Adaptive traffic control systems are employed in cities worldwide to improve the efficiency of traffic flows, reduce average travel times and benefit the environment via a reduction in fuel consumption. One of the main and most common functions of such systems lies in adaptive control of traffic lights. This ranges from simple lengthening or shortening of green and red light durations in an intersection according to the actual presence of cars in the respective lanes, to coordination of green light phases among neighboring intersections on main thoroughfares (to avoid frequent stopping of traffic), and manual intervention in response to abnormal incidents, such as traffic jams caused by accidents or large public events. This adaptivity is made possible with the use of sensors (typically in the form of magnetic loop detectors embedded under the road pavement) that feed data to roadside traffic light controllers, and a communications infrastructure that connects among the intersections and a traffic management centre, as well as, in some cases (typically in large cities), a hierarchy of regional computers (RC) that perform the control decisions for respective portions of the system. In addition, the traffic control system may include devices other than traffic lights, such as variable message signs or variable speed limit signs, which can be configured dynamically in response to incidents.

Traditionally, the communications layer of traffic control systems has been based on wired connections, either private or leased from public telecommunications operators. While for many years such leased lines (operating at 300bps) have served their purpose well, they have several shortcomings, such as a significant operating cost, inflexibility, and difficulty of installation in new sites. In certain cases, alternative solutions, operating over public infrastructure, have been deployed for specific sites where private or leased lines were not a viable option; these ranged from ADSL, regular dialup, or cellular (GPRS). However, using public network for traffic control could suffer from problems such as inconsistent delay jitters and and reliability issues. Such problems can generally be traced to the fact that, in networks intended for public use, the traffic control application is a minuscule component of the data, and, therefore, of the revenues for their operators; consequently, there is little incentive for providers to offer a special service level agreement to the road authority at competitive prices.

In recent years, there has been considerable interest in wireless mesh networks and their deployment in metropolitan areas, from both a commercial and a research perspective. Trials in several major cities in the US (e.g. Philadelphia, New Orleans, and others) and worldwide (e.g. Taiwan) have shown mesh networks to be a viable technology that can compete well with alternative ``last-mile'' connectivity solutions to the public. Correspondingly, most of the research on metropolitan-area wireless mesh networks (MAWMN) has focused on maximising the throughput that can be extracted from them, in the anticipation that their major use will be public, for purposes such as accessing the Internet or conducting voice calls. On the other hand, little attention has been directed to the aspects of reliability, security, and latency, which are most important if MAWMN are to be considered for replacement of mission-critical infrastructure, such as traffic control system communications.

The Smart Transport and Roads Communications (STaRComm) project at National ICT Australia (NICTA), started in August 2005, sets out to develop protocols that enhance the reliability and reduce the latency of mesh networks, and thereby enable them to be used as the communications layer of traffic control systems. Note that STaRComm is one part of the larger STaR project, which seeks to enhance a variety of facets of the traffic control system, including optimal traffic light scheduling and vehicle sensing. A multi-hop wireless testbed has been built in the first stage of the project to facilitate the aforementioned research directions. We are currently collecting measurements from the testbed to understand factors that exert the most influence on the signal quality and network performance. The measurement data allow us to draw important insights about the suitability of different radio technologies for the purposes of establishing a mission-critical, reliable communications layer based on a wireless mesh network.
 

Developed and maintained by the Roads and Traffic Authority (RTA, formerly Department of Main Roads) of the state of New South Wales, the Sydney Coordinated Adaptive Traffic System (SCATS) is one of the most popular traffic management systems used worldwide. Its main task is to adjust, in real time, signal timings in response to variations in traffic demand and system capacity. The current generation of SCATS connects not only the traffic signal controllers, but also a wide variety of other roadside devices, such as vehicle detectors, fog detectors, closed-circuit TV (CCTV) cameras, and more. Real-time data from all these devices are collected and transported to a central traffic management centre (TMC) for analysis and optimum control of road traffic. The performance of SCATS, therefore, depends critically on the capabilities of the underlying communication system that transports roadside data to and from the TMC.

The existing communication system of SCATS relies strongly on third-party wired infrastructure (provided by Telstra, Australia's largest telco). The bulk of the communications to the intersections, namely the traffic light controllers and vehicle detectors, are predominantly made using serial point-to-point connections over standard voice-grade telephone lines, using 300bps modems. This is also the most common method of connecting between the TMC and other low-bandwidth devices, including variable message signs, variable speed limits, ramp meters, and over-height detectors. High-bandwidth devices, namely CCTV cameras (offering image input to the traffic operators), are connected using ISDN or dedicated lines, including optical fibers in a limited region of the city centre.

