and Artificial
Intelligence from the School of Cognitive and Computing Sciences, University of Sussex, England. He
took a virtual reality course at Salford University, England and a graphics techniques internship at the
Madrid Polytechnic University, Spain. Miguel has been a visiting professor at the University Of Ontario
Institute Of Technology, Canada. He has been teaching computer science courses, and doing research

331
 About the Contributors

mainly on virtual reality and multimodal interfaces at the University of Colima. Miguel has published
various scientific papers in major journals and a book, and has directed a video documentary about an
introduction to virtual reality.

Denis Gingras received his BS and MS degrees from Université Laval in 1980 and 1984 respec-
tively, and his Dr. Eng degree, in 1989, from Ruhr-Universität Bochum, Germany, all in electrical
engineering. Awarded a STA fellowship in Japan, he was invited researcher at CRL, Tokyo, from 1989
to 1990. From 1990 to 1998, he directed the digital and optical systems department at INO, in Canada.
In 1999, he was invited professor at INRIA in Sophia-Antipolis, France. Since 2000, he is professor
at Université de Sherbrooke, where he has also been director of the Intelligent Materials and Systems
Institute from 2000 to 2008. Dr. Gingras has published two books and more than eighty publications.
His research interests cover intelligent systems and information processing. Dr. Gingras was leader of
the Intelligent Vehicles & Intelligent Vehicle-Highway Systems Working Group within the ATLANTIC
Canada International Network and is leading the Intelligent Systems and Sensors research program in
the Canadian AUTO21 NCE.

Apolinar Gonzalez-Potes holds a PhD from Polytechnic University of Valencia, Spain. Since
2004, he has been with the Faculty of Mechanical Electrical Engineering of the University of Colima,
where he is currently a research-professor in informatics. His current research interest includes mobile
computing and embedded software.

Cristian Gorgorin is a software engineer working for S&T in Bucharest, Romania. He obtained
an MSc degree in 2008 and a BSc degree in 2006 from “Politehnica” University Bucharest, Computer
Science Faculty. He studied VANET routing protocols for his graduation thesis and co-authored several
research papers in the field of VANETs. His research topics include VANET simulators and algorithms
for intelligent traffic lights.

Wilfried Gouret received the MSc degree in electrical engineering from University of Rennes,
France in 2004 and and the PhD degrees from National Institute of Applied science (INSA), Rennes,
France in 2007. From 2004 to 2007, he was with the research laboratory IETR (Institute of Electronics
and Telecommunications of Rennes) where he worked on the optimization of power line communica-
tions for automotive application. A demonstrator has been realized and demonstrates the feasibility
of such system. Currently, he is research engineer and his current research focuses on high data rate
wireless communications

Victor Gradinescu is a software developer for S&T Romania. He has received a BSc degree and
MSc degree in computer science from Politehnica University of Bucharest. His research interests have
focused on the design and simulation of VANET applications, such as intersection cooperation and
adaptive traffic lights, with a particular emphasis on mobile wireless networks simulation, analysis and
optimization.

Todd Hubing is the Michelin Professor of Vehicular Electronics at Clemson University. He holds a
BSEE from MIT, an MSEE from Purdue, and a PhD from North Carolina State University. He began
his career as an EMC engineer for IBM in 1982, where he did EMC testing and troubleshooting on a

332
About the Contributors

variety of computer and network communications products. In 1989, he became a faculty member at the
University of Missouri-Rolla (UMR) where he worked with other faculty and students to analyze and
develop solutions for a wide range of EMC problems affecting the electronics industry. Since coming
to Clemson in 2006, he has continued his work in electromagnetic compatibility and computational
electromagnetic modeling, particularly as it is applied to automotive and aerospace electronics. Dr.
Hubing is a fellow of the Institute of Electrical and Electronics Engineers (IEEE) and a past president
of the IEEE Electromagnetic Compatibility Society.

