Journal of Intelligent Transportation Systems, 13(3):111126, 2009 Copyright C Taylor and Francis Group, LLC ISSN: 1547-2450 print

/ 1547-2442 online DOI: 10.1080/15472450903084212

A Framework for Performance Evaluation of Communication Alternatives for Intelligent Transportation Systems
YONGCHANG MA,1 YAN ZHOU,2 MASHRUR CHOWDHURY,2 KUANG-CHING WANG,2 and RYAN N. FRIES3
1 2

IEM, Inc., Morrisville, North Carolina Clemson University, Clemson, South Carolina 3 Southern Illinois University, Edwardsville, Illinois

The communications infrastructure is critical for effective and efcient operation of intelligent transportation system (ITS) functions. Relatively large investments from the public sector are usually required to construct and maintain or provide leasing fee to private telecommunication companies for such infrastructure. Consequently, a rigorous analysis and evaluation of communication alternatives is of utmost importance. This research sought to develop a systematic framework and associated platform that will facilitate the evaluation of the communication alternatives for ITS. The framework includes identifying alternative architectures, such as distributed, centralized, or a combination of both, and identifying communication mediums including wired, wireless, or a combination of both. This article presents, as a part of the framework, important measures of effectiveness (MOE) for making objective comparisons between alternatives based on the performance of the communication systems related to ITS. A simulation platform that integrated a trafc and communication simulator was used for efciently evaluating communication alternatives with different architecture and mediums by generating important MOEs. A case study was performed for a test network in Greenville, South Carolina. The proposed evaluation framework identied four communication alternatives, namely, the centralized wired, distributed wired, centralized wireless, and distributed wireless, to generate the selected MOEs, such as throughput, delivery ratio, and throughput cost ratios, for comparing and analyzing these alternatives. Keywords Wireless; Intelligent Transportation Systems; Evaluation; Architecture

INTRODUCTION Ensuring that the highway transportation system remains a productive part of the nations infrastructure in the coming decades without costly expansion is of paramount importance. Increasing population, more vehicles, and urban sprawl now impact a highway system already overburdened and inadequately maintained. Many countries and areas around the world have been using technologies and systems to better manage and control their surface transportation network under the umbrella
Address correspondence to Mashrur Chowdhury, Associate Professor of Civil Engineering, 216 Lowry Hall, Clemson University, Clemson, SC 29634. E-mail: mac@clemson.edu

of intelligent transportation systems (ITS). The operation of numerous key components of ITS, such as real-time trafc surveillance and management, incident management, traveler information, natural and human hazard evacuation, heavily relies on the support of an effective and efcient communication system. State-of-the-art trafc surveillance systems around the world have been built with an emphasis on centralized observation and control (City of Cape Town, 2005; Road and Trafc Authority, New South Wales, 2006; Tokuyama, 1996; U.S. Department of Transportation [DOT], 2006). Transportation agencies deploy as many sensors as affordable along the highway and establish trafc management centers (TMCs) at central locations to collect data from sensors for making centralized control

111

112

Y. MA ET AL.

decisions. Substantial investments have been made to connect all sensors to central or regional controllers with dedicated communication links. Following predetermined schedules, roadside sensors transmit data to TMCs, where human operators identify possible incidents from the continuous data streams and initialize reaction decisions. Several problems arise from the existing centralized trafc surveillance network. First and foremost, the required dedicated communication infrastructure is prohibitively expensive as a system grows in coverage and number of sensors, thus making wide deployment difcult as a system expands to broader suburban and rural areas. A dedicated communication infrastructure and centralized control centers are also vulnerable to terrorist attacks and natural disasters. Furthermore, the response time of utilizing a centralized decision-making system is generally long. Lastly, human operators who monitor the sensors endure high working stress, which in turn decreases the system reliability. Distributed control concepts are not new to trafc control systems. To locally optimize trafc delays, trafc signal controllers have long been organized in local clusters. State-of-the-art of such trafc signal control systems include SCOOT (Split, Cycle, Offset Optimization Technique; Siemens, 2006), Sydney Coordinated Adaptive Trafc System (SCATS; Tyco Integrated Systems, 2006), and RHODES (Real-time Hierarchical Distributed Effective System; Mirchandani and Head, 1998). Though these methods may be effective for todays trafc control, they have been limited to the scope of xed signal control clusters and have required expensive communication infrastructure. In Coifman and Ramachandran (2004), the authors outlined a vision of deploying intelligent sensors along highways that could engage in distributed sensing and local data processing to report only concise information to TMCs or other responsible controllers if an anomaly is detected. The strength of this approach lies in the ability of sensors and controllers to make collaborative decisions without human intervention. The tradeoff between constant centralized control capability and communication cost needs to be carefully balanced. In existing on-line centralized trafc management systems, communication links continuously send data from trafc sensors to staffed centralized TMCs for assessment. Because these data frequently require no trafc management action, unnecessary communication costs are incurred. In addition, these systems are vulnerable to a single point of failure and suffer from scalability issues. With distributed-only systems, there is no single point of control; however, they are more difcult to implement for system-wide optimization. On the other hand, there exist communication medium options, which can be grouped into two categories: wired and wireless. Although the current prevalent communication medium for trafc control systems is wired cables in copper or optical ber, technology advances make wireless communication progressively more attractive to the transportation community (Cai and Lin, 2005; Smith et al., 2004; Stephanedes, Douligeris, and Takaba, 1996; Yang et al., 2000). The characteristics of different mediums in terms of

performance, cost, and reliability require a thorough investigation under specic ITS application scenarios. The importance of evaluating alternative communication schemes in a quantitative fashion can never be overemphasized. However, there remain important technical knowledge gaps to accomplishing this task. There has been limited work that addresses the detailed assessment of communication system for trafc management in terms of detailed measures of effectiveness, such as throughput, delay, and throughput-to-cost ratio. The objectives of this research are to present a framework for identifying and evaluating communication alternatives and applying an integrated trafc and communication simulation platform for evaluation of the communication alternatives. This paper used a case study of trafc surveillance and incident management as an example of evaluating communication alternatives for ITS applications.

