You are on page 1of 13

Dynamic Congestion Pricing using Wireless Sensor Networks Ashok Kumar Ashapreethi, Leong Qian Wei, Leong Zhi Yi

Abstract Congestion pricing as a means of traffic management is aimed at curbing urban ro ad demand. In the case of Singapore, it has been successful in reducing commutin g time and number of vehicles on road. From an economic standpoint, the rational e behind congestion pricing is to reflect the societal costs of road usage to th e user in the form of levies. These levies aim to manage demand by discouraging motorists from using the road during peak hours. Static congestion pricing models are time or geographic based and are inflexible to current traffic conditions. Dynamic congestion pricing on the other hand adj usts for different congestion patterns and more appropriately addresses the prob lem of congestion. This report examines the use of a wireless sensor network (WS N) to meet the needs of a dynamic congestion pricing model. The first part of th e report introduces our idea for flexible congestion pricing, the second part de scribes in detail how our idea works and the third, fourth and fifth part descri bes how we intend to implement this system. Finally, we conclude the report by e xploring how our idea can be better achieved using other technologies and how we can implement this elsewhere. TABLE OF CONTENTS Abstract i 1 Introduction 1.1 Overview of congestion pricing 1.1.1 Static pricing 1 1.1.2 Dynamic pricing 1 1.2 Overview of wireless sensing 2 1.3 Proposed idea 2 2 Theory of Operation 2.1 Detection methods 2.1.1 Infrared 3 2.1.2 Ultrasound 3 2.1.3 Video/ image processing 3 2.1.4 Millimetre wave radar 4 2.1.5 GPS 4 2.2 Communication Network 5 2.3 Dynamic congestion pricing 5 3 Technical Specifications 3.1 Hardware Specifications 3.1.1 Power supply 7 3.1.2 Microprocessor 8 3.1.3 Transceiver 9 3.1.4 Sensor 10 3.1.5 Enclosure selection 10 3.2 Software Specifications 3.2.1 System architecture 11 3.2.2 Programming 12 3.2.3 Memory 13 3.2.4 Communication protocol 13 3.2.5 Power 13 3.2.6 Suggested operating system - Nano RK 4 Modification of Existing In-Vehicle Units 15 5 Future developments 16 6 References 17

14

1. INTRODUCTION 1.1 Overview of Congestion Pricing The main objective of congestion pricing is to promote a more rational use of ro ad resources given the limited network capacity. This is achieved by persuading road users to modify their travelling decisions. Positive responses from road us ers include altering their travelling routes, changing departure time or switchi ng to public transport. In urban cities where vehicle density is high and road capacity is low, congesti on pricing has been used as an effective strategy to manage traffic. Congestion pricing works by charging motorists for the negative externalities they create a nd this reduces demand for road resources. This has been applied with success in major cities around the world such as London, Tokyo, Milan and Singapore. For e xample, Singapore employs an electronic road pricing (ERP) scheme that levies mo torists automatically when they pass by gantries that are located across the cit y state. Technically speaking, congestion pricing consists of both parking prici ng and road pricing but for the purposes of this report, we shall focus on road pricing. One of the reasons why congestion pricing is popular is because it manages conge stion without having to increase road resources. It also reduces greenhouse gas emissions and overall carbon footprint, making it a viable option for a greener and more sustainable land transport system. 1.1.1 Static pricing Static congestion pricing refers to systems of road pricing that charge a fixed levy based on periods of peak hours or when motorists enter particular zones in the city. These levies are expected not to vary with time and are generally impo sed as a flat rate. One such example is the ERP system in Singapore. The nature of these static pricing systems makes it predictable and that can cau se more problems. Motorists who choose to avoid these systems inevitably cause d iversion of traffic elsewhere which can lead to another congested scenario. Simi larly, time based pricing systems can cause unnecessary financial burdens to roa d users when there is no congestion. 1.1.2 Dynamic pricing Dynamic pricing systems on the other hand consider the variable nature of road d emand and impose a cost that is proportional to road conditions. These road cond itions include the time it takes to travel a particular stretch of road, the veh icle density on road, the average speed of vehicles or the number of links the r oad makes. To sum up, the costs of using the road should reflect the severity of congestion in real-time. Additional research has shown that the rate of allevia ting congestion also contributes to the external costs of road usage. Dynamic congestion pricing is not constrained by space or time. One key point th at is highlighted in associated literature is the concept that the amount of lev y should be equal to the marginal costs of using the road. Static pricing system s are unable to capture this key concept and are less effective. There are several challenges to design and implement this system. For instance, a proper communication network is required to relay the varying toll rates to mo torists at proper intervals of time. These challenges are addressed in the follo wing sections below. 1.2 Overview of Wireless Sensing Wireless sensing consists of a complicated network of sensor nodes that provide cumulative information. Each sensor node is able to collect data on speed or pos ition and communicate with other nodes to form a collective network that has far greater computational power than a single node. A sensor node typically consists of sensors, a power source, a transmitter and r eceiver, a microprocessor and its enclosure. When it is inactive, it conserves e