Javier Ibañez-Guzmán obtained his PhD at the University of Reading on a SERC-UK assistantship
and his MSEE at the University of Pennsylvania (USA) as a Fulbright scholar. He is currently working
for Renault S.A where he is responsible for localisation technologies used in advanced driving assistance
systems. Dr Ibañez-Guzmán has worked as senior scientist at SimTECH, a national research institute
in Singapore, where he spearheaded research on autonomous ground vehicles for unstructured envi-
ronments. He was responsible for several projects sponsored by the ministry of defence. Concurrently,
he was adjunct associate professor at Nanyang Technological University. His research interests are in
localisation and perception for intelligent vehicles. He has worked in manufacturing robotics in France
and was Maître de Conferences at the engineering school ESA IGELEC (part of the Ecole Polytechnique
de l’Université de Nantes). Dr. Ibañez-Guzmán is Fellow of the Institute of Engineering Technology
and a Chartered Electrical Engineer (C. Eng.) of the Engineering Council of the U.K. He is member of
several societies within the IEEE.

Liviu Iftode is an associate professor of computer science at Rutgers University, New Jersey. He
received his PhD in computer science from Princeton University in 1998. His research interests in-
clude distributed systems, operating systems, vehicular networks, mobile and pervasive computing.
He is a senior member of the IEEE Computer Society and a member of the ACM. Liviu Iftode is the
vice-chair of IEEE Technical Committee on Operating Systems and a member of the editorial board of
IEEE Pervasive Computing. In the area of vehicular computing, Dr. Iftode led the TrafficView research
project, one of the first traffic monitoring and warning system based on cat-to-car communication, and
is a co-author of VITP (Vehicular Information Transfer Protocol). More information can be found at
http://www.cs.rutgers.edu/~iftode/.

Shivkumar Kalyanaram is a professor in the Department of Electrical and Communications Sys-

tems Engineering at Rensselaer Polytechnic Institute. His research interests are in the area of computer
networking, concentrated around the theme of traffic management and high performance wireless net-
working. In particular, he works on topics of congestion control, reliability, connectionless traffic engi-
neering, quality of service (QoS), last-mile community wireless networks, low-cost free-space-optical
networks, automated network management using online simulation, multicast, multimedia networking
(including peer-to-peer multimedia systems), and performance analysis.

Lars Kulik is a senior lecturer at the Department of Computer Science and Software Engineering
at the University of Melbourne. Until June 2004, he was a postdoctoral researcher at the Department
of Spatial Information Science and Engineering, University of Maine. In 2002, he completed his PhD
at the University of Hamburg. His overall research goal is to develop a theory of spatial information
for building pervasive computing systems that anticipate, adapt and respond to the needs of users, and

333
 About the Contributors

provide services based on the user’s location and context. Specifically, he focuses on efficient algorithms
for moving objects, optimization of traffic throughput using sensor-enabled cars in VANETs, informa-
tion dissemination algorithms in sensor networks, location privacy, and more generally, efficient spatial
algorithms in pervasive computing environments.

Antonio Lioy holds a MSc degree in electronic engineering (summa cum laude) and a PhD in com-
puter engineering. Currently he is full professor at the Politecnico di Torino where leads the TORSEC
research group that investigates information systems security. This group has taken part to several inter-
national security projects, the most recent ones being the European projects Positif, Deserec, Open-TC,
and Stork. Since September 2007 Prof. Lioy is a member of the PSG (Permanent Stakeholders’ Group)
of ENISA (the European Network and Information Security Agency) and often acts as a consultant and
reviewer for public and private bodies on several security issues. His current research interests are in
the fields of network security, PKI, and policy-based system protection.

Xiaobo Long is a PhD student in the Department of Electrical, Computer and System Engineering
at Rensselear Polytechnic Institute, Troy, NY, USA. She got her bachelor's degree (2001) and master's
degree (2004) both from University of Electronic Science and Technology of China. Currently, she
is working with Professor Biplab Sikdar in Networking Lab. Her research is focused on link level
characteristics of wireless communications and computer networks, such as link failures detection in
computer networks, wireless channel quality prediction, shadow region detection and intrusion detec-
tion at PHY layer in wireless networks. She is also interested in automatic detection and decision and
pattern recognition.