RELATED WORK There are a variety of documents developed under Federal Highway Administration (FHWA) sponsorship and intended to help transportation agencies obtain better understanding of wired and wireless communication for ITS applications (Gordon et al., 1993; Gordon and Tighe, 2005; Klein, Mills, and Gibson, 2006; Leader, 2004; Neudorff, 2003). The Communications Handbook for Trafc Control Systems surveyed various available communication medium for trafc control applications (Gordon et al., 1993). Another handbook, Trafc Control Systems Handbook (Gordon and Tighe, 2005), reviewed the emerging technologies and control concepts for planning, designing and implementing trafc control systems. The Telecommunications Handbook for Transportation Professionals (Leader, 2004) briey introduced the history and basic concepts of telecommunications systems used to transmit voice and data information. The Trafc Detector Handbook (Klein, Mills, and Gibson, 2006) comprehensively surveyed the operation, application, design, installation, and maintenance of trafc sensor technologies. Another study sponsored by FHWA evaluated the performance of various digital subscriber line technologies (xDSL) with both laboratory experiments and eld tests (Jones, 2002). The study implemented high-speed data services (e.g., 2 Mbps) with xDSL on the existing twisted pair wire for transferring trafc video images, and their eld studies showed that the xDSL technologies were able to maximize the DSL throughput and subsequently to optimize the video motion/quality relation. The Texas Department of Transportation sponsored the development of a reference guidebook and training workshop to establish a fundamental level of understanding of wired communication concepts and technologies among state transportation engineers and an evaluation framework for wired communication alternatives (Brydia et al., 2005). This guidebook recommended the different criteria for choosing wired technologies (e.g., serial, ISDN, DSL, T1/T3 twisted pair, and ber), based vol. 13 no. 3 2009

intelligent transportation systems

COMMUNICATION ALTERNATIVES

113

on the number of devices, bandwidth, latency, distance, and cost. The California Department of Transportation (CalTrans) and FHWA conducted a eld operational test (FOT) between June 1994 and September 1998 to evaluate the benets of a mobile surveillance system with wireless communication interface (Mastako and Klein, 1999). In 1998, the Philadelphia Satellite Communication Demonstration project evaluated the effectiveness of using very small aperture terminal (VSAT) Ku-band satellite communications for trafc and incident management on I-95 corridor (Habesch, Jehanian, and Ziemer, 1998). Compared to terrestrial-based copper and ber opticbased closed circuit television (CCTV) systems, the VSAT-based CCTV system was found to be superior in terms of quality of service assurance, delay, and jitter control. In 2002, the Kentucky Transportation Center at the University of Kentucky implemented and evaluated a base station based wireless communication technology as part of the TRIMARC trafc management system (Hunsucker, 2002). This study investigated the use of a 220-MHz wireless communication system to transmit trafc measurements from eld sensors to trafc management center to support real-time trafc management in Louisville, Kentucky, nding that this 220-MHz communication system was equal to or better than the leased phone line in terms of functional reliability and cost effectiveness. The Wisconsin Department of Transportation (WisDOT) has implemented a statewide digital microwave backbone infrastructure that is used to transport voice and trafc data for 161 public safety agencies throughout the state (Verhyen, 2005). Among the existing evaluation efforts, some measures of effectiveness (MOEs) were recognized as the most important indicators of the performance of the communication system. Gordon et al. (1993) summarized possible attributes such as bandwidth, signal attenuation, latency, power consumption, signal-to-noise ratio, bit error rate, and error control technique as the fundamental MOEs for evaluating performance of the communication network. The authors also suggested that reliability, maintainability, and expandability were important for overall effectiveness of communication system. In addition, quality of service assurance, delay, and jitter control of video motion image were also widely used MOEs to assess the performance of a communication network (Habesch, Jehanian, and Ziemer, 1998). Mastako and Klein (1999) found that the portability and reliability of a communication system were key factors to realizing the expected functionality of the mobile surveillance system. Hunsucker (2002) evaluated the owned and leased wireless network in terms of functional reliability and cost effectiveness. Jones (2002) considered throughput and video image/motion quality as the MOEs for evaluating communication systems supporting trafc surveillance systems using CCTV. Texas DOT identied the number of devices, communication link bandwidth, and latency as important criteria for evaluating communication alternatives (Brydia et al., 2005).

In addition to eld studies and analytical tools, a few studies have attempted to develop an integrated simulation platform connecting vehicle and communication simulators for the purpose of communication alternative evaluation. Earlier work on integrated vehicle and communication simulations has been used to create simplied models of communication characteristics (Ghaman et al., 2003; Hsin and Wang, 1992; Sukthankar et al., 1996). Fitzgibbons et al. (2004) integrated CORSIM vehicular trafc simulator and the QUALNET communication network simulator for modeling vehicle ad hoc network. For intervehicle communication in vehicle ad hoc networks, Yin et al. (2004) implemented software objects for modeling vehicular trafc ows, driver behavior, and network protocols, and Avila et al. (2005) adopted vehicular trafc models based on realistic street maps and simplied communication models based on state machines. However, none of these studies addressed communication among xed eld equipment. More recently, simulators integrating microscopic vehicular trafc and detailed network protocol modes were developed for vehicle-to-vehicle communication (Choffnes et al., 2005; Fujimura and Hasegawa, 2005). These authors made a convincing case that an integrated vehicular trafc and communication networking simulator revealed important ndings not otherwise observed. The choice of vehicle and communication modelers is critical to the accuracy of an integrated simulation platform. In this article, the PARAMICS vehicular trafc modeler and the Network-Simulator version 2 (ns-2) were used to develop an integrated simulation platform. Previous studies have found PARAMICS to be superior in its detailed trafc modeling, which closely corresponds to real-world scenarios (Boxill and Yu, 2000; Chowdhury et al., 2007).

METHODOLOGY The proposed framework is intended to be complementary to the national ITS architecture (USDOT, 2007b), based on which regional and project ITS architecture has been dened to specify the communication requirements for information ows between different physical entities (subsystems and terminators). The proposed framework serves to assist ITS developers to determine the right communication technology to meet the specied requirements in the most cost-effective manner. The choice of communication alternatives concerns the communication medium, topology, and protocols of specic communication technologies. In spite of the wide variety of available communication technologies, the framework allows one to characterize the communication systems medium and topology, such that demanded information ows can be evaluated with selected MOEs using an integrated simulation platform. By incorporating cost information of the compared communication alternatives, a cost effectiveness can be conducted to aid decision-makers in choosing the most cost-effective communication alternatives for any specic ITS application needs. vol. 13 no. 3 2009

intelligent transportation systems

114

Y. MA ET AL.

Purpose of a Communication Alternative Evaluation Framework The purpose of the proposed framework is to enable evaluation of a wide range of alternative communication solutions for use by a given ITS application. To maximize its general applicability and adaptability, the framework allows modeling of any communication systems, rather than being limited to any specic protocol. The communication model also allows modeling of a wide range of ITS application properties (including achievable data rate, communication drop out probability, and network topology constraints). Consequently, the framework can model any ITS deployment given its data trafc, network, and environmental/weather constraints against its communication performance and communication infrastructure costs. To focus our presentation of the design ow rather than an excessive set of design parameters, the case study presented later does not exercise all such parameters; instead, it demonstrates the modeling of a practical network in South Carolina and discusses the expected impact of various important parameters that can be modeled by the framework. Though different public and private ITS components incur different communication patterns and requirements, the framework is capable of modeling any such diverse applications by conguring the corresponding communication and trafc models for each respective application. The proposed framework analyzes the performance and cost of only the communication infrastructure. The cost of the ITS devices is expected to be the same or differ little regardless of their underlying communication infrastructure. The adoption decision should depend on the set of required devices, the adequacy of network throughput for the devices, and the device plus infrastructure cost. Because the device cost is considered invariant with the alternative communication infrastructure, the constant cost will not affect the cost effectiveness of selected communication technology.