nergy by switching off higher functions and enters idle or sleep mode. When activate d, a sensor node collects data, does some minor processing and passes it along t he network until it reaches an access point where it can transfer this data to e nd users. Types of sensing systems include vision-based detection systems, motion-based sy stems, acoustic sensors and electromagnetic sensors. These systems are able to d etect real-time kinematic conditions of objects on road and can detect the prese nce of vehicles and objects. To gain a high resolution of traffic patterns, it i s crucial to deploy these microsensors in large numbers and over a huge region. Therefore the design of microsensors must be cost-effective and energy efficient . 1.3 Proposed Idea Our proposed idea is to implement real-time dynamic pricing by deploying a dense network of microsensors along roads to gather information on traffic conditions . This information is communicated node to node to an access point which process es and passes it to the traffic management centre. The traffic management centre is usually the land transport authority and has two responsibilities to fulfil. Firstly, it needs to determine the level of congestion and the related toll rat es. Next, it has to communicate the updated toll rates to motorists approaching the congested zone using some electronic means (electronic billboards, text mess ages, etc.). 2. THEORY OF OPERATION 2.1 Detection Methods There are many commercially available sensing systems that utilise different ele ctromagnetic, acoustic and magnetic sensors to detect objects. Each type of sign al has its own advantages and drawbacks. The sub-sections below detail how some of the popular signals work and their limitations or advantages. 2.1.1 Infrared System Infrared systems are able to detect EM waves with wavelength around 750 nm - 1 m m and are superior to conventional video imaging systems. An infrared system can function independent of solar illumination since it relies on blackbody radiati on for detection. There are two types of infrared detection systems, the active and the passive. Active infrared systems emit infrared rays that reflect off ground surfaces and are received by a sensor. When an object appears in the rays, the reflection of the rays are distorted and results in a shorter reflection period and based on t he time taken for the ray to return, the distance to the object can be calculate d. Passive infrared sensors detect ambient temperature and when an object appears i n the vicinity, the sensors are able to register a change in radiation or the te mperature change to the surrounding temperature and respond to it. However, this is only effective in conditions whereby the ambient temperature remains unchang ed. Infrared systems are also able to detect motion and can respond to dynamic s ituations. Compared to visible imaging systems which are vulnerable to poor weat her conditions such as rain, infrared systems are capable of showing a complete image of an object and has huge potential for complementing image data fusion sy stems. 2.1.2 Ultrasonic System Ultrasonic sensors emit sound waves at a frequency between 25 50 KHz which is ab ove the human audible range. These sensors emit pulsed sound waves and measure t he time taken for a reflected pulse to return. Using this information, the dista nce to a surface can be calculated. To differentiate between the background road surface and a vehicle, the receiver is gated on and off with a user-adjustable interval. The distance to the background road surface is used as a reference and if any other distance is measured, a vehicle is detected. Doppler ultrasound is also used to calculate the velocity of objects by measuring the change in frequ ency of ultrasonic echoes due to moving objects but is more expensive. Some disa