At the core of the SCATS operation is a periodic exchange of messages between the controlling computer and each and every intersection (via the point-to-point links). This exchange happens every 1sec, and is initiated by the computer which sends to the intersection's local controller a command message, instructing it about the next phase it should switch to and the timing of that switch. The controller, in turn, is required to reply with an acknowledgment, which includes information from the intersection's sensors. If an acknowledgment is not received within one sec from the time the command message is sent, it is retried once; after the second time an acknowledgment fails to arrive, the communications link is declared failed, and SCATS instructs all controllers at the respective cluster of intersections to fall back into a `default' self-controlling mode, where decisions about the timing of green light phases are made locally and independently. Likewise, a controller will fall back to this mode upon not receiving a command message. Once triggered, a controller will stay in the self-controlling mode for at least 15 minutes; if another communications failure happens during this time, the duration of this mode will be extended by another 15 minutes, and so on. Obviously, the self-controlling mode, where the decisions at intersections are uncoordinated, can lead to a severely suboptimal traffic control, particularly in a busy thoroughfare during rush hour. Accordingly, though the bandwidth required from the communication links is quite low (comfortably handled by 300bps modems), this one-sec latency is critical for an efficient operation of the system.

The currently used SCATS infrastructure, based on wired communications, suffers from the following main problems:

• Slow installation and inflexibility. In most cases, installing a new line at a road site (especially a remote site) involves earth excavation, which is very slow and with adverse effects on existing infrastructure. Moreover, due to traffic growth and pattern changes over time, it is necessary to expand and adapt the network topology from time to time, which is a costly and cumbersome process for a wired network. Consequently, network topology adaptations tend to be put off and do not respond quickly enough to demand changes, resulting in a far less than optimal traffic control performance.
• High capital and operating cost. The installation of a wired connection at a new site, or repairs at an existing one, carries a high cost due to the material and labor required. More importantly, the ongoing fees for leasing the wires from the telephone company run very high; currently, RTA pays nearly $40 million annually to Telstra in leasing fees for connecting the traffic signals and other roadside devices to SCATS.
• Low bandwidth. Modem-based leased lines support bandwidth less than 32 Kbps. While these low-bandwidth telephone lines are adequate for connecting traffic signal sensors, they cannot provide adequate support for connecting high-bandwidth applications, e.g. high-resolution video cameras, that increasingly becoming necessary to effectively monitor traffic pattern on our roads.

With wireless solutions, there is no cabling involved. Wireless can therefore provide fast installation and exceptional flexibility. Cost can be reduced significantly by building a private wireless network, because there will be no monthly charges to be paid to telephone company (some small license fee may apply). Moreover, the installation cost will be low because there will be no cabling-related labour. Finally, it should be noted that recent advances in wireless technology provide bandwidth that is more than adequate for connecting many high-resolution roadside cameras to SCATS.

Given a wide spectrum of available wireless technologies and services, there can be several options for going wireless: third-party service provider and in-house dedicated networking. Existing voice/data cellular service providers, e.g. GSM/GPRS/3G operators can provide circuit or session based wireless connectivity between roadside equipment and TMC using the standard services. However, these would attract either monthly charges (subscription fees) and/or volume charges. Additionally, the latency over such public networks is usually much higher than the 1-second requirement of traffic controller communication. MobitexTM is a data-only cellular service that can meet the latency requirement, but is not immune from the third-party charges. The cellular services may not be able to full-fill the high-bandwidth requirement of roadside video servers.

A more attractive option for going wireless is to build dedicated a wireless network using widely available, standards-based, low-cost wireless technologies, e.g. IEEE 802.11x and 802.16x. 802.11x equipment is cheaper, less complex, and operates entirely in the unlicensed spectrum (no licensing fee). On the other hand, 802.16x is more reliable (has multiple carrier frequencies to avoid interference), has longer range, and better features to cater for a diverse range of communication needs of future roadside equipment. It is possible to operate 802.16x in both license and unlicensed spectrums.

