EFFICIENT MAC PROTOCOL FOR WIRELESS BODY AREA SENSOR NETWORKS

SEMINAR REPORT
Submitted by LADVINE D ALMEIDA in partial fulfillment for the award of the degree of

Bachelor of Technology
in ELECTRONICS AND BIOMEDICAL ENGINEERING of

COCHIN UNIVERSITY OF SCIENCE AND TECHNOLOGY

DEPARTMENT OF BIOMEDICAL ENGINEERING MODEL ENGINEERING COLLEGE COCHIN 682 021

2012

GOVERNMENT MODEL ENGINEERING COLLEGE THRIKKAKARA KOCHI

DEPARTMENT OF BIOMEDICAL ENGINEERING Cochin University of Science and Technology

BONAFIDE CERTIFICATE This is to certify that the Seminar entitled Submitted by is a bonafide work done by him/her under our supervision. Dr. Jessy John
HEAD OF THE DEPARTMENT

Mrs. Suja Markose

CO-ORDINATOR

Mrs. Sincy P S
GUIDE

ACKNOWLEDGEMENT
At this moment of accomplishment, I am presenting my work with great pleasure. I would like to express my sincere gratitude to all those who helped me in the successful completion of my seminar. First of all, I would like to thank our Principal Prof. Dr. V P Devasia, who provided us with all facilities and support for the development of my work. I would like to thank Dr. Jessy John, Head of Department of Electronics and Biomedical Engineering for helping me in the successful accomplishment of this seminar. I would also like to thank my guide Mrs. Sincy P S, Assistant professor who gave me constant guidance and support throughout this journey of turbulence. I thank my seminar coordinator, Mrs. Suja Markose, Faculty staff who gave timely and valuable suggestions. I also thank Ms. Honey Hycinth Thomas, Ms. Sruti S, Mrs. Sajitha S, Ms. Nisha Krishnan Assistant Professors of Electronics and Biomedical Engineering for their support and encouragement of seminar. Above all, I thank God almighty for constantly motivating me with his love, and giving me courage at each stride to step forward with confidence and self belief.

ABSTRACT

This Paper presents a novel energy- efficient MAC protocol for wireless body area sensor Networks (WBASN) focused towards pervasive healthcare applications. Network adopts master-slave architecture where body worn sensors send readings to central node. To reduce energy consumption, all the sensor nodes are in standby or sleep mode until the centrally assigned time slot. A standard Listen-before-transmit algorithm is used to avoid network collisions. Using single-hop communication and centrally controlled sleep/wakeup times leads to significant energy reductions for this application compared to more flexible network MAC protocols such as 802.11 or Zigbee. As duty cycle is reduced, the overall power consumption approaches the standby power. System implemented as a part of Sensium System-on-chip WBASN IC showing easiness in implementation.

ii

TABLE OF CONTENTS

SL NO.

TOPIC LIST OF FIGURES

PAGE NO. v

LIST OF TABLES

vi

LIST OF SYMBOLS AND ABBREVATIONS

vii

1. 2. 3.

INTRODUCTION MOTIVATION METHODOLOGY 3.1 ATTRIBUTES OF WBASN 3.2 DESIGN GOALS FOR EFFICIENT MAC 3.2.1 SCALABILITY 3.2.2 DELAY PREDICTABILITY 3.2.3 ADAPTABILITY 3.2.4 ENERGY EFFICIENCY 3.2.5 RELIABILITY

1 2

5 5 6 7 7

iii

3.3 MAC PROTOCOL DESIGN 3.3.1 NETWORK ARCHITECTURE 3.3.2 BASIC OPERATION A. LINK ESTABLISHMENT B. WAKEUP SERVICE C. ALARM 3.3.3 WAKEUP FALL BACK TIME 3.3.4 CROSS LAYER FUNCTIONALITY 3.4 MAC PROTOCOL IMPLEMENTATION 3.4.1 IMPLEMENTATION ON SENSIUM PLATFORM 3.4.2 MAC COMPLEXITY 3.4.3 SYSTEM POWER AND DUTY CYCLE ANALYSIS 3.4.4 MEASURED RESULTS 3.4.5 COMPARING WITH EXISTING SYSTEM 4. 5. 6. APPLICATIONS CONCLUSION FUTURE SCOPE 15 17 18 22 24 25 26 27 10 11 12 12 14 8

REFERENCES

28

iv

LIST OF FIGURES

FIGURE 3.1 FIGURE 3.2 FIGURE 3.3 FIGURE 3.4 FIGURE 3.5 FIGURE 3.6 FIGURE 3.7 FIGURE 3.8 FIGURE 3.9 FIGURE 3.10

PROPOSED MAC TOPOLOGY LINK ESTABLISHMENT WAKE UP SERVICE ALARM PROCESSING SENSIUM SYSTEM ON CHIP DUTY CYCLE VS SLEEP TIME DUTY CYCLE VS SYMBOL RATE TRANSMIT POWER PIE CHART APPLICATION DEMONSTRATION BOARD COMPARISION BAR CHART