dvantages of using ultrasound are that it is vulnerable to temperature and air t urbulence. 2.1.3 Video/ Image Processing Video and image processing methods involve the use of video cameras that monitor a certain region of space for changes in the images taken. When an obstacle app ears, a sequence of images is captured and variations in successive images are a nalysed to determine whether an object has appeared in the field of vision. For instance, data association tracking systems identify and track vehicles through a series of images. It begins by tracking groups of interconnected pixels in dif ferent frames and identifies one or more vehicles based on gradient or morpholog y. Gradient identification uses the edges of pixels to determine different objec ts whereas morphological identification uses features of known vehicles. Video and image processing is relatively popular with the prevalence of low cost , high performance microprocessors. Image segmentation also makes it possible fo r additional features such as the vehicle trajectory to be determined. However f or it to be effective a high mounting panel is required to prevent obstacles fro m obscuring the field of vision. It is also vulnerable to environmental conditio ns such as wind, dust and rain. Also, it is only cost effective when multiple de tection zones are required in the field of vision. 2.1.4 Millimetre-wave Radar technology Panasonic and Mitsubishi have developed a new millimetre-wave radar technology f or automotive applications. The new millimetre-wave radar incorporates a signal processing unit in the radar head and has a more compact and lightweight radar s ystem that is easy to install in vehicles. It is also designed to provide stable detection of targets under inclement weather conditions such as rain or snow. T he new technology can also measure the distance to a target and the targets relat ive velocity simultaneously. Current radar technologies can detect and measure the distance to a nearby vehic le but cannot detect the human body with high resolution due to very weak radar reflection of human body. Panasonics new automotive radar technology is capable of detecting humans and veh icles simultaneously despite weak reflections from the human body and that is on e of the reasons this technology is chosen for our project. 2.1.5 GPS GPS is a navigation system consisting of 24 GPS satellites that broadcast signal s to receivers in navigation systems. GPS detection of vehicles and objects is d one by calculating the time it takes for three or more synchronized signals to r each a receiver. This system of detection is able to determine the position and velocity of static and dynamic objects. However, GPS signals propagate by line of sight and communication is lost when t here is an obstruction in the propagation path such as travelling under tunnels. Furthermore, GPS suffers from multipath errors that occur because radio signals are reflected off surfaces and as a result take a longer time to reach the rece iver. It also has a relatively long signal acquisition time (30s to 15mins) and is less effective in real-time systems. 2.2 Communication network Sensor nodes are activated when a vehicle is within range and will start transmi tting traffic data wirelessly node to node until it reaches the traffic manageme nt centre, as shown in Fig.1.

The traffic management centre is the central processing station where most of th e data processing and analysis are done. After obtaining the traffic information from the wireless sensors, the management centre will determine the level of co ngestion on a certain road or highway. Computer algorithms will categorize the i ntensity of congestion into different tiers and calculate the corresponding toll rates. A few metrics that are of interest are vehicle density per square metre

of the road and the rate at which this value changes. Since traffic conditions a re dynamic, the toll rates vary with time. These toll rates should also vary for different classes of vehicles since a larger vehicle contributes to larger occu pancy of road and greater congestion. The computed toll charges is communicated to individual vehicles and displayed on the in-vehicle unit. There should be a r efresh rate that is frequent enough to display any changes to the tolls. 2.3 Dynamic congestion pricing model To design a suitable pricing scheme, we need to first determine what constitutes congestion and what credible metrics we can use to measure it. There is no spec ific definition for congestion but perhaps it is aptly defined as the state at w hich the network capacity of the road is reached and the flow of vehicles is at a minimum. From this definition we can infer several characteristics of congesti on. Firstly there is the issue of capacity. This capacity does not merely deal w ith the width of a road but the manoeuvrability of vehicles. It is concerned wit h the number of exits and alternative travelling routes. While road capacity con tinues to increase judiciously to meet rising demand, it is costly and represent s only a short term solution. Road pricing on the other hand aims to achieve a m ore permanent solution by forcing drivers to rationalise their travelling decisi ons. Road users should be penalized more if there are alternative routes and exi ts available but they choose to ignore it. Hence, we need to consider the number of nodes in a transportation network and charge a higher rate for roads with gr eater network capacity since it causes a greater negative externality. Secondly, pricing should be tagged with elements concerning the flow of vehicles . Elements include time, speed, volume, vehicle density, traffic cycle failure, the number of times drivers apply brakes and distance between vehicles. Some of these elements cannot be properly defined as a metric hence we shall only look a t time, speed and vehicle count. In conventional congestion pricing, time usually refers to the peak hours of tra ffic. However, rates set based on peak hours are not responsive enough to captur e the dynamic nature of congestion. Rather than using peak hours, we can measure an average time a vehicle takes to clear a stretch of road and tag the value of the toll to it. We would require information on the spatial distribution of veh icles, the distance of the road, the speed of the vehicles. In addition, there n eeds to be a virtual boundary to determine the time of entry and the time of exi t. A simple model of pricing would be to classify different speeds under differe nt strata and give a corresponding value. The toll rate should display an invers e relationship with the vehicle speed to reflect the severity of the congestion. There is also a possibility of continuously charging the vehicle for its presenc e on the road. Although this method does not tag the toll to the marginal cost o f externality, it incentivises the road user to remove his presence from the roa d as fast as possible. Such fees are expected to be low but variable for differe nt levels of congestion. In conclusion, the pricing model must be designed such that at the very least ma tches the marginal cost of adding another car to the congested zone. This margin al cost is largely determined using several metrics associated to congestion but can also vary vehicle to vehicle. 3. TECHNICAL SPECIFICATIONS Fig.2 Basic Infrastructure of a WSN node 3.1 Hardware Specifications 3.1.1 Power supply Due to advancements in micro-electronic mechanical systems (MEMs), microprocesso rs and peripheral components have become smaller and less costly. However, a lar ge number of wireless sensor nodes available in the market these days are still using batteries. No doubt these batteries are simply designed; they are not made to function as long-term power sources. Moreover, since the sensor nodes are us ually mounted on road fixtures, the maintenance and replacement of batteries bec omes a costly issue. Hence, we need to consider power sources that have long ope