Centralized Network Sensor

Distributed Network Controller

Figure 1 Topology of centralized and distributed communication network.

maintains an adaptive data rate based on the decisions of each device. Moreover, in distributed systems, multiple sensors and controllers may share the bandwidth of communication links among them. Here, competition for communication resources might occur. Figure 1 illustrates the typical topology of centralized and distributed communication networks. For selecting the communication medium, although wired lines are the most prevalent communication medium used in vehicular trafc control system, wireless communication has become a popular technology for recent ITS applications. Wired communication can be costly for large-scale implementation. It also causes inconvenience for maintenance and system extension. However, wireless communication can be affected by terrain, obstructions, and weather. The design layout and parameters of the proposed alternative will be input into a Geographic Information System (GIS) database. After identifying all the alternatives that need evaluation, an interface between the GIS software and the simulation software would convert the ITS assets into a format ready for simulation. MOE Selection The performance and costs for different communication alternatives must be evaluated with respect to the specic communication needs of the ITS application in question. The problem of quantifying, measuring, and controlling the performance metrics of a network has been studied extensively in the context of quality of service (QoS) analysis (Peterson and Davie, 2003). The MOEs for the ITS communication system must therefore be selected in terms of the proper QoS metrics with respect to the application requirements. An ITS communications system must transfer information from eld components to the trafc operations center, which then transmit responses and commands to various eld components (Gordon et al., 1993). According to the respective components functionalities, MOEs for the communication system can include its bandwidth and data rate, where bandwidth of a network is given by bits that can be transmitted over the network in a certain period of time (Peterson and Davie, 2003). The reliability of timely monitoring and response operations is also critical. Reliability is affected by environmental (terrain, weather) as well as human factors. For example, though wired communication is typically considered reliable, its communication can completely break down due to physical damage to the vol. 13 no. 3 2009

Alternatives Identication The communication infrastructure for an ITS system can adopt either a centralized or distributed architecture or a combination of both at different scopes. To choose among alternative communication architectures, one must evaluate the advantage and disadvantage for each and balance the tradeoffs between them. A centralized communication infrastructure allocates dedicated bandwidth to connect a central controller with a set of controlled eld devices, which are in general referred to as sensors in this article. A distributed communication infrastructure, on the other hand, makes no distinction among central controllers and eld devices. Any device is connected to nearby peer devices for relay, sharing, and conrmation of their sensing information. Using a trafc camera system as an example, a centralized solution aggregates all trafc surveillance data to one place for centralized processing. In contrast, a distributed solution assigns each device with certain distributed decision-making ability and

intelligent transportation systems

COMMUNICATION ALTERNATIVES

115

wires during construction or adverse weather, and such damages are time consuming to locate and repair. Wireless communication is sensitive to terrain and weather conditions even during its off-peak operation. Communications can occasionally be lost or contain errors in between correct communications. The degree of such errors/loss increases as the adverse conditions worsen, yet at all times, a fraction of communications can be made successfully, which poses a signicant opportunity to enhance the systems reliability under all conditions. The communication system is the most expensive part of a trafc surveillance system (Gordon et al., 1993). The cost of a communication alternative must be justied with respect to its application QoS requirement. Hence, MOEs must also be dened to quantify the relationship between costs and performances, such as throughput per unit cost (e.g., megabytes per dollar), to facilitate the system planning process.

Simulation Study
Figure 2 Architecture of the integrated simulator.

The integrated simulation platform is based on the PARAMICS trafc modeler and the ns-2 network simulator. PARAMICS is a detailed microscopic simulator that provides realistic trafc ow and detector modeling, with an extensive API for plugging in customized control procedures and external interfaces (Quadstone Limited, 2006). ns-2 is an open-source, packetlevel, and event-driven network simulator, allowing modular incorporation of newly developed protocol components and interface with other software (Sobeih et al., 2005). PARAMICS simulates a transportation system using a number of network les that dene all aspects of a transportation system, including its infrastructure geometrics, trafc control methods, ITS components, driver characteristics, and trafc volumes. In PARAMICS, many different driver vehicle units (DVUs) interact with each other and roadway infrastructure as well as trafc control devices in the simulation model to realistically represent the trafc conditions in the real world. Userdened functions are programmable via an API add-on module in PARAMICS and a plug-in C++ source le, with which the integrated simulator implements trafc statistics logging, synchronizes sensor data, and exchanges control commands with ns-2. ns-2 simulates network protocols at packet level among nodes in a specied network topology. Network protocols are modeled with individual source les in C++ and TCL languages. The integrated simulator architecture is illustrated in Figure 2. To initiate a simulation, in PARAMICS, users build, calibrate, and validate a trafc network model, whereas in ns-2, users dene the wireless network protocol stack, the network topology, and the execution time and interval. Figure 3 shows the execution owchart of the interactive process of PARAMICS and ns-2. Basically, the two simulators perform synchronized locked-step executions to model simultaneously the vehicular trafc dynamics and network communications incurred for trafc control operations. At ns-2 side (right-hand side of Figure 3b), user-dened functions, includ-

ing synchronization and data trafc routing and scheduling, can be inserted at specic protocol layer with C++ source les, which enables ns-2 to implement synchronization and message exchange with PARAMICS. At PARAMICS side (left-hand side of Figure 3a), a similar synchronization function is to be dened in the API program, which interact with a PARAMICS modeler or processor module to support sequential running and exchanging data and command with ns-2 domains. The synchronization le formats are customizable for different control semantics.