9 11 11 12 17 21 21 22 23 24

LIST OF TABLES

TABLE 3.1 TABLE 3.2 TABLE 3.2

APPLICATION REQUIREMENTS POWER CONSUMPTION FOR SENSIUM POWER MANAGEMENT UNITS

19 23 23

vi

LIST OF SYMBOLS AND ABBREVIATIONS

m - Micrometre
WBAN - Wireless Body Area Networks WSN - Wireless Sensor Networks MAC - Medium Access Control CCA - Clear Channel Assessment ISM - Industrial Scientific and Medical Band

vii

1. INTRODUCTION
The wireless communications revolution which is leading the convergence of all media and data services appears to be gaining wide acceptance. The healthcare sector is becoming increasingly interested in using this new technology to more effectively administer healthcare delivery. In particular, wireless vital signs monitoring is an area of modern healthcare that is growing very fast. This is due to its potential for slowing down the unsustainable growth of healthcare spending due to an increasing number of people living for years or even decades with chronic conditions that require on-going clinical management [1], [2]. Evolution of wireless, medical and computer networking technology has merged into an emerging horizon of science and technology called Wireless Body Area Networks (WBANs). Miniaturization and connectivity are notable parameters of this eld. Vital signs monitoring using wireless sensor network technologies have previously been described, but these systems are typically bulky and power hungry and rely on MAC protocols such as Bluetooth and 802.11 which are inefficient for such WBASN applications [3], [5].More general Wireless Sensor Network (WSN) MAC protocols, which have been the focus of fairly intensive research, are also not well suited to these specific biomedical WBASN applications either. Zigbee/IEEE 802.15.4[5] which is designed for similar networks does not have sufficient network device flexibility in non-beacon mode. It also lacks the cross-layer optimization features which the proposed protocol brings to this particular area. This paper describes a novel MAC Protocol designed specifically for wireless body area sensor networks focused on pervasive healthcare applications. Like other wireless sensor network MAC protocols, a primary design goal was low power consumption. This is achieved through a focus on collision avoidance (a primary source of energy wastage, and the use of centrally controlled time slotting for sensor nodes. The complete hardware MAC also incorporates cross-layer optimization, performing some ISO/OSI upper layer functions (from session layer down to PHY) at the hardware MAC layer to reduce the power overhead of software implementations. As a result of the network topology adopted in the MAC protocol, many of the traditional problems that plague wireless sensor networks have been either eliminated or significantly reduced.

2. MOTIVATION
MAC Protocol design is a very broad research area, and a lot of recent work has focused on the area of wireless sensor networks [1], [2], [3], [5]. Coming along with the urgent development of wireless technology, wireless devices have invaded the medical area with a wide range of capability. Not only improving the quality of life of patients and doctorpatient efficiency, wireless technology enables clinicians to monitor patients remotely and give them timely health information, reminders, and support potentially extending the reach of health care by making it available anywhere, anytime. As widely reported major causes of energy wastage in wireless sensor networks are collisions, idle listening, overhearing, traffic fluctuations and protocol overhead. In the more specific area of wireless body area networks, the first three sources of wastage can be eliminated by using master-slave architecture with time division multiple access with clear channel assessment (TDMA/CCA) network access scheme. The main goal of the proposed MAC Protocol is to reduce power consumption from sources like idle listening, overhearing and collision. In the more specific area of wireless body area networks, the first three sources of wastage can be eliminated by using master-slave architecture with time division multiple access with clear channel assessment (TDMA/CCA) network access scheme. MAC protocol proposed by Lamprinos [2] imposes a limitation on the duty cycles of the slaves on the network that some would have low duty cycle because they are serviced first while others would have a higher duty cycle since they are serviced later in the Rx slot. The limitation of energy supply on-board the sensor nodes, has motivated a lot of the research on sensor networks. Such a research can be classified into two general categories addressing the main causes of energy consumption; signal processing and radio communication. The first category of work is dedicated to extending the life of the network through the selective engagement of a subset of the sensors in monitoring the environment. Unselected sensor can switch to a low-power sleep mode. The goal of this work is to maintain enough live sensors to have organization. Sensor organization often involves signal processing techniques. The second category includes research on energy efficient radio communication. The network and link layers have received

the most notable attention with the bulk of the work focusing on energy-awareness and minimization through clever route setup. Energy-efficient link layer protocols tackle the energy wastage due to collisions among the radio transmission of nodes, keeping the receiver unnecessarily active and the excessive state changes of the radio circuit. We are hereby proposing a new MAC protocol which overcomes all the limitations discussed earlier. Specifically, idle listening and over-hearing are not an issue in this protocol as traffic is managed centrally. In the following sections, the proposed MAC Protocol is presented in more detail, from conception to design, implementation together with measured results.