rating lifetimes and/or develop protocols to reduce energy consumption of sensin g networks. The section below describes and compares an energy harvesting power source (solar energy) and a conventional lithium battery. Power density (uW/cm3) One year lifetime Solar (direct sunlight) 15000 Batteries (non-rechargeable lithium) 45 Batteries (rechargeable lithium) 7 Table.1 Comparison of energy scavenging and battery sources Fig.3 Comparison of power from various energy sources The data in Table.1 indicates that a typical solar cell placed under direct sunl ight has a much higher power density than a conventional lithium battery. Furthe rmore, Fig.3 shows that batteries are only viable for short lifetimes whereas so lar cells offer better power density for longer lifetimes. Although these figure s discount for environmental factors such as the weather and the efficiency of t he solar cell, they do however impress the fact that energy harvesting sources r epresent a far more sustainable option than batteries. To power a sensor node, hybrid energy power sources are used. These consist of e nergy harvesting solar panels and rechargeable batteries to store and supply pow er. Fig.4 IXYS Semiconductor Solar Cell Solar panels such as the one shown in Fig.4 are 22mm by 7mm by 1.4mm and are con nected in series to provide a maximum peak power of 16.6mW/cm2. These are energy efficient and can be considered for our purposes. For batteries, a combination of supercapacitors and rechargeable batteries can b e used to store energy. This reduces capacitive discharge and supports a longer battery lifetime. For example, two AA 2.7Ah NiMH rechargeable batteries in serie s can be connected in parallel with supercapacitors to lower the internal impeda nce and increase sensor node lifetime. Compared to lithium batteries, NiMH batte ries are cheaper and have simpler charging schemes. 3.1.2 Microprocessor Microprocessors incorporate units of control and data processing such as memory, converters, digital I/O interfaces and peripherals on a single integrated circu it. They are typically used for the sampling, analog-to-digital conversion, data processing and control of electric signals. Basically, a microcontroller system reads inputs, performs processing and writes to output nodes. The design of mic rocontroller interfaces is largely influenced by factors such as power consumpti on, functionality, size, cost and complexity. In order to carry out large-scale sensing, we can make use of an integrated lowpower wireless microsensor mote. One such example is the MICA2DOT. MICA2DOT by C rossbow Technology integrates a low-power processor (Atmel ATmega128L), a low-po wer radio platform (Chipcon CC1000), an open-source software operating system (T inyOS), transceivers and on-board peripherals. The integration of the processor and radio transceivers on a single product makes it attractive in terms of cost and energy efficiency. As such, MICA2DOT largely fulfils our criteria for traffi c surveillance and congestion management. Table.2 Specifications of MICA2DOT 3.1.3 Radio transceiver Radio communication is responsible for more than 90% of the power consumption of a sensor node. Using radio transceivers which include an integrated microproces sor, we can reduce the operating power of a sensor node. We aim to reduce power consumption of the transceivers to extend the lifetime of the sensor node. Howev er, reducing the power of the transceivers usually reduces the range of the sens or nodes; hence it is a challenge to devise a balanced system. The communication distance of the sensor nodes is determined by the transmission power, sensitivity of receiver, gain and efficiency of the antenna. Based on th

