Predictive or reactive dynamic assignment?
LEMOINE ; MANGEAS ; HAJ-SALEM
Type de document
ARTICLE A COMITE DE LECTURE NON REPERTORIE DANS BDI (ACLN)
Langue
anglais
Auteur
LEMOINE ; MANGEAS ; HAJ-SALEM
Résumé / Abstract
Advanced traveller information systems (ATIS) and advanced traffic management systems (ATMS) have the potential to contribute to the solution of the traffic congestion problem. It is important to be able to study the impact these systems have in traffic condition as well as in driver's choices. In order do this, some dynamic traffic assignment (DTA) algorithms have been proposed in the literature. These assignment algorithms can be classified into two categories : reactive assignment and predictive assignment. The first one uses instantaneous travel times and can be solved using the optimal control discrete-time theory or by a feedback control. The second one uses experience travel time calculated by 'floating vehicle's method'. It considers time varying conditions all along the path and the impact of future traffic. It is solved by heuristic methods as the method of successive averages. This paper presents the advantages and disadvantages of these two approaches using a second order macroscopic traffic models. The interactions between links and spillback congestion are taking into account. Results of these different methods are presented for a southern part of the Paris highway. At the first stage, the underlying traffic model is described and the assignment algorithms presented. Then, for the test network, the travel times between each origin and each destination are compared to the minimum travel time for the two methods. Convergence of the two algorithms is also evaluated. Finally the paper shows that predictive approach is closer to a wardrop's like assignment and more appropriate for planning purposes. Even though no mathematical proof of convergence is given, the algorithms properly converge.
