Sie sind auf Seite 1von 11

Public transport priority schemes comparing microsimulation

with traditional TRANSYT and LINSIG models

Sonal Ahuja
Tom van Vuren

Mott MacDonald

Neil Priest
Wolverhampton City Council

1.0 INTRODUCTION

Traffic signals are essential traffic management tools in urban conditions. The
design of signals has been an evolving science, driven by the need to make best
use of available road space and changing policy objectives; moving away from
capacity maximisation or delay minimisation for private vehicles, towards priority
operations for public transport, recognising the passenger numbers transported.
In addition, many authorities now actively wish to pursue policies aimed at modal
shift towards non-motorised and public transport modes, whilst environmental
and safety concerns also come into play.

Wolverhampton City Council identified two congested movement corridors where


the need optimise movement of vehicles (buses and cars) and pedestrians was
desired. A comprehensive study was conducted to evaluate the provision of bus
priority and test alternative signal plans allocating a greater priority to the
movement of public transport vehicles.

Traditional analysis of signalised intersections was carried out using TRANSYT


and LINSIG models. However, the models showed poor fit to the base year
situation due to the inability to explicitly model blocking back from one junction to
another. As an alternative design tool a VISSIM microsimulation model was set
up to assess the feasibility of public transport prioritisation along the corridor, and
to identify the appropriate signal timings. Comparative results indicate that
microsimulation is a more accurate tool to analyse the operation of public
transport vehicles within closely located congested signalised intersections.

2.0 PROBLEM DEFINITION

The key objective of the exercise was to assess provision of public transport priority
and test alternative signal designs on the Waterloo Road corridor and the Stafford
Road (A449) corridor in Wolverhampton City. Both corridors at present experience
severe congestion to traffic movement into the city in the AM peak and out of city in
the PM peak.

Association for European Transport 2003


Public transport priority was targeted to be achieved by the provision of priority
signal control and dedicated bus lanes, removal of on-street parking and other
infrastructure improvements. The key tasks of the exercise were.

A validated base model of current situation was desired to be built in the first
case
The model was utilised to test alternative signal and intersection designs.
Detailed signal plans explaining the operation of bus priority were output.
An indication of future bus journey times and average delays and queues at
intersections was provided.

The intersections examined are shown in Figure 1.

Figure 1: Case study intersections.

3.0 LINSIG AND TRANSYT MODELS

Initially only LINSIG and TRANSYT models were built to model the traffic on the
corridor. LINSIG is a software package assists in the design of isolated
intersections and helps in assessing the performance of signalised intersections.
LINSIG can accurately model signal controllers that are based on UK design and
standard specifications. LINSIG an be utilised to optimise isolated signals

TRANSYT is a computer program for determining and studying optimised fixed-


time co-ordinated traffic signal. A traffic model of the network calculates a
Performance Index (PI) in monetary terms. TRANSYT can be utilised to optimise
traffic signals that results in minimum PI subject to signalling constraints.

Association for European Transport 2003


Both TRANSYT and LINSIG are based on statistically derived driver-vehicle
behaviour rather than exact stochastic relationships. They both calculate saturation
flows based on TRL research report 67.

Both LINSIG and TRANSYT models for the case study areas failed to model the
intersections accurately. The base year model could not be validated against the
observed delays and queues. LINSIG and TRANSYT models indicated much
smaller delays and queues than observed on the ground (Figure 2). Even fine-
tuning and adjusting input parameters could not match the model conditions to the
observed. LINSIG and TRANSYT failed to represent the congested corridor
properly due to following reasons.

Inaccurate representation of blocking back;


Vertical queuing assumptions lead to underestimating the interactions
between closely spaced junctions; and
Inability to model dynamic changes in demand, particularly pedestrian
demand

In addition the models were not capable of modelling demand dependent priority
based on the arrival of vehicles. To model public transport priority stages one had
to make assumptions like the number of times the stage shall be called per hour.
This lead to a bias against the priority stage and inaccurate forecast of the working
of the signal controller. One could not assess the impact public transport priority on
the remaining road traffic as all traffic was modelled as one user group and not
individual vehicles. Furthermore details of the operation of priority control, like the
impact of the location of priority detectors and the effect of alternate signal design
could not be clearly assessed. The signal control logic was proposed to be
developed with public transport priority based on IF THEN ELSE statements that
could drive the controller to specific stages based on demand. However, the above
could not be modelled accurately.

LINSIG and TRANSYT models took no account of impact of signals from public
transport operation viewpoint. One could not assess the impact of delayed or early
arrivals of priority vehicles. The Stafford Road corridor also involved roundabouts in
and dedicated bus lanes in the model. These could not be modelled accurately
within the above conventional software. The greatest limitation of the above
software was its inability to model the above corridors as one unit.

