Comparison of path flow reassignment ethods for dynamic user equilibrium traffix assignment based on mesoscopic simulation

Abstract

The iterative process proposed to solve the dynamic user equilibrium traffic assignment has two fundamental components: the dynamic network loading and the flow reassignment method. The reassignment component takes much less time and computational resources than the dynamic network loading; however, it has more direct influence to achieve the dynamic user equilibrium conditions. For this reason this paper is mainly concerned with path flow reassignment and how this affects solving dynamic user equilibrium. To investigate the performance of the reassignment, limitations of the method of successive averages are analyzed, paying special attention to the indiscriminately diversion of flow from used paths to the new found best path, and how this affects the convergence of later departure time intervals. Consequently, a new method of successive averages is developed including the improvements suggested. A method for paths flow reassignment that has been used in the literature, is compared with the new proposal embedding them in a dynamic traffic assignment scheme. For this purpose it is essential to incorporate a dynamic network loading model in order to complete the scheme: we have used the mesoscopic traffic simulation model included in Mezzo. The evaluation of these alternative flow reassignment algorithms has been computationally tested in the network of Södermalm (Stockholm). The results obtained show that, the proposed modification of the method of successive averages improves the observed bad convergence for latter departure time intervals.