3. METHODOLOGY
3.1 Attributes of Wireless Body Area Sensor Networks The main goal of the proposed MAC Protocol is to reduce power consumption from sources like idle listening, overhearing and collision. The closest existing MAC Protocol to the one presented is IEEE 802.15.4 [6]; however it had 3 differences which were not well suited to this specific application. 1) Data reliability isnt handled in the MAC layer. 2) Multiple communication modes increase the complexity of implementation. Hence, this new scheme is easily implemented in hardware. 3) Time-slotting is limited (16 slots in a super frame) and must all be equally spaced Before describing the MAC Protocol, assumptions about wireless body area networks are outlined. In specifying this MAC Protocol, the following attributes can be inferred about the wireless body area sensor network. 1) All wireless sensor nodes are attached to the body. 2) The data being monitored is of low frequency, thus the network does not need to respond immediately to changes. 3) Sensors monitor a range of vital signs which are typically at a low data rate kB e.g., temperature, pressure or heart-rate reading. However some higher data rate applications must also be catered for, such as streaming of electrocardiogram (ECG) signals. 4) The nodes are miniature; battery powered and need to run ideally for days from very low capacity batteries such as flexible printed battery technologies or miniature coin cells. 5) Sensor nodes are resource constrained, i.e., they have low processing power and limited memory. 6) Data from the wireless sensor nodes is forwarded to a central master node for processing; this central node is significantly less resource and power constrained relative to the wireless sensor nodes. These listed attributes are the main influences leading to the specific MAC Protocol implementation described in this paper. These attributes also differentiate the particular application from more generic wireless sensor network protocols, and other protocols which have been deployed in biomedical applications such as Bluetooth and Zigbee.

3.2 Design Goals for Efficient MAC Protocols In this subsection we briefly discuss the main design goals for the MAC protocols of sensor networks. As will become clear, some of these goals may be conflicting and may force a trade-off. 3.2.1 Scalability It is envisioned that most applications of unattended sensor networks will involve large number of nodes. Therefore, scalability of the employed protocols is crucial. The resources, i.e. time and bandwidth, sharing method and the arbitration strategy have to allow for fair access to the medium and to prevent excessive collisions. In addition, the potential for large set of communicating nodes would impose a restriction on the use of some MAC schemes such as CDMA. In almost all sensor networks, nodes relay other sensors data and can even perform data aggregation. Pursuing a pure CDMA scheme would require a sensor to store many code sequences, which may be impractical for tiny sensor devices with very limited computational resources. It is worth noting that the scalability of the link layer protocols is influenced by the network architecture and routing methodology. Hierarchical network structures can allow the employment of multiple resource sharing strategies and shape the network flow into patterns that can be exploited at the link layer. For example, grouping sensors into disjoint clusters allows designating nonoverlapping frequency bands to clusters, similar to FDMA, and applying a TDMA or CSMA schemes for intra-cluster communication among sensors. In addition, the methodology for route setup can rule out some MAC schemes. For example, flooding-style data dissemination makes time-based medium arbitration strategy impractical. 3.2.2 Delay-predictability A number of applications of sensor networks such as target tracking require delaybounded delivery of data. Ensuing timeliness of data reception is typically handled at multiple layers in the communication stack. For example, special consideration at the network layer would alleviate the burden of long queuing time that affect the overall end-to-end delay. However, the link-layer would play a major role through careful packet scheduling and predictable strategy for medium arbitration.

The employed MAC scheme determines the schedule for packet transmission not only for the individual node but for the entire network. At the node level, a suitable packet classification and priority mechanism is the base for a service differentiation that allows delay centric handling of out-going packets. On the network level, a well-defined and easily enforced strategy is needed to prevent inter-node competition for medium access from causing contention that makes the time for packet transmission and reception nondeterministic. For example, collision based medium arbitration mechanisms such as CSMA would not be appropriate for large and densely populated sensor networks since it is not known how many times a node will back off until it successfully transmit. On the other, reservation based approaches such as TDMA would be a good match despite their scalability problems. 3.2.3 Adaptability In most applications of sensor networks traffic density varies significantly over time and from part of the network to another. Such observation is valid for both event-triggered and query-based models of network operation. For example in an ECG monitoring setup only periodic status updates are sent in normal conditions while many sensor reports are generated in case of detecting an arrhythmia. In a typical query-based operation, sensors transmit only in response to requests and little traffic is generated otherwise. In addition, generated data can be subject to aggregation on route to the sink. Data aggregation can take the form of averaging the reported data, picking the maximum value, removal of redundant report, etc. In some cases the traffic pattern will not change if the aggregating nodes are fixed at the time of route setup and the transformation of the data is many-to-one, e.g. averaging multiple readings. However, data aggregation is mostly performed when applicable and thus can cause variability in the traffic flow. For example, elimination of redundant sensor readings filters out repetitions and prevents resource wastage. Needless to say that dropping the packets of unneeded data depends on the type of sensors and the detected events. The MAC scheme should adapt to such high fluctuation in traffic and should allow medium access rescheduling to efficiently handle burst high-priority traffic.