Syed Masud Mahmud is an associate professor of electrical and computer engineering. He has been
long recognized for his teaching excellence in the University. He has been instrumental in educational
innovation, such as developing and teaching the first distance-learning course for Wayne State University.
In 2002 Dr. Mahmud received the University’s Presidential Award in Excellence in Teaching. In 1994
and also in 2001, he was the recipient of the College of Engineering’s Teaching Excellence Award and
received a number of Outstanding Faculty Awards from his College. During the last eighteen years, Dr.
Mahmud supervised ten research and development projects funded by local industries. These projects
involved developing embedded hardware and software, and simulation models for embedded systems.
He also served as a Co-PI of an NSF funded project to develop an undergraduate laboratory for teaching
Real-Time Embedded System Networking. He has published over 100 peer reviewed technical journal
and conference papers. He also published two editions of a book entitled “Lecture Notes On 16- And
32-Bit Microprocessors” through Simon & Schuster Custom Publishing.

Nicholas F. Maxemchuk, received the BSEE degree from the City College of New York, and the
MSEE and PhD degrees from the University of Pennsylvania. For the past 7 years he has been a pro-
fessor in the electical engineering department at Columbia University in New York City. He has a dual
appointment as a chief researcher in IMDEA Networks in Madrid. Prior to joining Columbia he spent
25 years at Bell Labs and AT&T Labs, as a member of the technical staff, department head, and techni-
cal leader. Prior to joining Bell Labs, he spent 8 years at the RCA David Sarnoff Research Center, as a
member of the technical staff. He has been the editor-in-chief of the IEEE Journal on Selected Areas
in Communications, an editor for the IEEE Transactions on Communications, and the JACM, was on

334
About the Contributors

the founding committee of the IEEE/ACM Transactions on Networking, and served on their steering
committee for 11 years. In 2006 he received the Koji Kobayashi Award for his work in computer com-
munications. He has also been awarded the IEEE’s Leonard G. Abraham Prize Paper Award in 1985
and 1987, for his papers on data and voice on CATV networks and the Manhattan Street networks, and
the William R. Bennett Prize Paper Award in 1997, for his paper on an anonymous credit card.

Margarita G. Mayoral-Baldivia received the MSc degree in telematics at the University of Colima
in 1999. Since 1996, she has been working in the networking area and actually she is a research-professor
in telecommunications networks and CCNA Cisco Networking Academy instructor. Her main research
focuses in security, wired and wireless networks and bandwidth efficiency.

Patrice Maziero was born in Toulouse, France, on December 18, 1984. He received the electronic
engineer degree from Ecole Nationale d’Ingénieur (ENIB) of Brest, France in 2007. From 2006 to
2007 he worked on VHDL school project and at TE2M on the creation of a magnetic sensor for crack
and corrosion detection on metallic surface covered by thick paint layer. Since 2008, he is a research
engineer at the Electronic and Telecommunication Laboratory of Rennes, France (IETR). He works on
the automotive network architecture for ECU’s Communication project, especially on PLC and wireless
intra-car communication medias. He is involves on the French project ANR CIFAER.

Lek Heng Ngoh obtained BSc (Honors) in computer systems engineering from University of Kent
at Canterbury, UK in 1986 and obtained MSc and PhD degrees in computer science from University of
Manchester, UK in 1987 and 1989 respectively. He has organized numerous international networking
conferences, and a review board member of several international journals. He is a founding member of
the Singapore Advance Research and Education Network (SingAREN) — an advanced Internet Initia-
tive. He is currently a senior research scientist at the Institute for Infocomm Research (I2R), Agency for
Science Technology and Research (A*STAR) in Singapore. Dr. Ngoh is also an adjunct associate profes-
sor with the Nanyang Technological University in Singapore. His research interests include: multimedia
network protocols, advanced internet technology, P2P mobile computing & wireless sensor networks.