CASE STUDY To illustrate the efcacy of the framework and associated simulation platform for implementing the framework, a case study on selecting the best communication alternative for ITS system in Greenville, South Carolina, is presented. Alternatives Identication Greenville is one of the largest cities in South Carolina. With three major national freeways, I-85, I-185, and I-385 passing by, one of the four SC transportation management centers (TMC) is located here to enhance the trafc management and operation. The regional ITS architecture in Greenville area requires all the ITS subsystems to be connected and to collaborate with each other over communication links reliably and efciently. All trafc information and command must be transmitted back and forth between the TMC and local ITS equipments. The research team rst started the alternative identication with collecting data, which included highway topography, ITS asset locations, and trafc volumes. The existing ITS equipment at the study site includes trafc cameras, trafc detectors, count vol. 13 no. 3 2009

intelligent transportation systems

116
Start

Y. MA ET AL.
Start

Read synchronization file

Read synchronization file

Is PARAMICS turn ? Yes

No

Is ns2 turn ? Yes

No

Reads detector and controller parameter files

Read detector and controller parameter files

Pause

PARAMICS simulate traffic: Vehicles, roadway and control devices interaction Generate and clear incidents Collecting traffic statistics

Pause

ns2 simulate networking: Determine data rate, destination, routing, and queuing Sending and receiving packets Collecting MOEs

No Is scheduled pause time reached? Yes Record detector and controller collected information Write ns2 into Synchronization file

No Is scheduled pause time reached? Yes Record detector and controller transferred information Write PARAMICS into Synchronization file

No Is scheduled stop time reached? Yes Stop

No Is scheduled stop time reached? Yes Stop

(a)
Figure 3 Integrated simulator process execution owchart.

(b)

stations, a trafc management center, variable message signs (VMS), and highway advisory radio (HAR) stations. The infrastructure information in the database includes the locations of facilities and such attributes as cost, bandwidth, latency, and power requirements, all benecial for analysis and evaluation. Table 1 shows the existing ITS equipment. The obtained ITS infrastructure and communication system information was geo-coded into the GIS software in different layers, each containing the available attributes (location,

Table 1 Example of list assets for communication analysis Asset 1 2 3 4 5 6 7 8 9 10 11 12 Trafc cameras Trafc detector Count station Trafc management center HAR transmitter VMS Trafc signal Drop cabinets DOT owned ber Leased ber Fiber node/hub Coaxial cable Resource SCDOT SCDOT SCDOT SCDOT SCDOT SCDOT SCDOT and operation city/county SCDOT and operation city/county SCDOT Commercial carriers SCDOT SCDOT and commercial carriers

bandwidth, cost, etc.) of a particular type of ITS device. This can help analyze existing and proposed communication alternatives. Figure 4 shows a GIS map with a layout of the existing ITS infrastructure of Greenville. Information such as future hubs is not shown in the GIS map. Considering the choice over two system architectures (i.e., distributed and centralized) and two communication media (i.e., wired and wireless), four alternative communication architectures were studied to support the trafc surveillance and incident management system comprised of trafc cameras and trafc management center. The four alternative architectures considered were wired centralized, wireless centralized, wired distributed, and wireless distributed. MOE Selection To select the MOEs for this study, the ITS application requirements and the various system variables for the four communication alternatives were analyzed. To evaluate both the performance and the cost effectiveness of the alternatives, the study selected three relevant MOEs related to QoS: the peak achievable throughput, the successful delivery ratio, and the throughput per unit cost. Figure 5 illustrates the MOE selection vol. 13 no. 3 2009

intelligent transportation systems

COMMUNICATION ALTERNATIVES

117

Figure 4 GIS map of ITS and communication infrastructure in Greenville, South Carolina.

process. Delay was not selected as an MOE because the magnitude of communication latency is far less than the time magnitude of the trafc surveillance events when the required data rates are below the provisioned capacity of the communication network.

Simulation Study Simulating the Communication Network The I-85 study network corridor in Greenville, South Carolina, consists of approximately eleven miles of freeway and six interchanges. This segment is the major corridor connecting Atlanta, Georgia, and Charlotte, North Carolina. It serves the trafc to and from the Greenville metropolitan area with a population of 601,986 according to the 2006 census estimate. After site selection, the research team used the PARAMICS microscopic trafc simulation software to build, calibrate, and validate the roadway network. Network building began by collecting various data including geometry, trafc control, and trafc volume. The geometric layout data for the roadway network were obtained from the South Carolina Department of Natural Recourses in GIS format. Next, aerial photos from multiple sources and information collected from site visits were used to verify correct geometric conditions, such as number of

lanes, lane widths, lane allocation, and curvature. The specic location of each trafc camera was also added to the network according to the South Carolina Department of Transportation (SCDOT) GIS database. The research team requested and received the vehicle volume and incident data from the SCDOT local planning organizations. The SCDOT provided hourly and average daily trafc count data, trafc signal timing data, and incident location, severity, and duration data. The local planning organizations provided a planning model for use in predicting the origins and destinations matrix of the future network trafc. Other data needs such as speed limits, rights of way, and stripping were met through
Requirements System Variables MOEs

Data Rate Reliability Cost

Area Topography Bandwidth Device Reliability Weather and Environment Device Unit Cost Traffic Condition

Peak Achievable Throughput (Bytes per second) Delivery Ratio (%) Mbytes/Dollar

Figure 5 MOEs.