3.2.4 Energy-efficiency Energy is a scarce resource for sensor networks. As explained earlier, medium access is a major consumer of sensor energy, especially for long-range transmission and when the radio receiver is kept on all the time. The output power of the radio transmitter is directly proportional to distance squared and can significantly magnify in a noisy environment. Energy-aware routing typically pursues multi-hop paths in order to optimize the transmission energy. On the other hand, energy-conscious medium access control (MAC) can save transmission and reception energy by limiting the potential for collisions, minimizing the use of control messages, utilizing most of the available frequency band to shorten the transmission time, turning the radio into low power sleep mode when it is idle and finally, avoiding the excessive transitions among active and sleep states. 3.2.5 Reliability Reliable delivery of data is a classical design goal for all network infrastructures. Guaranteed packet delivery is ensured by the careful selection of error free links, avoidance of overloaded nodes, and the detection and the recovery from packet drops. There is usually a trade-off between the control traffic overhead and the level of reliability. For example, acknowledging each packet minimizes the recovery time and limits its scope, at the price of a high control traffic that can lower the effective link bandwidth, boost end-to-end delay and increase energy consumption. In wireless networks, packet drops are mainly caused by buffer overflow and signal interference. Avoiding buffer overflow is the responsibility of both the routing and the MAC protocols. Balancing load among available routes would reduce the potential for reaching the maximum capacity of the in-bound traffic buffer in relay nodes. Meanwhile, the employed medium arbitration scheme determines the buffer management strategy and has to ensure a service rate for the outbound flow that is high enough to stop the number of backlogged packets from exceeding the maximum buffer size. Packet drops due to signal interference can be minimized through the use of sufficiently high transmission power and the prevention of contention for medium access among nodes.

3.3 MAC Protocol Design In the seven-layer OSI model of computer networking, media access control (MAC) communication protocol is a sub layer of the data link layer, which itself is layer 2. The MAC sub layer provides addressing and channel access control mechanisms that make it possible for several terminals or network nodes to communicate within a multiple access network that incorporates a shared medium, e.g. Ethernet. The hardware that implements the MAC is referred to as a medium access controller. The MAC sub layer acts as an interface between the logical link control (LLC) sub layer and the network's physical layer. The MAC layer emulates a full-duplex logical communication channel in a multi-point network. This channel may provide unicast, multicast or broadcast communication service. Much efficiency in data communication can be achieved through proper design of MAC layer as it comes as highest layer in OSI Model of network Architecture. The various characteristics of network architecture include types of interconnection between nodes, operation modes, cross-layer functionality etc. 3.3.1 Network Architecture Network architecture is the design of a communications network. It is a framework for the specification of a network's physical components and their functional organization and configuration, its operational principles and procedures, as well as data formats used in its operation. The Open Systems Interconnection model (OSI model) is a product of the Open Systems Interconnection effort at the International Organization for Network Standardization. It is a way of sub-dividing a communications system into smaller parts called layers. A layer is a collection of similar functions that provide services to the layer above it and receives services from the layer below it. On each layer, an instance provides services to the instances at the layer above and requests service from the layer below. The different layers that comes under OSI Model includes Physical Layer, Data Link Layer, Network Layer, Transport layer, Session layer, Presentation layer and Application layer. In our Work, as a result of the attributes in the previous section, a point to multi-point (star) network architecture is proposed. In this architecture, the central node acts as the master while the other nodes are slaves. The slave nodes are the actual WBASN nodes which acquire sensor data and transmit

to the central node for processing. Each individual master-slaves network is referred to as a cluster. For ease of management, the maximum number of slaves connected to a master in one cluster is 8 (many more can be connected, but the time-slotting would have to be managed outside the protocol). Also in this architecture, the network access is clear channel assessment [3] and collision avoidance with time division multiplexing (CCA/TDMA). This network access scheme significantly reduces the likelihood of collision and idle listening, leading to significant power savings. In addition time-slot allocation is dynamically controlled by the master, so a slave time slot could be changed every time it communicates with the master. This enables the system to better cope with fluctuating traffic. The penalty is increased complexity of the central node. However, this is not a major problem because the central node is expected to have significantly more power and processing resources. The key idea used in this network architecture is to move much of the network and protocol complexity away from the power constrained wireless sensor nodes and into the much more capable central node. This network topology is shown in Fig. 3.1. To accommodate for intercommunication between clusters, access to an IP network may be used. This way complex network structures can still be built which extend wide areas.

Fig 3.1. Proposed MAC Protocol Network topology

3.3.2 Basic Operation The proposed MAC protocol operations are based on three main communication processes. The first is when a wireless sensor node wants to join a cluster. This is called the Link establishment process. The second is when a slave and master wake-up after an assigned sleep period. This is called the wakeup service process. The last process is an exception process which occurs when a slave urgently wants to send information to the cluster master. This is called an Alarm process. In all three processes, communication can only initiated by the master. In addition only one slave can join the network at a time as the network is non- ad hoc. A. Link Establishment When a master node is first enabled, it continuously tries to establish a link with unattached slave nodes. It does this by first scanning for a vacant RF channel. When it finds one, it remains on that channel and starts sending out a beacon containing a unique address and configuration for a slave and then listening for a fixed time for a response. The sum of the masters beacon transmit and listen time is termed Tseek. Alternatively, when a slave node is enabled, it also scans the available RF channels to find the master beacon. If a channel is vacant for Tseek, it hops to the next one. If it is occupied, it listens for fixed period 2 x Tseek for preamble from the master beacon and if it doesnt receive it, it moves on again to the next channel. Once the beacon is received, it responds with an acknowledgement to the master. The master node then assigns a sleep time and ends the transaction. At the end of the link establishment process, the slave has a unique address, configuration information and sleep time [Fig. 3.2]. Subsequent additions to the network have to be specifically initiated (e.g., by software) on the master. After Link establishment, the RF channel of communication is fixed and can only be changed by higher layer intervention.

10

Fig 3.2. Stages in the 3 processes - Link establishment.