3.0 MICROSIMULATION MODEL FOR PUBLIC TRANSPORT PRIORITY

A VISSIM microsimulation model was set up as an alternative design tool to assess


the feasibility of public transport prioritisation along the corridor, and to design the
appropriate signal timings. This involved manipulation of two aspects of the model

Association for European Transport 2003


Modelling the operation of a bus lane, and the associated change in
behaviour by car drivers; and
Development of a traffic signal optimiser that minimises some measure of
delay for road traffic whilst allocating priority to public transport, pedestrians
and cyclists.

VISSIM is discrete - stochastic microscopic traffic simulation flow model developed


by PTV AG in Germany and is based on psycho-physical car following model
proposed by Wiedemann (1974, 1991) as described in Hoyer, Fellendorf, (1997).
Microscopic traffic simulation reflects greater detail as it requires individual driver,
vehicle and environment elements to be coded. Gap acceptance and saturation
flow characteristics, which are derived in a deterministic way; based on statistical
equations in LINSIG and TRANSYT software, is generated more accurately in
VISSIM based on road geometry and individual driver-vehicle responses to
stochastic relationships. In microscopic traffic simulation the entire network can be
modelled as a continuous entity with individual vehicle movements, hence one can
assess the effect of blocking back more accurately.

One of the key features of VISSIM is the signal state generator. In VISSIM one can
either create fixed time traffic signals or programme flexible traffic controller using
interactive and flexible Vehicle Actuated Program (VAP) programming tool. Using
the VAP one can develop signal controller with tram, pedestrian or vehicular
priority. Within a VISSIM simulation vehicle movements and their characteristics are
detected in the network through as series of detectors. VAP pools detectors for
traffic information to pass to the signal controller, which generates varied signal
responses based on programmed control logic. Using VAP one can develop exact
signal control logic to allocate priority to various road users. One can change stage
sequencing, allocate green times based on demand or even code signal
optimisation algorithms that minimise delays to traffic.

One of the main features of VISSIM microsimulation models is its ability to generate
an accurate graphical representation of the simulation that can be used to
interrogate the models and signal control logic.

4.0 COMPARISON OF BASE MODEL RUNS

For implementing public transport priority schemes where considerable time,


resources and money were allocated to be spent there was greater need to know
the accurate working of the traffic signals and check if the suggested signal and
road infrastructure made improvements. These effect of changes need to be
quantified accurately.

Based on the above reasons VISSIM microsimulation models were developed for
the case study locations for the AM and PM peak periods. The base year VISSIM

Association for European Transport 2003


models were calibrated to observed travel times, queues, saturation flows and
driver behaviour.

LINSIG and TRANSYT failed to represent the congested corridor properly as they
forecast much lower travel times and delays at intersections. This was due to
inaccurate representation of blocking back, the vertical queuing assumptions
underestimation of the interactions between closely spaced junctions. LINSIG and
TRANSYT model were unable to model dynamic changes in demand, particularly
pedestrian demand. The results gave poor validation, i.e. fit with observed
conditions (queues and delays). One could not quantify the full benefits to be
achieved from public transport priority due to poor model representation. Moreover
LINSIG and TRANSYT models had poor graphical visualisation of schemes.

VISSIM models of the base year were then developed. These showed a larger
accuracy of bus and car travel times and delays to the observed base situation.

The above results are presented in Figure 2.

16.56
13.43
Bus Inbound
12.67
16.89

9.83
8.50
Bus Outbound
7.09
10.12
VISSIM Base Year
TRANSTY Base Year
12.73 LINSIG Base year
7.40 2002 Observed AM
Car Inbound
7.26
12.09

4.26
2.73
Car Outbound
2.28
4.56

Journey Time (min)


0 2 4 6 8 10 12 14 16 18

Figure 2: Comparison of Base Year travel times across the network.

Association for European Transport 2003


Using VISSIM it was possible to develop a calibrated and validated model of the
study area. The model depicted accurate validation of flows, bus and car journey
times across the network and queues.

5.0 TESTING OF FUTURE OPTIONS

On the Waterloo Road corridor two design options were proposed to be tested,
namely:

Option 1: Lane improvements on Ring Road intersection and signal


optimisation on the same intersection; and
Option 2: Lane improvements on Ring Road and Newhampton Road
intersections and signal and offset optimisation on the same intersection.

Due to the fact that LINSIG and TRANSYT models depicted poor fit with the base
models LINSIG and TRANSYT models were developed only for Option 1.

6.0 OPTIMISING TRAFFIC SIGNALS WITH PUBLIC TRANSPORT PRIORITY

IN VISSIM

VISSIM by itself is not a signal optimisation tool. It is rather a signal evaluation tool.
However the Vehicle Actuated Programming (VAP) interface within VISSIM offers a
viable tool to develop and test optimisation techniques.