Relationships between communication requirements and evaluation

intelligent transportation systems

vol. 13 no. 3 2009

118

Y. MA ET AL.

observation during site visits. All of this information was used to build the trafc simulation model in PARAMICS. Calibration and validation steps are critical to ensure that the simulation model reects trafc conditions accurately. Calibration involves face validation of the trafc model animation and comparison of trafc volumes collected from trafc sensors in the eld and virtual loop detectors in the simulation model. As shown in Figure 6, the measured and simulated daily volumes at the twelve freeway and sixteen arterial links did not vary significantly. The freeway links locate on both directions of six I-85 basic segments that are separated by ve major interchanges. The arterial links locate on the roads that have access to I-85. The overall simulated vehicular trafc volumes were within 1 percent of the measured volume, and the highest individual volume error was no more than 10 percent, with most less than 5 percent. Statistical analysis also conrmed that there was no signicant difference in the mean of the measured and simulated volumes for freeway and arterial links. Additionally, the expert opinions from the local trafc management centers staff conrmed that the trafc model was representative of the reality. Validation of the system performance output was carried out by comparing observed travel times and queue length with the simulated ones. The authors collected the observed travel times using the oating car technique (Institute of Transportation Engineers [ITE], 2000) along each direction of the freeway during peak evening hours. The simulated travel times were recorded as average vehicular travel time along the corresponding freeway directions and lengths during the same peak evening hours. Based on the authors knowledge of the site conditions and inputs from local TMC operators, the driver aggressiveness and awareness distribution parameters were adjusted to achieve simulated travel times within 1 percent from the measured ones. This process is summarized in Table 2. Fine adjustments on the target headway and driver reaction time on individual links were also performed when necessary to ensure that the simulated speeds were similar to the observed average travel speed on each link. Furthermore, the simulation model accurately reected the observed queue at the bottleneck segment, which occurred at three signalized off-ramp intersections. Specically, the calibration utilizes the average annual daily trafc obtained from the SCDOT converted to hourly volume according to the typical trafc volume prole of an average weekday. The trafc scenario for this study was PM peak period during an average weekday because the peak trafc ow occurred between 4:30 PM and 6:30 PM at the study site. The matching simulations were started at 4:00 PM and allowed at least half an hour of warm-up time. After the trafc volumes were fully loaded into the network, incidents were generated at random locations and random times between 4:30 PM and 5:00 PM. The ns-2 communication simulator implemented the T1 data links with a bandwidth of 1.544 MHz as the medium for wired centralized and distributed alternatives. For the wireless centralized system, the authors assumed that trafc surveillance operating agencies will lease the CDMA2000 data links with a

bandwidth of 1.25 MHz. For the wireless distributed alternative, IEEE 802.11b protocol with a bandwidth of 11 MHz is assumed for communication among sensors and controllers in the eld. The study considered trafc surveillance data generated at constant bit rate and sent across the network using the user datagram protocol (UDP). Different data rates were simulated to examine the capacity of the four alternatives. For the study scenario with incidents, the vehicular trafc simulator randomly generated incidents on the segments under surveillance of trafc cameras during the AM peak hours through PARAMICS Programmers API interface. The API programs collect network running parameters, such as sensor measurements, from PARAMICS modeler or processor; execute various dened functionality using these data; and then convey commands to modeler or processor to affect the trafc network. Particularly, the developed API program selected various incident occurrence times, locations, and severities accounting for the effects of different incident scenarios. PARAMICS also determines the duration of incidents through the realistic simulation of interaction between DVUs including the vehicles involved in incidents and the vehicles in the queue. The duration of incidents, which is dened as the time between incident occurrence and the return to normal trafc condition, directly affects the communication cost in terms of data rate, which can be altered by the ns-2 during the simulation. In a centralized system, each device continuously generates constant rate data at the rate of 384 Kbps no matter whether there is an incident or not. On the contrary, in a distributed system, the devices send stationary images with a consequent date rate of 24 Kbps to the controlling center at a low frequency during the normal condition. Once an incident is identied or suspected, the corresponding trafc camera transmits full motion videos with a data rate of 384 Kbps, which is the same as the constant data rate of the centralized system, to control center. Within the two-hour simulation period, throughputs of the centralized and distributed systems for various incident durations were examined to compare their communication costs. Simulation was performed for multiple runs with random vehicle types, release rates, and driver behaviors. As a signicant measure of effectiveness in each domain, vehicular delay and data throughput were selected as the indicators to determine the number of simulation runs. The experiments for each scenario were repeated until the mean and variance of both vehicular delay and data throughput satised the following inequality (Bartin et al., 2006; Law and Kelton, 2000; Ozbay, Yasar, and Kachroo, 2004): (n, )/ X(n) /(1 ) where n denotes the number of simulation runs; denotes the required Type I error, which was assumed to be 0.05; (n, ) = tn1,1/2 2 (n)/n denotes the half-length of the condence interval depending on the variance of n simulation runs 2 (n); X (n) denotes the mean of n simulation runs; and is the required relative error of X (n), which was assumed to be 0.05 in this study. vol. 13 no. 3 2009

intelligent transportation systems

COMMUNICATION ALTERNATIVES

119

Figure 6 Comparison of measured and simulated daily volume in the trafc model.

MOE Generation Figure 7 through 13 present the generated MOEs for the communication alternatives. Figure 7 shows the throughput and delivery ratio in percentage for the wired centralized network at various data rates. The simulated centralized network ensures each camera with a bandwidth of 1.544 MHz no matter how many cameras are deployed per mile. The capacity of the T1

centralized network is 1546 Kbps. The delivery ratios in percentage for data rates under the capacity were 100 percent. As the data rates approached the capacity, an increasing number of packets were dropped. The delivery ratio decreased dramatically and the throughput remained 1546 Kbps. Figure 8 shows the throughput and delivery ratio in percentage for the wired distributed network at different data rates. For trafc surveillance system with a density of one camera vol. 13 no. 3 2009

intelligent transportation systems

120
Table 2 Process of calibration driver behavior distribution Iteration Aggression distribution Awareness distribution I-85 NB travel time (sec) Measured Simulated Percentage difference I-85 SB travel time (sec) Measured Simulated Percentage difference
0

Y. MA ET AL.

1 0 0

2 2 2

3 3 3

4 4 4

5 3 4

6 Customized Customized 529 522 1.30 532 528 0.80

529 529 529 529 529 745 612 576 502 557 40.80 15.70 8.90 5.10 5.30 532 532 532 532 532 738 633 570 499 551 38.70 19.00 7.10 6.20 3.60

through 4 represent the software preset distribution of driver aggression and awareness.

per mile, the T1 distributed network was able achieve a maximum throughput of 512 Kbps at the capacity data rate of 512 Kbps. When the data rate was over the capacity, the delivery ratio in percentage decreased from 100 to 43 percent with data rate of 512 Kbps and 1216 Kbps, respectively, and the throughput slightly increased up to 520 Kpbs when the data rate was over 960 Kbps. Similarly, the wired distributed network reached maximum throughput at capacity data rate of 256 Kbps when the camera density increased to two cameras per mile. Figure 9 describes the throughput and delivery ratio in percentage for the wireless centralized network, revealing a similar trend as that of the wired centralized network. However, a smaller maximum throughput of 1252 Kbps was obtained due to its assumed 1.25-MHz link bandwidth. Figure 10 shows the throughput and delivery ratio in percentage for the wireless distributed at different data rate. For trafc surveillance system with a density of one camera per mile; although the system was able to achieve a maximum throughput
1600