B. Wakeup Service After link establishment, both master and slave sleep timers start to count up to the sleep time. Hence, they both wake-up at about the same time, the difference in wake-up times determined by the offsets between both timers and the length of the sleep time. On wake-up, the master interrogates the slave which alternatively listens. It (master) may simply request for its (slaves) sensor data, or request status information. Whatever the communication, a new sleep time is assigned to the slave, setting the next wakeup time-slot [Fig. 3.3]. To mitigate long-term time-slot drift between the master and slaves in a cluster, there is an optional synchronization phase during every communication when the slave can synchronize its timer to that of the master. The masters timers never change. This dynamic time-slotting does not in any way preclude the use of the protocol in a fixed time-slotting application. It just offers this added functionality which may be used if required.

Fig 3.3 Wakeup servicing.

11

C. Alarm If the slave detects an alarm condition while performing some local processing, it may communicate with the master without waiting for its next wake-up. This alarm condition may be due to an out of bounds measurement (of say body temperature or blood glucose level) or a sensor memory overflow alert. When this mode is enabled, the master continuously sends out a request to all the slave addresses on the network sequentially. The slave listens for its address and communicates the alarm condition when it receives it. This only happens when the master is not busy servicing a scheduled wakeup, and would be terminated when a slave wake-up needs to be serviced. [Fig. 3.4].

Fig 3.4 Alarm Processing.

3.3.3 Wakeup Fallback Time The energy waste caused by idle listening is reduced by sleep schedules. In addition to its implementation simplicity, time synchronization overhead may be prevented with sleep schedule announcements. In the preamble sampling technique, a preamble precedes each data packet for alerting the receiving node. All nodes in a network sample the medium with a common period, but their relative schedule offsets are independent. If a node finds the medium busy after it wakes up and samples the medium, it continues to listen until it receives a data packet or the medium becomes idle again. The size of the preamble is initially set to be equal to the sampling period.

12

However, the receiver may not be ready at the end of the preamble, due to reasons like interference, which causes the possibility of over emitting type energy waste. Moreover, over emitting is increased with the length of the preamble and the data packet, since no handshake is done with the intended receiver. To reduce the power consumption incurred by the predetermined fixed-length preamble, this MAC implementation offers a method to dynamically determine the length of the preamble. That method uses the knowledge of the sleep schedules of the transmitter nodes direct neighbours. The nodes learn and refresh their neighbours sleep schedule during every data exchange as part of the acknowledgement message. In that way, every node keeps a table of sleep schedules of its neighbours. Based on neighbours sleep schedule table, MAC layer schedules transmissions so that the destination nodes sampling time corresponds to the middle of the senders preamble. To decrease the possibility of collisions caused by that specific start time of wake-up preamble, a random wake-up preamble is advised. Another parameter affecting the choice of the wake-up preamble length is the potential clock drift between the source and the destination. A lower bound for the preamble length is calculated as the minimum of destinations sampling period, Tw, and the potential clock drift with the destination which is a multiple of the time since the last ACK packet arrival. Considering this lower bound, a preamble length, Tp, is chosen randomly. The central management of time slotting can be a complex task for the master especially when complicated by the occurrence of sporadic alarm conditions. To ensure that every sensor slave node maintains a guaranteed time slot [3] even if another slave flags an alarm condition, the novel concept of wakeup fallback time (WFT) is proposed. If a slave wakes up and fails to communicate with the master (either because it is busy servicing an alarm, or the channel is temporarily occupied by an interferer), it goes back to sleep with a sleep time set by the WFT. During this time it continues to buffer the sensor data. After the WFT, it wakes up and searches for the master again. Similarly, if the master is unable to communicate with the slave at the wakeup time, it also defaults to the WFT. Hence, both master and slave wakeup at the common WFT and communicate, restoring the schedule.

13

The WFT is a programmable parameter and is a fraction of the shortest sleep time on the network to mitigate continuous time-slot collisions. Also it is global to the network and originally set by the master during the link establishment process. This scheme ensures that time slot overlaps are seamlessly managed and do not degrade the network in the long run. Also it allows a slave with a long sleep time more opportunities to communicate its data to the master without having to wait for the whole sleep-time again. 3.3.4 Cross Layer Functionality When a data packet transmission fails, the MAC automatically retries a programmable number of times before dropping the packet. In addition large packets can be automatically broken in to smaller frames and transmitted one at a time. The protocol also provides for the receiver to reassemble the fragmented data packets as they are received. One additional function provided is the control of the frequency and rate of sensor data acquisition depending on the application. These functions are usually handled by higher layers in the ISO/OSI protocol stacks. The cross layer approach transports feedback dynamically via the layer boundaries to enable the compensation for e.g. overload, latency or other mismatch of requirements and resources by any control input to another layer but that layer directly affected by the detected deficiency [1],[2]. In the original OSI networking model, strict boundaries between layers are enforced, where data are kept strictly within a given layer. Cross-layer optimization removes such strict boundaries to allow communication between layers by permitting one layer to access the data of another layer to exchange information and enable interaction. For example, having knowledge of the current physical state will help a channel allocation scheme or automatic repeat request (ARQ) strategy at the MAC layer in optimizing trade-offs and achieving throughput maximization. Cross-layer optimization shall contribute to an improvement of quality of services under various operational conditions. Such adaptive quality of service management is currently subject of various patent applications. The cross-layer control mechanism provides a feedback on concurrent quality information for the adaptive setting of control parameters. The quality aspect is not the only approach to tailor the cross-layer optimization strategy. The control adjusted to availability of limited resources is the first mandatory step to achieve at least a minimum level of quality.

