While advanced technologies made it possible to develop more sophisticated traffic management strategies, experience has shown that such strategies do not always result in improved performance [Gartner et al. (1995)]. When designing a particular traffic management system, many possible architectures and alternatives may be chosen. Evaluation is, therefore, an important element in the design process. Similarly, in assessing the impact of proposed changes to an existing system, the analyst faces various ``what if'' questions.
If a traffic management system is already fully operational, field operational tests can be used in assessing its performance and evaluating many of the practical aspects of systems. However, field tests tend to be expensive, and as a result, typically few alternatives can be tested. In addition, the test results depend on uncontrollable elements of the environment (e.g. weather conditions, travel demand, incidents).
In recent years simulation has become a tool commonly used to understand the characteristics of a traffic system and to select an appropriate design. However, most of these simulation models are not capable of evaluating integrated dynamic traffic management systems and other ITS applications. The need for a more advanced evaluation tool has been identified by a number of researchers such as [Lin (1993), Santiago and Kanaan (1993)]. Recently various approaches have been proposed and models are under development [Leonard II (1993), Jayakrishnan et al. (1995), Jayakrishnan and Rindt (1996)]. The needed ability to date is Unfortunately still lacking in the state of the art, particularly for integrated evaluation of ATIS and ATMS. This problem stems in part from the complex nature of modeling traffic flow in an integrated network, and in part from the lack of basic research concerning the dynamic interaction between traffic flows and ATIS/ATMS.
Figure 1: A computer-based simulation laboratory
for evaluating dynamic traffic management systems
In this research, a methodology for evaluation of dynamic traffic management systems is developed. The methodology utilizes a simulation laboratory (SIMLAB) for testing and evaluating designs of ATMS and ATIS. SIMLAB is a computer-based modeling system that integrates a microscopic traffic simulator (MITSIM) and a traffic management simulator (TMS) (see Figure 1). MITSIM simulates in detail the ``state of the network''. Vehicles in the network are moved from their origins to destinations and respond to the various traffic controls and guidance while interacting with each other. The vehicle movements are recorded by a surveillance system module that represents traffic sensors and probe vehicles. TMS models the candidate traffic control and routing logic under evaluation. Using the data obtained by the surveillance system, TMS generates control and routing strategies, which, in turn, are fed into MITSIM through the traffic control and route guidance devices and hence influence the traffic flows in the simulated network. The core of SIMLAB is modeling the dynamic interaction between the traffic management system (simulated by TMS) and the traffic flows (simulated by MITSIM). This interaction mimics the relationship between a traffic operation control center and the traffic flows in the road network.
The design features a separation of the simulation of surveillance and control system from the simulation of traffic flows. This modular structure provides flexibility for representing a variety of control and route guidance systems and facilitates distribution of tasks to multiple computer processors. In addition, a prototypical system that generates route guidance based on predicted traffic conditions is implemented in TMS and is evaluated as an application of the simulation laboratory.
The main contribution of this research lies in the development of the laboratory environment for evaluating dynamic traffic management systems. This simulation laboratory supports integrated ATIS and ATMS operations in general networks including freeways, arterial, and urban streets. One of its unique features is the use of a mesoscopic traffic simulation model for traffic prediction inside TMS, which provides a fundamental function necessary for evaluation of proactive dynamic traffic management systems. In addition, this default traffic predictor can be replace by other appropriate traffic predictors such as dynamic traffic assignment models. The modeling system developed in this research is the first of its kind that has been implemented and whose usage has been demonstrated though a case study using a real network.