New Algorithmic Approaches for the Anticipatory Route Guidance Generation Problem

The emergence of Intelligent Transportation Systems (ITS) has considerably changed the field of transportation modeling during the past ten years. By Intelligent Transportation Systems (ITS) we refer to transportation systems which apply emerging information technologies to alleviate congestion problems (see, for example, Transportation Research Board 1999). They combine advanced surveillance systems collecting real-time traffic data, Advanced Traffic Management Systems (ATMS) and Advanced Traveler Information Systems (ATIS) in order to achieve that objective. The complexity of these systems requires the development of sophisticated tools for their optimal explo itation. Among the many ITS applications, providing real-time travel information is a particularly i m-portant one. It consists in using available data from the surveillance system in order to generate information that will be disseminated by the ATIS (variable message signs, highway advisory radio on standard broadcast bands, on-board GPS-based computers, etc.) The messages may contain complete or partial route recommendations, qualitative or quantitative information about traffic conditions over the network. We are interested here in generating consistent anticipatory route guidance (CARG). Anticipa-tory guidance accounts for the probable evolution of traffic conditions over time and throughout the network. By consistent, we mean that the anticipated traffic conditions used to generate the guidance must be similar to the traffic conditions that the travelers are going to experience on the network. The problem is non trivial because, contrarily to weather forecast where the real system under consideration is not affected by information provision, the very fact of providing travel information may modify the future traffic conditions and, therefore, invalidate the prediction that has been used to generate it. Consequently, travelers response to information must be explicitly captured in the CARG generation process. If not, overreaction phenomena may be observed (Ben-Akiva, de Palma & Kaysi 1996). and the reliability of the information system will be affected. and Yang (1999) and Bottom (2000) describe the CARG problem as a fixed point problem. In practice the computation of this fixed point is computationally expensive, because it involves complex simulation tools. Consequently classical algorithms solving fixed point problems, for example based on the Banach contraction principle, are inadequate, particularly for their use for real-time traffic management systems. Bottom (2000) has tested the most efficient fixed point algorithms on the CARG problem illustrating their slow behavior. Our idea is to consider this problem as a large scale optimization problem without derivative, exploiting the specific characteristics of the CARG problem. Due …