14

Communication systems that need to operate over media with non-stationary background noise may benefit from having a close coordination between the MAC layer (which is responsible for scheduling transmissions) and the PHY layer (which manages actual transmission and reception of data over the media). In some communications channels (for example, in power lines), noise may be nonstationary and might vary synchronously with the 50 or 60 Hz AC current cycle. In scenarios like this, overall system performance can be improved if the MAC can get information from the PHY regarding when and how the noise level is changing, so that the MAC can schedule transmission during the periods of time in which noise levels are lower. In this protocol, hardware implementation directly at the MAC layer is preferred as significant power savings over software implementations is achieved. This is because the processor would normally need to run continuously (significantly increasing standby power) to perform these functions like determining when to take the next sensor reading, how many should be taken and when to switch to another sensor. Also the delay involved in communicating through the protocol stack layers is eliminated [5].

3.4 Mac Protocol Implementation 3.4.1 Implementation on Sensium Platform Wireless sensor nodes gather, store and locally process vital signs data, before transmission to a central base-station node. Although prototype modules for such WBSN applications are becoming available, these devices tend to be multi-chip solutions manufactured from off-theshelf components, and suffer from excessive power consumption and relatively large form factors [1], [2]. Improvements to the patients quality of care can be achieved through miniaturization and a reduction in power consumption. These objectives dictate the development of a custom system-on-chip (SoC). Although ultra-low-power wireless transceiver ASICs have previously been reported [3-5], this paper describes the integration of a system solution with a full-custom hardware MAC, digital microprocessor core and I/O peripherals, on-chip memory, micro power ADC, wireless transceiver and custom sensor interfaces. This SoC platform device is capable of achieving ubiquitous medical monitoring when interfaced to

15

appropriate body-worn sensors, and represents the state of the art in terms of functionality and ultra-low-power consumption. The encapsulated wireless sensor node is in the form of a thin and flexible patch, comprising sensors, SoC, battery and antenna as shown in Figure 3.5. The patch is attached to the patient for a period of typically four to seven days, after which it is thrown away and a new patch attached, if necessary. The battery is manufactured from environmentally friendly materials such that it can be safely disposed of or recycled. It provides typically 3mAh/cm2 at 1.4V, dropping to 0.9V at end of battery life. The limited energy capacity means that the average current drain must be of the order of micro amps to achieve the target operating lifetime. In addition, the battery peak currents must be limited to be no higher than a few milliamps to avoid battery collapse. These energy constraints require a novel low-power design methodology to be applied at all levelsnetwork protocol, system architecture, circuit topology and implementationin order to guarantee reliable and robust operation within the batterys maximum peak current discharge capacity. The MAC Protocol was implemented as a key part of a custom system-on-chip (SoC) ASIC for biomedical WBASN applications. This mixed-signal SoC, known as Sensium, integrates a half-duplex transceiver, programmable sensor interface circuitry and a digital block containing the hardware MAC plus a low power 8051 microcontroller integrated with 32 kB of code and 32 kB of data memory. The data memory is directly accessible via a DMA controller by both the Sensor Interface ADC (to write sensor readings) and by the hardware MAC (to read/write sensor readings for direct transmission/reception). Having direct access to system memory allows the slave devices to operate entirely without processor intervention. The processor can therefore be switched to a low clock frequency and used to service irregular events like link errors. On the master, processor intervention is also minimal, and so it is freed up to handle higher layer functions or transferring acquired data to a PC for further processing. Which blocks are active in a given mode is controlled by the power management unit. The Sensium system block diagram is shown below in Fig. 3.5. The physical layer for the radio operates in the 870/900 MHz SRD/ISM bands, employing FSK modulation with 50 KHz deviation to give an over air bit rate of 50 kbps. The sensor interface block features

16

sensor driving and interface circuitry for a range of biomedical sensors, and includes a 10bit, 50500 Hz sampling rate DSM-ADC. For error control, a hamming code is implemented in the MAC hardware together with CRC frame checking. This provides 2 levels of error correction and detection.

Fig.3.5 Sensium System on chip block diagram.