It was proposed that the future signals would be running using SCOOT signal
optimisation program. Bus and pedestrian priority would be built into the SCOOT
controller that would allocate priority stages according the demand and developed
logic. However the SCOOT optimisation algorithm is not available in the public
domain. Thus a SCOOT type controller was developed within VISSIM that
optimised cycle times, stage green times offsets and delays to vehicles based on
Webster and Cobbe, 1966.

Additional pedestrian and bus priority stages were programmed that could be called
in when required. The developed signal optimisation controller is shown in Figure 3.

Association for European Transport 2003


1 .5 L + 5 1 .5 L + 5
c = = .......... .......... ........( 1 )
0 1 y 1 y 2 y 3 ...... y n 1Y
where..... ..c o - optimum cycle time (in seconds)
L - Lost time per cycle (in seconds)
y 1, y 2 , y 3 ,......, y n are the maximum ratios of flow
to saturation flow for stages 1,2,3,.... ...., n , Y = y
q1 q 2 qj
therefore , yi,......... , =) max(
s1 s 2 sj
where..... q j = Flow of phase j of stage i .
sj = Saturation flow of phase j of stage i .
and
yi
gi =
( c 0 L )......... .......... .......... .......... .( 2 )
Y
where..... ..... g i is the effective green time for stage i

(Webster-Cobbe)1

SIMULATION/DETECTION OF
GROUND SITUATION (flows, speeds) VISSIM

Allocate Initial green green times and


Individual Junction

cycle times (C0)


Aggregate Data every three cycle durations

Analyse detector data (mean speed, flow),


Measure current green times, Cycle times

Check Flow conditions, PT demand, Breakdown,


Stationary Queues (Break Down, PT subroutine)
VAP
Analyse flow conditions from all junctions
Area Level

Calculation of Area level optimum C0 and


green time ( Webster & Cobbe)

Allocate offsets to prefixed movements


Output
Allocate green splits locally based on LOG
Each SCJ

ground data ( Webster & Cobbe)

Assign data back to controllers in VISSIM

Figure 3: Developed Signal Time Optimiser In VISSIM.

Association for European Transport 2003


Within VISSIM the network performance and traffic data was evaluated through
detector polls. Within the controller this data was aggregated and analysed every
five minutes and alternate stage green times and splits were calculated. The above
was limited to a maximum relative change of 10 seconds or 10%, whatever was the
lower, to inhibit sudden fluctuations in signal working. Whenever public transport or
pedestrian demands were detected, the controller switched to priority mode,
immediately jumping to and satisfying the priority stage.

7.0 COMPARISON OF OPTION TEST MODELS

Comparative analysis of results indicates that VISSIM produces realistic estimates


of delays. Using the VISSIM based optimiser produces much lower delays.
Compared to the present base model that is correctly calibrated in VISSIM, LINSIG
and TRANSYT underestimate the future delays at intersections. Compared to
Option1 (where only one signal is optimised, Option 2 (where signal co-ordination
and optimisation is in place at all intersections in the corridor), yields better results.
(Figure 4).

Average delays per vehicle (seconds)


a t W a t e r l o o R o a d -R i n g R o a d i n t e r s e c t i o n i n t h e A M p e a k h o u r

Base VISSIM AM peak Option 1 VISSIM AM peak Option 1 LINSIG AM peak O p t i o n 1 T R A N S Y T A M p e a k Option 2 VISSIM AM peak

250

191
200
Delay (seconds)

150
106
98
100 88 84 82
76 76
72
67 69
59 57 56 59
50 50 51 51
44 44 46 45 45 4 4 44 45
39
50 38 36 32 35 35 38 36
24 2 4
18 16 17

0
N -E N -E N -S N-S N -SN-S
W W E-N E -N E-S W
E -SW E -S S -N W - E W -E W - N W -N

Figure 4: Comparative analysis of average delays.

Association for European Transport 2003


Compared to LINSIG and TRANSYT forecast travel times, VISSIM produces more
robust results. Using VISSIM one can accurately forecast the impact of priority
schemes. Figure 5 gives a comparative analysis of travel times across the network.
One can clearly see improved travel times for busses as well as cars with the
priority signal control that has been more accurately modelled in VISSIM
microscopic simulation.

11.26
14.076
Bus Inbound 11.45
12.17
16.56

6.76
7.864 VISSIM Option 2
Bus Outbound 7.93
VISSIM Option 1
6.19
9.83
Option 1 TRANSYT

6.57 Option 1 LINSIG


7.638
Car Inbound 7.48 VISSIM Base Year
7.26
1. The measure used to quantify travel time variability (TTV);1 2 and
.73
2. The value of TTV benefits.
3.27
3.408
Car Outbound 2.35
1.92
4.26

Journey Time (min)


0 2 4 6 8 10 12 14 16 18

Figure 5: Comparative analysis of travel times.