Fabienne Nouvel received the MSc and PhD degrees in electrical engineering from National In-
stitute of Applied science (INSA), Rennes, France in 1984 and 1994 respectively. From 1984 to 1990,
she was an engineer and worked on DAB, networks and embedded application. In 1990, she joined the
research laboratory IETR (Institute of Electronics and Telecommunications of Rennes) and worked on
automotive application and the optimization of high data rate networks. In 1995, she joined INSA and
currently she is professor in the Department of Telecommunications and Electronic Engineering. She
is involved in several European and national research projects in the field of hardware implementation
on heterogeneous embedded plate-form and power line transmissions. Her research interests lie both in
digital implementation and multicarrier modulation, spread spectrum and multi-user communication.
Current research focuses on power line for automotive combined with wireless system, implemented
on a reconfigurable plate-form.

Kotagiri Ramamohanarao (Rao) received PhD from Monash University. He was awarded the
Alexander von Humboldt Fellowship in 1983. He has been at the University Melbourne since 1980 and
was appointed a professor in computer science in 1989. Rao held several senior positions including

335
 About the Contributors

Head of Computer Science and Software Engineering, Head of the School of Electrical Engineering and
Computer Science at the University of Melbourne and Research Director for the Cooperative Research
Centre for Intelligent Decision Systems. He served on the editorial boards of the Computer Journal. At
present he is on the editorial boards for Universal Computer Science, the Journal of Knowledge and
Information Systems, IEEE Transactions on Knowledge and Data Engineering (TKDE), Journal of
Statistical Analysis and Data Mining and Very Large Data Bases (VLDB) Journal. He was the program
Co-Chair for VLDB, PAKDD, DASFAA and DOOD conferences. He is a steering committee member
of IEEE ICDM, PAKDD and DASFAA. He received distinguished contribution award for data mining.
Rao is a fellow of the Institute of Engineers Australia, a fellow of Australian Academy Technological
Sciences and Engineering and a fellow of Australian Academy of Science.

Víctor Rangel-Licea received the B.Eng (Hons) degree in computer engineering in the engineering
faculty from the National Autonomous University of Mexico (UNAM) in 1996, the MSc in Telemat-
ics at from the University of Sheffield, U.K. in 1998, and the PhD in performance analysis and traffic
scheduling in cable networks in 2002, from the University of Sheffield. Since 2002, he has been with
the School of Engineering, UNAM, where he is currently a Research-Professor in telecommunications
networks. His research focuses on fixed, mesh and mobile broadband wireless access networks, QoS
over IP, traffic shaping and scheduling.

Sara Sandoval-Carrillo graduated from the Technological Institute of Colima with a BE in infor-
matics, received his MS degree in telematics from the University of Colima, in 1998. Since then, she
has been with the College of Telematics, at the University of Colima, where she is currently a research-
professor in information system.

Biplab Sikdar received the B. Tech degree in electronics and communication engineering from
North Eastern Hill University, Shillong, India, the M. Tech degree in electrical engineering from In-
dian Institute of Technology, Kanpur and PhD in electrical engineering from Rensselaer Polytechnic
Institute, Troy, NY, USA in 1996, 1998 and 2001, respectively. He is currently an associate professor in
the Department of Electrical, Computer and Systems Engineering of Rensselaer Polytechnic Institute,
Troy, NY, USA. His research interests include computer networking protocols, network security and
queueing theory. Dr. Sikdar is a member of IEEE, Eta Kappa Nu and Tau Beta Pi and currently serves
as an associate editor of the IEEE Transactions on Communications.

Patcharinee Tientrakool is a PhD student at the department of Electrical Engineering, Columbia

University, New York, NY. She received the BS degree in telecommunication engineering from King
Mongkut’s Institute of Technology Ladkrabang, Bangkok, Thailand in 2000 and the MS degree in
electrical engineering from Columbia University in 2006. Currently she is an engineer at CAT Telecom
Public Company Limited, Thailand. Her research interests include vehicular networks and mobile ad
hoc networks.