The Sensium SoC is implemented in a 0.13 um CMOS technology and occupies an area of 16 mm2. Full functionality for centre-processed samples has been verified down to 0.85V; initial yield across corner lots is greater than 95% at a test time of <3s on a Teradyne J750. In WBSN applications this SoC is able to provide typically one to two orders of magnitude lower power consumption than competing solutions, and thus offers the possibility for truly unobtrusive and disposable vital-sign monitoring. 3.4.2 Mac Complexity When MAC protocol is implemented in hardware, the processor needs to execute many instructions in sequential order, the complexity of instructions determined by different protocols in use. A more generalized MAC protocol will have long sequence of instructions. A processor that executes every instruction one after the other (i.e. a non-pipelined scalar architecture) may use processor resources inefficiently, potentially leading to poor performance. The performance can be improved by executing different sub-steps of

17

sequential instructions simultaneously (this is pipelining), or even executing multiple instructions entirely simultaneously as in superscalar architectures. Further improvement can be achieved by executing instructions in an order different from the order they appear in the program; this is called out-of-order execution. As often implemented, these three techniques all come at a cost: increased hardware complexity. Before executing any operations in parallel, the processor must verify that the instructions do not have interdependencies, for example, a first instruction's result is used as a second instruction's input. Clearly, they cannot execute at the same time, and the second instruction can't be executed before the first. Modern out-of-order processors have increased the hardware resources which do the scheduling of instructions and determining of interdependencies. In our work, the entire Hardware MAC protocol, including the error control and framing block, was under 12 K-gates for the slave and ~25K-gates for the master. The gate count of the hardware implementations points to the simplicity of the protocol. Since no hardware implementations of 802.15.4 were found, we should compare the software implementations of both protocols. The proposed protocol can be implemented in around 16 kB of code (including application code) while 15.4 would require at least 32 kB. The power consumption for this implementation is around 500 W while it is 15.4 is 10 mW. This is because of the difference in clock frequency required to run both protocols. 3.4.3 System Power and Duty Cycle Analysis The average power consumption is dependent on the duty cycle of operation. So even though a sensor node has a very long sleep time, but also has a long active time, the duty cycle would be high and hence average power. This can be computed for spot measurement applications like for temperature and glucose and also for continuous monitoring applications like ECG. Table 3.1 compares common applications. A detailed analysis of the relationships between the parameters that affect duty cycle and average power computation follows. From the above analysis, it has been shown that the duty cycle in continuous monitoring applications like ECG is affected mainly by the communication symbol rate.

18

TABLE 3.1 Typical Application Requirements

Table 3.1 illustrates this using typical numbers for 3 important applications. For spot measurement applications, we can reduce duty cycle by increasing the sleep time because more payload data means that the overhead time becomes less. This is however not the case for continuous monitoring applications like ECG as the amount of sensor data must increase with sleep time. For applications like this, the sleep time is usually limited by the system memory resources available for storing the sensor data. The payload size was kept fixed; while the sleep time was changed. It can be concluded that the power is dependent on the sleep time as well as the number of retransmissions. Also, the power consumption approaches the standby power as sleep time increases. PAVE = average power PA = active power PSB = standby power (sensors acquiring data) TS = allocated sleep time (time between wakeups) NR = number of retransmissions TFO = Frame overhead time (RF Setup, preamble, sync, CRC)

19

FADC = ADC sampling frequency NS = Number of samples taken in a sleep period Es = Effective sampling rate = NS/TS Nbps = Number of bits per sample FSYM = TX/ (RX bits/sec) RECC = Error Overhead Ratio ~ 1.5 DC = Duty Cycle = TA/TSB (1+NR) The general equation for average power is PAVE = PA*DC PSB * (1-DC) Expading the DC equation further, we have, DC = (TC+ 3/2 * N2/T2 * (Nbps*TA)/FSYM ) * (1+NR) TS = 2*TFO*FSYM + 3*NS*Nbps * (1+NR) 2* TS * FSYM For spot measurement applications, TFO is significant because of the small data payload and hence cannot be ignored. However for continuous monitoring applications like ECG, where the payload bits are >> frame overhead bits, NS = FADC * TS and so TFO becomes insignificant. Equation (3) then becomes, (3) (2) (1)

DC =

3 FADC * Nbps (1+NR) 2 FSYM

(4)

The transmit time for the data payload is 40 ms (100 samples), giving a typical duty cycle (for 1 sec sleep time) of 4%. The majority of target applications however have much

20

longer sleep times, so the duty cycle would be much smaller and hence lead to greater power savings. A more realistic plot is shown below in Fig. 3.6. The plot shows that as sleep time increases, the duty cycle decreases, quickly converging even for 9 retries. In the case of an ECG streaming application, the duty cycle is fixed by the Transmit/Receive symbol rate. Fig. 3.7. is a plot of duty cycle versus symbol rate for this implementation. In all applications, the duty cycle would determine the time slot allocations to the slave devices in a cluster network and ultimately limits how many can be supported. Hence, network scalability is mainly application dependent. Also an application like blood glucose monitoring (0.0014% duty cycle) could have up 255 slave nodes which is the maximum number that can be supported by the master node.

Fig. 3.6 Duty Cycle versus Sleep time

Fig. 3.7 Duty Cycle versus symbol rate

21

3.4.4 Measured Results The fabricated chip was mounted on a demo board with other interfaces for SPI, UART and USB [4], [5] as well as a bread boarding area for connecting the application sensors. The constructed demo board is shown below in Fig. 3.9. Table 3.2 below gives the component and system standby and active power with a 1 V supply. These are actual measured current consumptions from the fabricated PCB including the sensor currents for body temperature sensing and ECG streaming applications. As shown in Table 3.2, there are 3 power states; active, sleep/standby and deep sleep. In active mode, all the blocks are turned on. For sleep mode, the 16 MHz clock is turned off, but the sensor interface remains on. The sensor interface and 16 MHz clock are turned off in deep sleep which is the lowest power mode. In all the modes, the MAC timers remain active as they control when to enter or exit the different power modes. These are run off a separate 32 kHz XTAL. So the power consumption of the digital block is significantly reduced. On wake-up, the clocks are turned on again. Table 3.3 illustrates these power modes and the states of all the blocks. The relative power contributions are illustrated in Fig. 3.8 below.