A more thorough (experimental) study is required to assess the performance of available wireless options. It is advisable to first complete a comprehensive and thorough technology evaluation study, preferably involving laboratory and/or field tests, to identify the most suitable wireless option (e.g. 802.11 or 802.16 etc.) to be used as the underlying physical platform over which to build the ultimate communication infrastructure. One of the objectives of this study would be to look for features in the standard/service that can be exploited later in the specific research efforts. The study may reveal that more than one standard be used to meet the requirements of diverse ITS applications in the most cost-beneficial way.

Because of the enormous benefits, wireless has been considered as a potential solution to connect roadside equipment to TMC. Wireless, however, a few new challenges that must be overcome before it can be deployed in a mission-critical application like SCATS, irrespective of which wireless option is selected. Below we identify the communication problems or challenges that arise when wireless is adopted instead of wired solutions.

• Latency. Wireless can potentially increase latency due to several reasons. If a new circuit or session needs to be set up each time a message is transmitted, the circuit/session set up delay can extend the latency by several seconds in a public cellular network (e.g. GSM/GPRS). Some technology, e.g., IEEE 802.11x, uses a common wireless channel (it is cheaper to share channel) among many contending devices causing potential conflict. To avoid such conflicts, some form of medium access control (MAC) is implemented by these technologies. MAC introduces some delay before data can be transmitted on the wireless channel.
• Reliability. Wireless signals are susceptible to interference from other signals in the vicinity operating in the same or adjacent spectrum. Given that ITS equipment is deployed in public area, such interference will be the norm rather than exception. Interference can corrupt messages transmitted over the wireless medium. Some frequencies do not work well (or at all) if there is no direct line-of-sight between the two communicating end points. In a dynamic context of public roads, roadside equipment may frequently face line-of-sight problems due to transient obstructions, e.g. a high vehicle carrying a tall crane etc. Also in vehicle-to-roadside communications, a car in the near-lane may obstruct communication between a far-lane car and the roadside equipment. Temporary outages, i.e., periods when no wireless signal is available, therefore, is a real issue to deal with.
• Security. What makes wireless so vulnerable is the fact that the attacker does not have to gain physical access to the channel from any predefined access point. Roadside wireless components are well within the wireless range of passing motorists and pedestrians, which make them vulnerable to intrusion, denial of service, and other forms of security threats. Every communication system has some unique security holes that are created by system-specific features and functions. There may be new categories of security threats when wireless is deployed in SCATS. Given increased dependence of the new generation of SCATS on intelligent roadside equipment, a compromised device (or a set of compromised devices) can cause serious inconvenience and even loss of lives on our roads.
• Scalability. As mentioned earlier, wireless systems are sensitive to interference from other communicating devices operating in the vicinity. Additionally, if a common wireless channel is shared among all devices within a given area (cell), the MAC delay increases rapidly as the number of competing devices increases. Another scalability issue arises from the processing overhead that is required at a central radio base-station. The more remote radios there are in communication with the central radio, the more processing that must take place. The radio controller at the base-station will simply not be able to process all incoming radio signals if there are too many of them. Due to these reasons, sudden increase in the number of communicating devices within a limited geographical area (sometimes referred to as flash crowd), e.g., during an emergency, can cause a wireless system to collapse when it is needed most.