of 435 Kbps, the delivery ratio in percentage began to drop from 100 percent when the data rate was 384 Kbps. Therefore, when the rate is over 384 Kbps, the surveillance system might encounter considerable delay and jitter. Therefore, the capacity rate for this scenario was 384 Kbps. When the camera density increased to two cameras per mile, the system could only support 128 Kbps data rate, which is lower than the typical full motion video data rate of typical trafc cameras. To enhance the capacity, one solution is to increase the wireless link bandwidth from 11 MHz to 54 MHz, which is compatible with the IEEE 802.11g standard. Another solution is to partition the sensors into smaller groups and have each group communicating with a different radio channel, such that bandwidth contention within each group is reduced. A feature of the wireless distributed system different from other alternatives is that its throughput decreased after reaching a peak value due to the increased random transmission collision at high per device transmission rates. For cost analysis, installation and operation costs for the different communication infrastructures have been estimated with best effort according to vendor advertisements and are summarized in Table 3. The cost information was obtained from various resources, including the FHWA ITS online database (USDOT, 2007a), communication device vendors, and communication service carriers, and was converted into 2007 dollars using an ination rate of 3 percent. These costs were then summed to account for the amount of infrastructure in a given network architecture. For example, the cost of communication alternative for a wireless centralized communication alternative with trafc cameras at half-mile intervals is the sum of communication element unit cost, installation cost, and monthly service fee. In this case, the average cost of a modem and corresponding installation fee are $300 and $100, respectively. Considering a ten-year modem lifetime and an ination rate of 3 percent, the
110%

1400

100%

Throughput(Kbps)

1200

90%

1000

80%

800

70%

600 600

800

1000

1200

1400

1600

1800

2000

60% 2200

Data Rate Per Camera (Kbps)

Throughput Delivery Ratio

Figure 7 Throughput and delivery ratio for wired centralized network at different data rates.

intelligent transportation systems

vol. 13 no. 3 2009

Delivery Ratio

COMMUNICATION ALTERNATIVES
Table 3 Cost estimate in 2007 dollars of the communication infrastructure Unit Cost ($) Element Wired distributed/centralized Fiber optical cable Transmitter Optical regenerator Optical receiver Wireless access Monthly service Modem Lifetime (years) 20 10 10 10 20 10 Low High Installation Cost ($) Low 5000 150 High

121

O&M Cost ($/year) 1000 2000

Wireless distributed Wireless centralized

172/100 ft 1000 4000 Optional 800 1200 400 1000 60 140 200 400

15000 250 Optional 150 250 200 900 50 150

1000

600

120%

500

100%

Throughput (Kbps)

300

60%

200

40%

100

20%

0 0 200 400 600 800 1000 1200

0% 1400

Datat Rate Per Camera (Kbps)

Throughput of 1 camera per 1.5 mi Delivery Ratio of 1 camera per 1.5 mi Throughput of 1 camera per 0.6 mi Delivery Ratio of 1 camera per 0.6 mi

Figure 8 Throughput and delivery ratio for wired distributed network at different data rates.
110% 1350 100% 1200

Throughput (Kbps)

90% 1050 80%

900 70%

750

60%

600 700

900

1100

1300

1500

1700

1900

50% 2100

Data Rate Per Camera (Kbps)

Throughput Delivery Ratio

Figure 9 Throughput and delivery ratio for wireless centralized network at different data rates.

intelligent transportation systems

vol. 13 no. 3 2009

Delivery Ratio

Delivery Ratio

400

80%

122
500

Y. MA ET AL.
120%

100% 400

Throughput (Kbps)

80% 300 60% 200 40%

100 20%

0 0 100 200 300 400 500 600 700

0% 800

Data Rate Per Camera (Kbps)

Throughput of 1 camera per 1.5 mi Delivery Ratio of 1 camera per 1.5 mi Throughput of 1 camera per 0.6 mi Delivery Ratio of 1 camera per 0.6 mi

Figure 10 Throughput and delivery ratio for wireless distributed network at different data rates.

yearly cost of a modem is $46.89. Assume the monthly wireless service fee is $100/unit. The overall yearly cost of communication infrastructure for twenty trafc cameras that cover ten miles of highway is (46.89 + 100 * 12) * 20 = 24,937.80 dollars. Moreover, the yearly cost for each network element can be further converted to value in dollar per operation second, based on the operation schedule of specic trafc surveillance system. To assess the cost effectiveness of different alternatives, their throughput-to-cost ratios were computed and shown in Figure 11. The wired alternatives were less cost effective than the wireless ones; for a camera density of one camera per one and a half miles, the wireless centralized and distributed has comparable cost effectiveness for camera rates up to 480 Kbps, where the distributed throughput saturates. With the centralized solution, per camera rates can be as high as 1400 Kbps. The wireless distributed network, however, increases its cost effectiveness as the density of supported devices increases. The per camera capacity reaches inection point at a capacity of 480 Kbps. Compared to the wireless alternatives, the per camera rates in a wired centralized network can be 1700 Kbps, whereas the same rates reach the maximum capacity of 512 Kbps in a wired distributed network. Figure 12 shows the throughput-to-cost ratios for different camera densities for the wireless distributed system. The ratio increases substantially when the density increases and the optimal density depends upon the desired camera data rate. If the expected data rate is low and the density is high, a wireless distributed solution will be economically more preferable. The throughput-to-cost ratio will also increase if higher rates with IEEE 802.11g are achievable in the eld. As shown in Figure 13, the centralized and distributed wireless system demonstrated signicant differences in throughput

during the simulated two-hour peak vehicular ow period. The solid line that represents the centralized system maintains a high throughput of 384 Kbps throughout the entire period. On the other hand, the distributed system can maintain a much lower throughput by transmitting images at a low frequency with a data rate of 64 Kbps during the normal vehicular trafc condition. Once an incident occurs, the data rate of the trafc cameras that have detected the incident increases to 122 Kbps. As the incident duration increased, the throughput of distributed network increased linearly, because the incident duration determined the portion of time that needs high data rate.