Fig 3.8. Transmit power (~10 dBm) pie chart.

22

Fig 3.9. Application demonstration board photo

TABLE 3.2 Power Consumption for SENSIUM SOC

TABLE 3.2 Power Management Unit Modes

23

One of the observations from measurements is that on average, the number of retries is very low, but ultimately depends on the agility of the radio when there is relative movement between the communicating nodes. In addition, the overall measured packet error rate is 0.04%. This is however detectable using CRC and so the data can be retransmitted at a later communication. All performance measurements used a dipole antenna with both master and slave nodes stationary. Also the separation distance was 5 meters and RF transmit power was ~10 dBm. 3.4.5 Comparing With Existing Systems The observations about traffic in sensor networks impact the design of the MAC protocol as well as the network and transport layers. Instead of optimizing for high throughput, low latency, and fairness, MAC protocols for sensor networks must first and foremost be energy efficient. Consequently, they should be optimized for the case that there is little or no network traffic. Classical MAC protocols for WLANs like 802.11waste a lot of energy by so-called idle listening, that is, listening to receive messages that are never sent. The power consumption of this work compared with other systems is bar charted in Fig. 3.10. The comparison is based on extensive simulation driven by traffic that varies over time and location; sensor nodes are inactive unless they observe some physical event, or send status updates to the sink node providing the connection to the wired world. One of the key differences that comes out of this is that the RF power requirement is significantly low for this work. This makes it possible for much smaller batteries like flexible-thin or zinc-air which cannot be used for any of the other standards. It is concluded that power is the penalty these protocols pay for their generality.

Fig 3.10. Bar chart comparing battery power requirements for different wireless standards

24

4. APPLICATIONS
The low-power protocol proposed here provides a saving in transmission power, which can then be used to carry out some supplemental signal processing at the sensors in a WBAN. This allows near real-time monitoring of patient condition and can reduce the work load on the medical staff by providing only the information of importance, while keeping power usage to a minimum. Low power usage means that small devices equipped with small and light batteries can still have acceptable lifetime. Given the low power usage of the proposed protocol, battery power is now available to implement signal processing algorithms, whose power usage is in the order of a few milliwatts. This means that for the same power as using Zigbee or similar protocols, we can both transmit and process data. Furthermore, if the results of the signal processing in central node are exploited fully, the amount of data that needs to be transmitted can be greatly reduced, thus lowering power usage still further. Therefore, the power saved by transmitting using the proposed protocol can be used to provide clinically meaningful information and greatly reduce the overall power usage of the device.

25

5. CONCLUSION
This paper presents a new energy-efficient MAC Protocol targeted at wireless body area sensor networks focused on pervasive healthcare applications. The protocol exploits the attributes of this type network to implement a very low power architecture which is still capable of fast reaction to sporadic Alarm events. Reducing the power requirements for the communication part of the system allows allocation of more energy to more accurate DSP for different applications. The proposed scheme also results in very reliable data transfer, which is crucial in medical applications. The novel concept of wakeup fallback time is also presented as a means of reducing the complexity of time-slot management in the presence of link failures resulting from Alarm events or other interference. The MAC has been implemented as part of a larger SoC (Sensium), and measured results have validated the effective operation of the new MAC protocol. The New Protocol was also compared with three other popular protocols, and results showed improvements in communication power consumption and duty cycle.

26

6. FUTURE SCOPE Future Scope of work includes incorporating more efficient forward error correction codes to improve reliability of data transfer, as well as research into the best strategies for sensor to central node communication. The central node can be made more intelligent by incorporating more capable signal processors to analyse data. If the need for the more exible network should arise, various solutions for network forming and management should be investigated.

27

REFERENCES
[1] W. Ye, J. Heidemann, and D. Estrin, An energy-efficient MAC protocol for wireless sensor networks, in Proc. IEEE 21st Ann. Joint Conf. IEEE Comput. Commun. Soc., 2002, vol. 3. [2] I. E. Lamprinos, A. Prentza, E. Sakka, and D. Koutsouris, Energy efficient MAC protocol for patient personal area networks, IEEE Eng. Med. Biology Soc., vol. 2005, pp. 37993802. [3] N. Chevrollier and N. Golmie, On the use of wireless network technologies in healthcare environments, in Proc. 5th IEEE ASWN, France, Jun. 2005, pp. 147152. [4] E. Jovanov, A. Milenkovi, C. Otto, P. D. Groen, B. Johnson, S.Warren, and G. Taibi, A wban system for ambulatory monitoring of physical activity and health status: Applications and challenges, in Proc. 2005 IEEE Eng. Med. Biol. 27th Ann. Conf., pp. 38103813. [5] T. v. Dam and K. Langendoen, An adaptive energy-efficient MAC protocol for wireless sensor networks, in Proc. 1st Int. Conf. Embedded Netw. Sens. Syst., 2003, pp. 171180.

28