e Friis Transmission Equation, we are able to determine the relationship between the received power from the signal and the distance between antennas of nodes. A typical receiver sensitivity of -85dBm allows for a free space range of 25-50m . Using this estimation, we can determine the deployment range of the microsenso rs and the power required. 3.1.4 Sensor Sensors are devices that perform tasks like signal sampling and analog-to-digita l conversion. After much deliberation, we recommend the use of the millimetre wa ve radar. As described under section 2.4, the millimetre wave radar is an electr omagnetic wave with wavelength 10 to 1mm and frequency 30 to 300GHz. It has supe rior performance under poor weather conditions and is able to detect objects wit h high enough resolution to determine its velocity and location. The sensors are able to sense objects from ten to hundreds of metres and can be placed in different orientations. These can be mounted on a high place such as a lamppost or at the sides of the road. Millimetre wave radar systems can be eith er pulse radar, frequency modulated continuous wave (FM-CW) and spread spectrum radar. The choice of radar system affects the antenna gain and transmission powe r and this will have an impact on the power supply. Although the sensors in these systems are less rigorous than their ad-hoc counte rparts, they have to include a signal processing component. The signal processin g component converts the analog signal to a digital one via Fast Fourier Transfo rm (FFT) where the velocity and distance are calculated. 3.1.5 Enclosure selection The sensor node is expected to be small and cost effective; hence the enclosure must also be small. Apart from that, it needs to be durable enough to protect th e sensors and integrated circuits from variable weather conditions. An example o f an enclosure that fits our criteria is the Radar Box, a commercially available product from Legal Speeding. It is made of painted thermoformed polycarbonate e nclosure and offers protection against rainwater, UV rays, rust and other debris damage. It has interior dimensions of 134.6mm by 91.4mm by 31.8mm and is big en ough to fit most of the commercial radar detectors like the Valentine One and th e Escort RedLine. Solar panels can be easily mounted on top of the enclosure. 3.2 Software Specifications Wireless sensor nodes are capable of rudimentary computation, data acquisition a nd communication of data in a single or multi-hop wireless network. Usually, dat a is acquired over a distributed network and is transferred node to node until i t reaches an access point, a node that has superior computational ability and en ergy resources. General nodes are less privileged and are resource constraint in terms of processing ability and power available. Hence, there is a need for an operating system that is able to mitigate some of these disadvantages and employ smart strategies for resource management. This section discusses some of the op erating systems and algorithms that are currently available to wireless sensor n etworks (Table.3) and suggests a suitable operating system for the proposed sens or network. Platform MCU RAM Code Memory RF Transceiver Frequency Radio Range (feet) mica Atmel ATMega128L 4 KB mica2 Atmel ATMega128L 4 KB 500 mica2Dot Atmel ATMega128L z 500 MICAz Atmel ATMega128L 4 KB Cricket Atmel ATMega128L 4 KB TelosA TIMSP430 2 KB 60 KB TelosB TIMSP430 10 KB 48 KB BTnode3 Atmel ATMega128L 64 KB 868 MHz 328/ 500 EYES TIMSP430 4 KB 60 KB 128 KB TR1000 433, 916 MHz 200 128 KB CC1000 315, 433, 916 MHz 4 KB 128 KB 128 KB CC2420 CC2420 128 KB 128 KB CC1000 315, 433, 916 MH CC2420 2.4 GHz 410 CC1000 433 MHz 500 2.4 GHz 410 2.4 GHz 410 Zeevo-BT/ CC1000

2.4 GHz/

TR1001 868 MHz 984

intelmote ARM7TDMI (Core) 64 KB 512 KB Zeevo-BT 2.4 GHz 328 intelmote2 PXA27x (Core) 256 KB 32 MB CC2420 2.4 GHz 410 MANTIS nymph Atmel ATMega128L 4 KB 128 KB CC1000 315, 433, 868, 9 15 MHz 500 XYZ mote ARM7TDMI (Core) 32 KB 256 KB CC2420 2.4 GHz 410 ECR TIMSP430 2 KB 60 KB TR1001 868 MHz 984 ESB TIMSP430 2 KB 60 KB TR1001 868 MHz 984 Table 3: Sensor node platforms 3.2.1 System Architecture A suitable architecture requires coding that is versatile enough to handle diffe rent hardware components and allow for reuse of code across different applicatio ns. A flexible architecture allows for easy modification of code, occupies littl e memory and makes it possible to improve on any existing designs. Another feature that we wish to emphasize is the importance of an event-driven a rchitecture. An event-driven system is one that remains in idle mode until trigg ered by a change in environment. This way power limited systems can remain in a low power state until required. A handler determines the transition between idle to active modes. A faster transition time translates to better responsiveness a nd higher frequency of idle mode, hence less wastage of energy. Some types of ar chitecture are discussed as follows. A microkernel architecture (Fig.5) is one that distributes functionalities and s ervices to user-space programs located on servers. This reduces the kernel size and provides a high level of customizability. Services communicate via communica tion channels and if one server fails, the rest of the system is unaffected. How ever transfers between user and kernel maybe high and can cause latency problems .