Satish Ukkusuri is an assistant professor in the Department of Civil and Environmental Engineer-
ing at Rensselaer Polytechnic Institute. Professor Ukkusuri’s expertise is in transportation network
modeling with applications to planning and operations. Ukkusuri received the SWUTC Robert Herman
Award for outstanding research on transportation network modeling and leadership in transportation

336
About the Contributors

during the year 2005. He also holds the Blitman Career Development Chair at RPI which facilitates to
conduct research on transportation systems. Professor Ukkusuri is research centers around large scale
transportation network modeling and its applications to practice. Apart from his active research, he has
been involved in various educational and outreach activities. He presently serves as a faculty advisor
for ITE with an overall interest in contributing to the community.

Ziyuan Wang received her BEng degree in electric and information engineering from Dalian Jiao-
tong University in 2005. Currently, she is a PhD student at the Department of Computer Science and
Software Engineering at the University of Melbourne. She is working on vehicular ad hoc networks
(VANETs), in particular, how to optimize traffic throughput when there are conflicting traffic streams.
Her research goal is to develop proactive traffic control strategies for sensor-enabled cars to mitigate
traffic congestion and reduce fuel consumption in a decentralized manner.

Theodore L. Willke received the BS and MS degrees in electrical engineering and computer science
from the University of Illinois, Chicago, in 1993 and 1995, respectively, and the MS degree in electrical
engineering from the University of Wisconsin, Madison, in 1998. He is currently working toward the
EngScD degree in electrical engineering at Columbia University, New York, NY. In 1998, he joined
Intel Corporation, DuPont, WA, where he is currently a principal engineer with the digital enterprise
group, where he focuses on I/O technologies and standards. His research interests include wireless
communication networks, I/O protocols, and computer system architectures.

Ghais El Zein received the PhD and Habilitation a Diriger des Recherches (HDR) degrees in tele-
communications signal processing and electronics from Rennes 1 University, France, in 1988 and 1998,
respectively. From 1985 to 1987, he was a lecturer and he worked on the first communication systems
using spread spectrum in different applications: wireless, power line, multiplexing. In 1988, he joined
INSA and currently he is professor in the Department of Telecommunications and Electronic Engineer-
ing. Since 2001 he is the assistant director of the Institute of Electronics and Telecommunications of
Rennes (IETR) which depends on the French National Center for Scientific Research (CNRS). He is a
member of Commissions C and F of CNFRS-URSI. He is involved in several European and national
research projects in the field of radiowave propagation, combined with mimo applications and their ef-
fects on communication systems performance. His research interests lie in spread spectrum modulation,
multi-user communication, space time coding.

337
338

Index

A co-operative vehicle infrastructure systems

(CVIS) 201,  206,  218,  253,  261,  308
ad hoc networks 109,  113,  114,  117,  118,  communications, car-to-car 6,  9,  96,  98,  99, 
119,  121,  122,  123,  124,  129,  132,  100,  101,  103,  104,  159,  265,  266, 
135,  136,  141,  142,  160,  164,  163,  279,  306,  307,  309,  317,  320
178,  120,  178,  137,  178,  179,  180,  communications, intra-vehicle (InV)
237,  263,  280,  281,  282,  305,  306,  1,  2,  3,  9,  1,  3,  4
310,  312,  313,  314,  315,  317,  318,  communications, vehicle-to-infrastructure
322,  323,  326,  327,  329 (V2I) 1,  2,  3,  4,  5,  6,  8,  9,  3,  4, 
advanced driving assistance systems (ADAS) 200,  201,  202,  205,  213,  217,  248
201,  210,  215 communications, vehicle-to-vehicle (V2V)
advanced vehicle safety systems (AVS) 105 1,  2,  3,  4,  5,  6,  8,  9,  3,  4,  105, 
anti-lock braking system (ABS) 12,  13,  17,  200,  201,  202,  204,  205,  213,  216, 
18,  19,  24,  25,  39,  76,  97,  224,  217,  248
242,  306 consortia, LIN (Local Interconnect Network)
ATIS (advanced traveler information systems) 2,  29,  62,  63,  71,  72,  75,  76,  84,  85,
162,  163,  164,  179,  325   86,  87,  89,  90,  98,  104,  284,  288, 
ATIS, decentralized 163,  164 302,  319,  327
automated highway systems (AHS) constant headway state 174
105,  188,  189,  198,  319 controller area network (CAN) 2,  4,  9,  10, 
automotive consortia 8 29,  39,  40,  42,  62,  63,  70,  71,  72, 
B 73,  75,  76,  84,  86,  87,  88,  89,  90, 
91,  92,  93,  94,  95,  96,  97,  98,  99, 
bottlenecks 72,  75,  180,  181,  186 100,  101,  103,  104,  284,  286,  287, 
brake by wire (BBW) systems 13,  17,  19,  22, 288,  296,  297,  298,  299,  300,  301, 
  24,  39,  45 302,  303,  304,  305,  307,  308,  309, 
311,  317,  320,  323
C cooperative adaptive cruise control (CACC)
car communications 2,  6,  9,  69,  83,  87,  96, 201
  98,  99,  101 crosstalk 65
CISPR 12 standard 49,  67,  314 cyber-physical systems 139,  140,  141,  159
CISPR 25 standard 49,  50,  51,  55,  81

Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.
Index

D G
data exchange 1,  2,  3,  5,  9,  123 generalized prediction (GP)
dedicated short range communication (DSRC) 33,  34,  35,  37,  38,  39
2,  3,  4,  5,  7,  9,  11,  64,  99,  100,  geographic information system (GIS)
103,  106,  112,  142,  159,  164,  181,  116,  220,  270,  273
250,  251,  256,  261,  263,  269,  278,  greedy perimeter stateless routing (GPSR)
281,  282,  304,  310,  314,  326,  327,  115,  116,  121,  124,  129,  132,  134, 
328,  329 135,  270,  280,  316
distance routing effect algorithm for mobility grid location service (GLS) 115,  121
(DREAM) 115,  121,  270,  279,  305
drive-by-wire system 12 H
drive by wire (DBW) 12 headway state, constant 174
drive by wire (DBW) systems 12,  13,  14,  17, headway state, intermediate 174
  22,  23,  24,  25,  26,  28,  29,  30,  31,  headway state, random 174
39,  40,  44,  45 human-machine interface (HMI) 220
hybrid-electric propulsion 12
E
hybrid electric vehicles (HEV) 13,  14,  17
electric vehicles (EV) 13
electro-hydraulic brake (EHB) system 13,  14 I
electro-mechanical brakes (EMB) 13 infotainment applications 69
electromagnetic compatibility (EMC) 48,  49,  intelligent cruise control 161,  329
50,  53,  54,  56,  57,  60,  61,  62,  67,  intelligent transportation systems (ITS) 2,  3, 
61,  64,  79,  83,  87,  311,  54,  62,  65,  4,  5,  7,  8,  10,  11,  105,  121,  142, 
67,  66,  67,  64 159,  163,  179,  202,  206,  209,  262, 
electromagnetic interference (EMI) 49,  51,  310,  311,  314,  324
54,  55,  56,  64,  67 interference 42,  48,  49,  51,  54,  55,  56, 
electronic control units (ECUs) 1,  2,  1,  4,  57,  58,  60,  62,  63,  64,  66,  67,  77, 
29,  43,  69,  70,  72,  75,  82,  87,  91,  78,  79,  84,  85,  106,  107,  162,  163, 
93,  95,  96,  100 165,  166,  167,  168,  167,  168,  169, 
electronic noise 17,  26,  39,  49,  52,  54,  170,  171,  173,  174,  175,  177,  178, 
55,  56,  57,  62,  63,  64,  65,  77,  78,  179,  232,  235,  250,  258,  269,  305, 
48,  78,  66,  78,  79,  83,  86,  90,  98,  306,  311,  315,  320,  321,  323
171,  221,  223,  224,  225,  227,  229,  internal combustion (IC) engine
232,  233,  235,  240,  243,  308,  325 13,  14,  67,  309
electrostatic discharge (ESD) 52,  63,  64,  54 intersection decision systems (IDS) 105
Intra-Vehicle (InV) 1,  9
F
fault tolerant communication 43 K
fault tolerant control 25,  28,  29,  30,  38,  Kalman filter 220,  221,  224,  225,  226,  227,
39,  45   228,  229,  230,  231,  235,  236,  240, 
FlexRay 2,  10,  39,  40,  41,  42,  43,  44,  63,  245,  326
64,  67,  73,  75,  88,  91,  98,  103,  287,
  292,  296,  299,  302,  311 L
4-way stop problem 156
Ladars (Laser detection and ranging sensors)
201,  203