Summary of the Case Study The results from the case study showed that, given the reliable performance for wireless communication links under normal weather and site conditions, the wired alternatives are less cost effective than the wireless ones in both centralized and distributed communication topologies. For the particular camera density (one camera per 1.5 miles), the wireless centralized and distributed topologies have comparable cost effectiveness before per camera rate reaches up to saturation throughput. The wireless distributed network, however, increases its cost effectiveness as the density of supported devices increases. Throughput-to-cost ratio for different camera densities increases substantially when the density increases, and the optimal density depends upon the desired camera data rate. A wireless distributed solution will be economically more preferable if the expected data rate is moderate and the camera density is high. Wireless networks, especially wireless distributed networks, provide a certain level of reliability during normal weather vol. 13 no. 3 2009

intelligent transportation systems

COMMUNICATION ALTERNATIVES
2800

123

Throughput Cost Ratio (MagaBytes/Dollar)

Total Cost = Data Rate Per Camera * Number of Cameras / Throughput Cost Ratio

2400

Increase until reach capacity of 1400Kbps

2000

1600

Reach inflection point at capacity of 480Kbps Reach maximum at capacity of 512Kbps

1200

800

Increase until reach capacity of 1700Kbps

400

0 0 100 200 300 400 500 600 700 800 900 1000

Data Rate Per Camera (Kbps)

Wired Distributed Network Wireless Distributed Network Wired Centralized Network Wireless Centralized Network

Figure 11 Throughput-to-cost ratios of different network architectures.

conditions; the performance degrades during adverse weather but is not expected to cease operation completely during disaster scenarios. On the other hand, wired networks provide substantially more than adequate performance during normal weather conditions but are susceptible to total blackout during adverse situations. It is the exact purpose of the proposed framework to aid decision-makers to construct such different scenarios using the simulation framework to be able to quantify the performance

metrics under different conditions to aid in their choice of proper network type, topology, density, and capabilities. For decision-makers to select the best communication methods for a given location and application, the results of an economic analysis should accompany technical results. Further, the ranking and selection of public infrastructure projects based on economic analysis and the MOEs related to QoS identied in this study would allow a user-friendly comparison between

1600

Throughput Cost Ratio (MegaBytes/Dollar)

1400

1200

1000

800

600

400

200

0 0 100 200 300 400 500 600 700 800

Data Rate Per Camera (Kbps)

1 Camera Every 1.5 Miles 1 Camera Every 0.6 Mile 1 Camera Every 1 Mile 1 Camera Every 0.3 Mile

Figure 12 Throughput-to-cost ratios of wireless distributed network.

intelligent transportation systems

vol. 13 no. 3 2009

124
400

Y. MA ET AL.

350

300

Throughput (Kbps)

250

200

Wireless Distributed Wireless Centralized

150

100

50

0 0 10 20 30 40 50 60

Incident Duration (Min)

Figure 13 Throughput of centralized and distributed networks during an incident.

communications installations and upgrades and other transportation infrastructure projects. Economic analysis methods such as benet-cost analysis or multi-criteria decision analysis can also be applied to MOEs using the previous steps presented in the framework. Cost data can be found within the IDAS database, the ITS Cost Database maintained by the FHWA (USDOT 2007a), or from retailers websites (CDW, 2008). All costs should be adjusted to the same base year and should include installation, maintenance, capital, and labor costs for all MOEs.

CONCLUSIONS This study developed a framework that summarized the evaluation procedures for communication alternatives using the simulation platform that integrated trafc and communication simulators, so that communication system performance under different vehicular trafc demand and conditions can be efciently analyzed. This article also described the identication of communication topologies and important MOEs as the performance indicators of various communication alternatives. The framework and the associated integrated simulator provide a systematic procedure and tool, respectively, to facilitate objective evaluation for ITS communication infrastructures. Additionally, this article presented a case study with surveillance cameras to demonstrate that the results generated through the application of the framework can help in deciding cost-effective communication alternatives for future ITS deployment. This is important today and in the future as public agencies are operating under budget constraints. The proposed framework provides the means to support a systematic study of the performance and cost tradeoff given the

known deployment criteria. Though based on the case studys conclusion, we can expect that the wireless distributed solution can be more cost effective if the bandwidth requirement meets the application demand. We by no means claim that wireless solutions are always superior to wired solutions. If regional policy and business considerations such as infrastructure and right-ofway sharing are factored in to substantially reduce the cost of wired infrastructure, then the wired infrastructure certainly provides much higher bandwidth capability and reliability (except in disaster conditions). The proposed framework is specically aimed to facilitate such a decision by evaluating whether a higher bandwidth infrastructure is indeed necessary, cost efcient, and meeting ITS application requirements of sustained operation under disaster situations. The integrated simulation platform presented in this article can be applied to evaluate a wide range of ITS applications involving communications, such as vehicle-to-vehicle and vehicle-to-infrastructure communications. To further study the reliability of such systems, environmental and human factors that impact communication performances must be included in the model. The integrated simulation platform has built-in capabilities to simulate communication errors of different patterns and extent. In future research, realistic communication error models should be measured and integrated into the platform for studying the systems reliability-related MOEs.

REFERENCES
Avila, A., Korkmaz, G., Liu, Y., Teh, H., Ekici, E., Ozguner, F., Ozguner, U., Redmill, K., Takeshita, O., Tokuda, K., Hamaguchi, M., Nakabayashi, S., and Tsutsui, H. (2005). A complete simulator architecture for inter-vehicle communication

intelligent transportation systems

vol. 13 no. 3 2009

COMMUNICATION ALTERNATIVES intersection warning systems. Proceedings of IEEE Intelligent Transportation Systems, Vienna, Austria. 461466. Bartin, B., Ozbay, K., Mudiganda, S., Yanmaz-Tuzel, O., and List, G. (2006). Modeling and simulation of unconventional trafc circles. Transportation Research Record, 1965, 201209. Boxill, S. A., and Yu, L. (2000). An Evaluation of Trafc Simulator Models for Supporting ITS Development. Houston, TX: Center for Transportation Training and Research, Texas Southern University. Brydia, R. E., Brackin, B. E., Johnson, J. D., Thomas, G. B., and Balke, K. N. (2005). Development of a Wireline Communications Design Guidebook for Intelligent Transportation Systems. Austin, TX: Texas Department of Transportation. Cai, H., and Lin, Y. (2005). Design of a roadside seamless wireless communication system for intelligent highway. Conference on IEEE Networking, Sensing and Control, Tucson, AZ, 342 347. CDW Corp. (2008). Proxim orinoco outdoor devices. Available at: http://www.cdw.com/shop/search/results.aspx?key=proxim+ orinoco+outdoor&platform=All&sr=1 (accessed May 2008). Choffnes, D. R., and Bustamante, F. E. (2005). An integrated mobility and trafc model for vehicular wireless networks. Proceedings of the Second ACM International Workshop on Vehicular Ad Hoc Networks (VANET), Cologne, Germany. 6978. Chowdhury, M., Fries, R., Ma, Y. Dunning, A., Hamlin, C., Ozbay, K., Ogle, J., and Sarasua, W. (2007). Benet cost Analysis of Accelerated Incident ClearanceFinal Report. Prepared for South Carolina Department of Transportation. Columbia, SC: Clemson University. City of Cape Town, South Africa. (2005). Trafc signal services. Available at: http://www.capetown.gov.za/atrams (accessed). Coifman, B., and Ramachandran, M. (2004). Distributed surveillance on freeways with an emphasis on incident detection. Conference on IEEE Intelligent Transportation Systems, Washington, DC, 773 778. Fitzgibbons, B., Fujimoto, R., Guensler, R., Hunter, M., Park, A., and Wu, H. (2004). Distributed simulation test bed for intelligent transportation systems design and analysis. Annual National Conference on Digital Government Research, Seattle, WA. 12. Fujimura, K., and Hasegawa, T. (2005). Performance evaluation of the MAC protocol for integrated inter-vehicle and road to vehicle. Conference on IEEE Intelligent Transportation Systems, Vienna, Austria, 308313. Ghaman, R. S., Zhang, L., Mchale, G., and Stallard, C. (2003). The role of trafc simulation in trafc signal control system development. Conference on IEEE Intelligent Transportation Systems, Piscataway, NJ, 1, 872877. Gordon, R., Reiss, R. A., Dunn, W. M., and Morehead, D. R. (1993). Communications Handbook for Trafc Control Systems. U.S. Department of Transportation, Washington, D.C. Gordon, R., and Tighe W. (2005). Trafc Control Systems Handbook. U.S. Department of Transportation, Washington, D.C. Habesch, N. O., Jehanian, K. M., and Ziemer, K. D. (1998). An evaluation of Philadelphias satellite communications demonstration project: The mobile platform. The 68th Institute of Transportation Engineers Annual Meeting, Toronto, Canada. Hsin, J. K., and Wang, T. R. (1992). Modeling concepts for intelligent vehicle highway systems (IVHS) applications. The 24th Conference on Winter Simulation, Arlington, VA. 1201 1209. Hunsucker, D. Q. (2002). Evaluation of 220 MHz Frequencies for ITS Experimentation. Kentucky Transportation Center, Lexington, KY.