On the other hand, a monolithic architecture kernel (Fig.6) defines a high level of virtual interface over services and applications. This means that each servi ce is associated with each other and bundled under a single large system. If any one component or service fails, the rest of the system is also affected. One of downsides is that the code is difficult to modify and understand.

3.2.2 Programming Since wireless sensor nodes are designed to be deployed over a huge area, it is not feasible to conduct software maintenance and reprogramming manually. Therefo re, support for remote reprogramming must be included in the selected operating system. How this is to be done is by flashing updates to several individual or c lusters of nodes and by propagating the updates via internode communication. Thi s is expected to be done during run-time and several concerns include propagatio n time, energy constraints, memory constraints and delays. Fig.7 outlines the ge neral propagation of such an update. Fig.7 Propagation of system upsate through WSN 3.2.3 Memory Memory is scarce in sensor nodes hence a dynamic memory management system is req uired. A dynamic memory management is flexible and allocates suitable memory siz es from a pool of available memory to applications. This raises the operational efficiency of the node and decreases the hardware required for a dedicated memor y space. Sensor nodes are expected to perform multiple execution threads therefore memory protection is required. The operating system must ensure that when a process ru ns, other applications cannot access its allocated memory. 3.2.4 Communication Protocol

Communication protocols for wireless sensor nodes have to account for the limite d power supply available and the overall lifetime of the node. A proper protocol should account for the different types of data received and sent by the nodes a nd provide a feasible framework to manage these data. With these considerations in mind, the protocol should also include MAC layered protocols and logical netw ork routing. Some examples of MAC protocols that come into mind include the sens or-MAC (S-MAC), timeout-MAC (T-MAC) and the traffic-adaptive medium access (TRAM A). All these protocols incorporate some form of sleep scheduling and are applic able for our purposes. Network routing can be done based on how the distributed data is collected and transmitted or can be done with energy considerations in m ind. 3.2.5 Power Most of the sensor nodes use batteries and are power limited. In order to prolon g sensor node lifetime, one of the ways is to employ a technique called duty cyc ling, where the processor and its components are put to idle or sleep mode to co nserve energy. Fig.8 shows the generic power consumption profile of a typical se nsor node. Fig.8 Power consumption profile The average power consumed is given by the following equation: where D = tactive/T. Thus, reducing the duty ratio lowers the average power cons umed. As mentioned above, the frequency of sleeping and the time it takes to wake up influences the energy overhead and are important parameters to be considered. Besides duty cycling, an energy efficient network can be designed. Parallel proc essing among the sensors is one such method. By distributing data computation an d analysis among the sensor nodes reduces energy dissipation and increases compu tational ability. Another way is to implement software to reduce computational e rror which contributes to energy wastage. 3.2.6 Suggested Operating System Nano-RK The above are some of the design considerations that are presented in a cursory manner and have not been expanded on. Designing a suitable operating system requ ires a thorough review of the demands of the wireless sensor network and prefera bly tested on a controlled and simulated environment before embarking on a large scale project. One popular operating system which we have found to be useful in our proposed idea is the Nano-RK. Among the many available operating systems, the Nano-RK is one that provides a r igorous framework for real time applications. In addition to being a lightweight system (using only 2Kb of RAM and 18Kb of ROM), it possesses preemptive multita sking, multi-hop network support and enforces energy usage limits to prolong sen sor node lifetimes. Preemptive multitasking refers to the execution of higher pr iority tasks which increases the responsiveness of the sensor node. However, pre emptive multitasking also implies that the current tasks need to be saved before engaging higher priority tasks and that could incur higher memory overhead. Thi s problem is further exacerbated due to a lack of memory management unit. The Na no-RK also provides custom MAC and routing protocols. The Nano-RK has its misgivings as an operating system but currently is one of th e more viable options for our wireless sensor network. 4. MODIFICATION ON IN-VEHICLE UNIT The in-vehicle unit acts as a wireless communication device between the traffic control centre and individual road users. After the traffic data from the WSN ha s been processed, the traffic control centre will communicate with vehicles on t he road via a vehicular telematics system . The system is connected to the centr al network and receives information that is updated every minute from the traffi c control centre. The associated toll charges computed from these traffic data a re then displayed on the screen of the in-vehicle unit, under the existing toll charges shown for the Electronic Road Pricing (ERP).