• Multi-hop wireless communication. Wireless links are much more prone to error and failure than the wired counterparts. How to provide reliable communication in the face of a wireless link failure between a field device and the wireless base-station (master device) for example becomes a challenging problem. Innovative network topologies and data routing techniques, e.g. dynamic multi-hop wireless communication, need to be investigated with a goal of quickly recovering from such failures without violating the latency bounds imposed by the ITS applications. Multi-hop can also address the scalability issue by reducing the number of direct wireless channels needed to connect a target base of field devices.
• Intrusion detection. Roadside wireless components are well within the wireless range of passing motorists and pedestrians, which make them vulnerable to intrusion, denial of service, and other forms of security threats. A compromised device (or a set of compromised devices) can cause serious inconvenience and even loss of lives on our roads. Robust security mechanisms must be in place to protect the wireless network from intruders. Given some of the devices may have very limited computing and energy resources, designing a security mechanism that does not deplete the resources is a challenging task.
• Characterisation of radio links. Movement of vehicles on the road, data traffic congestion, and weather fluctuation can continuously modify the characteristics of the radio channels that connect ITS devices to the fixed wired components or passing vehicles. In addition, tunnels and other underground environments pose new dynamics to radio wave propagation. Tunnel entrances, where a vehicle switches between surface and underground communication, can also exhibit special effects on radio communication. Although research on radio propagation models abound in the literature, characterisation efforts of radio channels under the above mentioned ITS scenarios are rare. Understanding radio propagation under these environments will help higher layer researchers to design more efficient communication protocols.
• Cross-layer communication. As mentioned earlier, radio characteristics in ITS wireless systems are unusually dynamic, and even more so when road-to-vehicle communication is concerned. This is manifested by the volatility of signal strength, available bandwidth, and even brief outages (e.g. due to temporary obstruction by a high-rise building). Given the high level of dynamics in the ITS radio links, higher layer protocols, such as routing or congestion control, need to constantly adapt their algorithms and parameters to remain synchronous with the current situations. Unfortunately, the strict layering principle of traditional communication systems prevents a layer from adapting its functionalities according to events happening in other layers of the system. To achieve this kind of adaptiveness at the higher layers, it is necessary to introduce cross layer communications. Cross layer communication is a recent area of research especially motivated by the introduction of wireless and mobile communications. There are many models for cross layer communication. This research will investigate an appropriate cross layer communication model that is suitable to ITS applications.
• Delay-tolerant protocols: In addition to informing road users of conditions through roadside displays, traffic information could also be delivered wirelessly to user devices using short-range communication systems (e.g. wireless LAN or PAN technology) when vehicles pause at traffic lights. This would require new delay tolerant protocols so that when connectivity is interrupted (e.g. vehicles moving between red lights), the transfers would be suspended rather than being aborted.
• Mobile communication. In ITS environments, there are a variety of information and data streams that need to be dispatched to or collected from a vehicle for safety and convenience. For example, a camera mounted near a rail crossing may send video stream to a screen in front of a train driver to help the driver view the rail crossing well before the train reaches the area. Conversely, there may be in-vehicle devices, for example a location tracking device, that regularly need to send important data to a central management server. Such data communication between roadside devices (or central servers) and vehicles presents special challenges, arising due to the movement of the vehicle. Indeed, as a vehicle moves, coming in and out of the ranges of roadside wireless links and crossing cells, the routing of the data stream between the vehicle and the network infrastructure changes constantly. The challenge is especially apparent for real-time data streams with high reliability and latency requirements. Existing solutions that support routing to mobile devices are, almost invariably, reactive; that is, they only act after a change of the wireless link is detected, and typically introduce delays of at least several seconds due to various overheads (such as registration of the new location in a “home agent”), which renders them inadequate for real-time data streams. Clearly, research is needed for more proactive and scalable methods of routing critical real-time data to a large number of moving vehicles. This research is particularly significant when 802 series technology is adopted instead of public cellular networks (802 was not designed for mobility).
• Reliable multicast. When the same information is to be sent to a group of devices (for example an intersection collision detector wants to warn three nearby cars approaching the intersection from three different directions), multicasting provides an elegant solution over point-to-point communication. Retransmissions are often used for reliability. However, retransmitting the information whenever one or more device receive it incorrectly due to error in the wireless channel makes the system very inefficient in terms of wireless usage and undue processing in the wireless devices. More elegant algorithms will have to be investigated to provide reliable multicast over wireless.

• Testbed on Google Map
  Our initial testbed covers seven traffic lights in the suburban area of Sydney. These intersections are chosen because they represent a typical suburban area with lots of traffic, foliages, pedestrians and high-rise residential buildings. In addition, the 200-500m range is representative of 90% of the distance between traffic controllers in the Sydney CBD area. In the next phase, we will extend our testbed to 15-20 nodes.
   
• Testbed Topology
  The initial testbed consist of three main components: Regional Computer (RC), mesh node and curbside controller box. Mesh node 1 (at intersection 521) also serves as a gateway to connect the testbed to the Regional Computer at NICTA through Sydney University network. Mesh routers are interconnected with each other using multiple radios operating at the frequencies of 900MHz and 2.4GHz. Each mesh node is also equipped with an Unwired modem (a WiMax variant, operating at 3.5GHz) for back-haul access. The curbside controller is connected to mesh node through Ethernet-over-powerline connection. We also equip the curbside controller box with a 802.11 radio so that it can be also accessed wirelessly.
  
  * node 1
  
  
  * node 2
  
  
  * node 3
  
  
  * node 4
  
  
  * node 5
  
  
  * node 6
  
  
  * node 7
  
  
  * curbside node
  
  
  
  
  
  
  
  
  * gateway node