125

IEEE. (2003). IEEE 802. 11g 2003: Further high data rate extension in the 2.4g GHz band. Available at: http://standards.ieee.org/ getieee802/download/802.11g2003.pdf Institute of Transportation Engineers. (2000). Manual of Transportation Engineering Studies. Institute of Transportation Engineers, Washington, D.C. Jones, W. S. (2002). Broadband Wireless: Is It an Option for Your ITS. Washington, DC: U.S. Department of Transportation. Klein, A. L., Mills, K. M., and Gibson, R. P. D. (2006). Trafc Detector Handbook: Version 3. U.S. Department of Transportation, Washington, D.C. Law, A. M., and Kelton, W. D. (2000). Simulation Modeling and Analysis (3rd ed.). McGraw-Hill, New York. Leader, S. (2004). Telecommunications Handbook for Transportation Professionals. U.S. Department of Transportation, Washington, D.C. Mastako, K. A., and Klein, L. A. (1999). Evaluation of mobile surveillance and wireless communication systems: Field operational test. Transportation Research Record, 1683, 1421. Mirchandani, P., and Head, L. (1998). RHODES: A real-time trafc signal control system: Architecture, algorithms, and analysis. Available at: http://www.sie.arizona.edu/ATLAS/docs/TRISTANIII.pdf (accessed). Ozbay, K., Yasar, I., and Kachroo, P. (2004). Modeling and PARAMICS based evaluation of new local freeway ramp metering strategy that takes into account ramp queues. Transportation Research Record, 1867, 8997. Peterson, L. L., and Davie, B. S. (2003). Computer Networks: A System Approach (3rd ed.). San Francisco, CA: Morgan Kaufmann. Quadstone Limited. (2006). Paramics V5.0 Modeler Users Guide. Edinburgh: Quadstone Limited. Road and Trafc Authority, New South Wales, Australia. (2006). Sydney coordinated adaptive trafc system (SCATS). Available at: http://www.rta.nsw.gov.au/trafcinformation/trafcfacilities/scats/ (accessed). Siemens. (2006). Principles of SCOOT. Available at: http://www. itssiemens.com/en/t nav224.html (accessed). Smith, B. L., Zhang, H., Fontaine, M. D., and Green, M. W. (2004). Wireless location technology-based trafc monitoring: Critical assessment and evaluation of an early generation system. Journal of Transportation Engineering, 130(5), 577584. Sobeih, A., Chen, W. P., Hou, J. C., and Kung, L. C. (2005). J-Sim: A simulation environment for wireless sensor networks. Proceedings of the 38th annual meeting on Simulation Symposium, San Diego, CA. 175187. Stephanedes, Y. J., Douligeris, C., and Takaba, S. (1996). Communications for the intelligent transportation system. IEEE Communications Magazine, 34(10), 2430. Sukthankar, R., Hancock, J., Pomerleau, D., and Thorpe, C. (1996). A simulation and design system for tactical driving algorithms. Conference on AI, Simulation and Planning in High Autonomy Systems, La Jolla, CA. Tokuyama, H. (1996). Intelligent transportation systems in Japan. Available at: http://www.tfhrc.gov/pubrds/fall96/p96au41.htm (accessed). Tyco Integrated Systems. (2006). Trafc managementSCATS. Available at: http://www.trafc-tech.com/pdf/scatsbrochure.pdf (accessed). U.S. Department of Transportation. (2006). Intelligent transport systems: Technology overview. Available at: http://itsdeployment.ornl. gov/technology overview/ (accessed).

intelligent transportation systems

vol. 13 no. 3 2009

126

Y. MA ET AL. Yang, Q. Y., Sisiopiku, V., Arnold, J. A., Pisano, P., and Nelson, G. G. (2000). Assessment of rural intelligent transportation system wireless communications solutions. Transportation Research Record, 1739, 5158. Yin, J., ElBatt, T., Yeung, G., Ryu, B., Habermas, S., Krishnan, H., and Talty, T. (2004). Performance evaluation of safety applications over DSRC vehicular ad hoc networks. Proceedings of the First ACM International Workshop on Vehicular Ad Hoc Networks, 19. Philadelphia.

U.S. Department of Transportation. (2007a). ITS costs database. Available at: http://www.itscosts.its.dot.gov/its/benecost.nsf/ AdjustedUnitCosts (accessed). U.S. Department of Transportation. (2007b). National ITS Architecture, Version 6.0. Available at: http://www.iteris.com/itsarch/ (accessed). Verhyen, J. F. (2005). State Patrol Mobile Data Communications Network: Final Report and Local Evaluation. Madison, WI: Wisconsin Department of Transportation, ITS-2003 (629).

intelligent transportation systems

vol. 13 no. 3 2009