Fig.9 Mitsubishi in-vehicle unit for ERP system Similar to present In-Vehicle Units, a credit or debit card can be inserted to t he device and tolls are automatically charged to the users account.

5. FURTHER DEVELOPMENTS One of the possibilities for further developments is to seek ways to lower the c osts of production of sensor devices as well as that of implementation. The adve nt of nanotechnology and microsystems has allowed the manufacture of devices on the micro and nano scales. Miniaturization of components is one of the ways to r educe the cost of production. This idea can also be used to promote a greener and more environmentally friendl y society. A higher toll rate can be imposed to reduce private road usage and en courage use of public transport. Rebates can be given to drivers who do not trav el on congested roads within certain time intervals. This also acts as an incent ive for drivers to avoid congested roads and seek alternative travel routes to g et to their destination. Moreover, we can extend our idea to other urban applications that involve crowd control since millimetre wave radar can be used for human detection. Human conge stion is a lesser known problem in urban cities and commercial districts and we believe our idea can be applied in this area. 6. REFERENCES [1] CIS 307: Views, Components, and Architectures of Operating Systems (No Aut hor), College of Science and Technology, Temple University http://www.cis.temple.edu/~giorgio/cis307/readings/intro.html (accessed 21.07.20 12) [2] A Chandrakasan , A Sinha, A Wang, E Shih, M Bhardwaj, R Min, S H Cho, Low -Power Wireless Sensor Networks, Department of EECS, Massachusetts Institute of T echnology http://www.mtl.mit.edu/researchgroups/icsystems/pubs/conferences/2001/rmin_vlsi0 1_paper.pdf (accessed 21.07.2012) [3] J J Lukkien , M Stolikj, P J Cuijpers, Efficient reprogramming of wireles s sensor networks using incremental updates and data compression, Department of M athematics and Computer Science, System Architecture and Networking Group, Eindh oven University of Technology http://www.win.tue.nl/~mstolikj/publications/CS-12-10.pdf (accessed 21.07.2012) [4] M A Perillo, W B Heinzelman, Wireless Sensor Network Protocols, Department of Electrical and Computer Engineering, University of Rochester http://www.ece.rochester.edu/courses/ECE586/readings/perillo.pdf (accessed 22.07 .2012) [5] S Y Cheung, P Varaiya, Traffic Surveillance by Wireless Sensor Networks: Final Report, California PATH Research Report, Institute of Transportation Studie s, University of Berkeley

http://www.its.berkeley.edu/publications/UCB/2007/PRR/UCB-ITS-PRR-2007-4.pdf (ac cessed 22.07.2012) [6] M O Farooq, T Kunz, Operating Systems for Wireless Sensor Networks: A Sur vey, Department of Systems and Computer Engineering, Carleton University Ottawa, Canada www.mdpi.com/1424-8220/11/6/5900/pdf (accessed 20.07.2012) [7] D. Manjunath, A Review of Current Operating Systems for Wireless Sensor N etworks, Department of ECE, IISc, Bangalore, India www.comp.nus.edu.sg/~doddaven/cata.pdf (accessed 20.07.2012) [8] Traffic Congestion in Singapore (No Author), a singapore economist blog en try, dated 26.07.2006 http://a-singapore-economist.blogspot.sg/2006/07/traffic-congestion-in-singapore .html (accessed 23.07.2012) [9] S Xu, Development and Test of Dynamic Congestion Pricing Model, February 2 009, Department of Civil and Environmental Engineering, MIT [10] B W Wie, R L Tobin, Dynamic congestion pricing models for general traffic networks, 04.02.1999, School of Travel Industry Management, University of Hawaii http://www.sciencedirect.com/science/article/pii/S019126159700043X (accessed 25 .07.2012) [11] Matthew B. Higgins, Heighting with GPS: Possibilities and Limitations. R etrieved from http://www.fig.net/commission5/reports/gavle/higgins.pdf [12] William David, FMCW MMW Radar for Automotive Longitudinal Control . Retr ieved from http://www.path.berkeley.edu/path/publications/pdf/PRR/97/PRR-97-19.p df [13] Graham M Brooker, Understanding Millimetre Wave FMCW Radars . Retrieved from http://www-ist.massey.ac.nz/conferences/icst05/proceedings/ICST2005-Papers/ ICST_111.pdf [14] Alfred K, Richard B, The limitations of GPS. Retrieved from http://gauss .gge.unb.ca/gpsworld/EarlyInnovationColumns/Innov.1990.03-04.pdf [15] Limitation of GPS , retrieved from http://www.emory.edu/BUSINESS/et/552f all2000/gps/limitations.htm [16] Earth Measurement Consulting. Retrieved fromhttp://earthmeasurement.com/ GPS_accuracy.html [17] Limitation and accuracy with GPS. Retrieved from http://oklahoma4h.oksta te.edu/scitech/docs/geospatial/limitations_accuracy.pdf [18] Global Positioning System. Retrieved from, http://en.wikipedia.org/wiki/ Global_Positioning_System [19] Limitation of GPS. Retrieved from , http://www.locatacorp.com/applicatio ns-of-gps/limitations-of-gps/ [20] Know the limitation from hiking. Retrieved from , http://www.gadling.com /2009/11/02/hiking-with-a-gps/ [21] Todd Litman, London Congestion Pricing. Retrieved from http://www.vtpi.o rg/london.pdf