339
 Index

location-based routing algorithm with cluster- Steer by wire (SBW) 13

based flooding (LORA-CBF) 121,  125, steer by wire (SBW) systems 13,  17,  21,  22, 
  127,  128,  129,  132,  134,  135 24,  30,  31,  35,  36,  37,  39,  45
location aided routing (LAR) 115,  121
T
M
throttle by wire (TBW) systems
medium access control (MAC) 5,  81,  105,  13,  18,  19,  39
106,  108,  109,  110,  109,  130,  132,  time-period vehicular degree (TPVD) 170
269,  273,  306,  111,  112,  113,  118,  traction control system (TCS) 12,  18,  19
110 traffic flow control 181,  187,  196
merging algorithms 182,  190,  191,  192,  194, traffic light control 276
  195,  196,  191,  194 traffic network simulator (TRANS) 265

N U
neighborcast 139,  151,  152,  157 unlicensed national information infrastructure
nodes, exposed 166,  167 (UNII) 105
nodes, hidden 110
nodes, mobile 109,  121,  122,  127,  130,  V
185,  247 vehicular ad hoc networks (VANETs) 120, 
121,  132,  141,  142,  160,  162,  163, 
P
166,  167,  168,  169,  170,  171,  173, 
particle filter 229 135,  165,  164,  177,  166,  175,  178, 
platooning 22,  143,  145,  159,  160,  144,  179,  180,  181,  182,  185,  196,  262, 
188,  189,  208,  312,  326 263,  264,  265,  266,  267,  268,  269, 
power line communication (PLC) 69,  75,  270,  272,  273,  274,  275,  276,  277, 
76,  77,  78,  79,  81,  82,  83,  85,  87,  278,  280,  281,  282
90 VANET applications 264,  265,  266,  268, 
273,  278
R Vehicle-to-Vehicle (V2V) 1,  9
Radar 185,  197,  201,  309 vehicular ad hoc networks (VANETs) 142, 
random waypoint model (RWP) 164,  165,  170,  173,  175,  177,  182, 
181,  182,  184 184,  185,  197,  198,  205,  206,  247, 
redundancy, analytical 25,  26,  28,  30,  31,  248,  249,  250,  251,  101,  178,  199, 
39,  45 252,  251,  255,  260,  261,  262,  263, 
redundancy, dynamic 27,  28,  29 304,  306,  307,  308,  310,  312,  313, 
redundancy, static 27,  29 316,  317,  318,  319,  322,  323,  324, 
road side units (RSU) 202,  208,  247,  255 326,  327,  328,  329
route planning 116,  140,  142,  265,  275,  277 vehicular degree, static (SVD) 170
vehicular degree, time-period (TVD) 170
S vehicular networks 1,  2,  3,  9,  91,  98,  105, 
109,  100,  110,  111,  112,  115,  116, 
sliding mode observer
118,  141,  248,  249,  263,  266,  274, 
30,  31,  32,  39,  46,  313
275,  281,  288,  290,  292,  296,  301, 
static vehicular degree (SVD)
310,  322,  323
170,  171,  172,  177

340
 341

W X
wireless access in the vehicular environment X-by-wire systems 46,  69,  70,  73,  74,  75, 
(WAVE) 2,  5,  11,  103,  104,  105,  84,  86,  87,  325
106,  107,  112,  118,  142,  159,  164,  X-tainment applications 69
310,  314,  328
wireless networks 81,  86,  109,  110,  119,  Y
123,  136,  137,  142,  143,  146,  160,  yaw stability control (YSC)
168,  179,  197,  198,  269,  280,  307,  12,  18,  19,  24,  25
308,  315,  316,  323,  324,  328
wireless technologies 1,  3,  9,  69,  86,  141, 
142