8.0 SUMMARY

Comparative results show that traditional, average flow rate based tools such as
TRANSYT and LINSIG have severe limitations in evaluating congested corridors.
Microsimulation, because of its dynamic and disaggregate nature, can produce a
much better fit to observed conditions. The VAP vehicle actuated programming
capability of VISSIM enables us to develop a signal optimiser that will both
calculate an optimum signal design and show its operation in reality. The
additional benefits of microsimulation are the ability to model new system

Association for European Transport 2003


components such as bus lanes, advance stop lines, puffins and toucans and
individual vehicle typologies and driver behaviour.

REFERENCES

Ahuja S., Kohli S.S., (2003), Optimising Traffic Signals in VISSIM -


A Comparison with LINSIG and TRANSYT Models, presented at 3rd VISSIM user
group meeting, London

Clegg, R.G., Clune, A.J.,MUSIC: Modelling signal setting policies to meet diverse
transport goals. York Network Control Group, Department of Mathematics,
University of York, Heslington, York, England, YO10 5DD.

Clegg, R.G., Clune, A.J.,The MUSIC project: Urban traffic control for traffic
demand management. York Network Control Group, Department of Mathematics,
University of York, Heslington, York, England, YO10 5DD.

Cornwell, P.R. & Luk, J.Y.K. & Negus, B.J. Tram priority in SCATS, Traffic
Engineering and Control, 1986, 11, 561-565.

Fellendorf, M. (1994). VISSIM: A microscopic simulation tool to evaluate actuated


signal control including bus priority, Proc. of 64th ITE annual meeting, Dallas,
October 1994.

Garbacz, R.B., Adaptive signal control: What to expect, I.T.S. Online.

Hoyer, R.; Fellendorf, M. (1997). Parametrization of microscopic Traffic Flow


Models through Image Processing, Proc. of 8th. IFAC Conference, June 1997

http://www.jctconsultancy.co.uk/jct/Software/linsat.htm (LINSIG Website)

http://www.trlsoftware.co.uk/ (TRL Software Website)

Jongenotter, E., Monsma, S., (2002). Green Logic and the benefits for traffic-
actualted control-the Dutch experience. In: Traffic Engineering and Control,
October 20002, Volume 43 No. 9, pp. 351-353.

LINSIG for Windiws, Software manual.

Niittymki J. (1998). Fuzzy Logic Application to Public Transport Priorities at


Signalized Intersections. Mathematics in Transport Planning and Control.
Pergamon,1998. pp. 47 57.

Niittymki J. and Pursula M. (1997). Signal-Group Control Using Fuzzy Logic.


Paper presented at the 9th Mini-EURO Conference, and accepted for publication
in the Journal of Fuzzy Sets in Traffic and Transport Systems, EURO Working
Group onTransportation. Budva, Yugoslavia. September 15 19, 1997. 17 p.

Association for European Transport 2003


Rahmat, R.A., Jumari, K., Hassan A., Basri, H., (2002), Optimising traffic control
for a congested intersection. In: Traffic Engineering and Control, October 2002,
Volume 43 No. 9, pp. 357-361.

Sane K. (1995). Public Transport Priorities in Traffic Signals. Theory and


Practice from the City of Helsinki. Paper presented in European Workshop (DG
XVIII), 16.3.1995, Barcelona, Spain. 9 p.

Smith, M.J., Vuren, T.V., Traffic equilibrium with responsive traffic control.
University of York, Heslington, York, England, YO10 5DD, Institute for
Transportation Studies, University of Leeds, Leeds LS2 9JT, U.K.

Webster, F.V., Cobbe, B.M., (1966). Road Research Technical Paper No. 56.
Her Majestys Stationery Office.

Wiedemann, R. (1991) Modelling of RTI-Elements on multi-lane roads. (ed


Comission of the European Community, DG XIII), pp 1007-1019 Advanced
Telematics in Road Transport, Elsevier, Amsterdam, 1991.

Wiedemann, R. (1974) Simulation des Verkehrsflusses Schriftenreihe des


Instituts fr Verkehrswesen, Heft 8, Universitt (TH) Karlsruhe, 1974.

ACKNOWLEDGEMENTS

The authors would like to thank Mr. Serbjeet Singh Kohli and Mr. Ackchai
Srikijpanichkul from Mott MacDonald for their help in data evaluation and building
the signal control logic within VISSIM and Dr. Paul Hoad for checking the
manuscript of the paper.

Association for European Transport 2003

Das könnte Ihnen auch gefallen