[22] Dr Chin Kian Keong , Road Pricing Singapores Experience. Retrieved from , http://www.imprint-eu.org/public/papers/imprint3_chin.pdf [23] Transport for London. Retrieved from, http://www.tfl.gov.uk/roadusers/co ngestioncharging/ [24] Tolling and Pricing Program. Retrieved from http://ops.fhwa.dot.gov/publ ications/fhwahop08047/02summ.htm [25] S Honma, N Uehara, Sensors are devices that perform tasks like signal sam pling and analog-to-digital conversion, June 2001, http://mitsubishielectric.com/ bu/automotive/advanced_technology/pdf/vol94_tr5.pdf (accessed 26.07.2012) [26] T Yamawaki, S Yamano, Y Katogi, T Tamura, Y Ohira, Millimeter-Wave Obsta cle detection Radar, 2000 http://www.fujitsu-ten.com/business/technicaljournal/pdf /15-2.pdf (accessed 26.07.2012) [27] Holger Karl, and Andreas Willig, Protocols and Architectures for Wireless Sensor Networks. John Wiley & Sons, 2005, pp. 15-329. [28] V. Raghunathan, C. Schurgers, S. Park, and M. Srivastava, Energy Aware W ireless Sensor Networks, Department of Electrical Engineering, University of Cali fornia, Los Angeles. [29] Mohammad Ilyas, and Imad Mahgoub, Handbook of sensor networks: compact w ireless and wired sensing systems. CRC Press LLC, 2005. [30] S. Roundy, P.K. Wright, and J. Rabaey, A study of low level vibrations as a power source for wireless sensor nodes, Computer Communications, vol. 26, pp. 1131-1144, July 2003. [31] Penella, M.T.; Albesa, J.; Gasulla, M.; , "Powering wireless sensor node s: Primary batteries versus energy harvesting," Instrumentation and Measurement Technology Conference, 2009. I2MTC 09. IEEE , vol., no., pp.1625-1630, 5-7 May 2009 [32] Mullner, R.; Riener, A.; An energy-ef cient pedestrian-aware Smart Street L ighting system, International Journal of Pervasive Computing and Communications V ol. 7 No. 2, pp. 147-161, 2011 [33] Microcontroller to Sensor Interfacing Techniques. BiPOM Electronics, Inc. (2006). http://www.bipom.com/documents/lectures/Microcontroller%20to%20Sensor%20 Interfacing%20Techniques.pdf. Accessed 20 July 2012 [34] Zigbee Protocol, New Circuits, http://www.newcircuits.com/article.php?id=t ut004. Accessed 20 July 2012. [35] Radar Box, Legal Speeding. http://www.legalspeeding.com/products.htm#radar box. Accessed 20 July 2012. [36] MICA2DOT Wireless Sensor Mote. Crossbow Technology, inc. https://www.eol .ucar.edu/rtf/facilities/isa/internal/CrossBow/DataSheets/mica2dot.pdf. Accessed 20 July 2012. [37] IXYS Semiconductor, Farnell Element. http://fr.farnell.com/ixys-semicondu ctor/xob17-04x3/cellule-solaire-0-63v-12-6ma/dp/1426532. Accessed 20 July 2012.

7. APPENDIX