Computational aspects in traffic simulation problems

We discuss topics related to the notion of shortest paths in large street networks. In the first part. we discuss the dynamic traffic assignment problem and describe an algorithm designed to determine user equilibria in simulation-based traffic models. The proposed algorithm is able to find in a reasonable amount of time equilibria in problems that are not treatable with conventional methods (e.g. Frank-Wolfe algorithm). The second part deals with an improved algorithm to find shortest paths in street networks with given link cost functions. The main idea is to reduce the network using the intrinsic properties of street networks and to perform a (backward) Dijkstra-algorithm on this reduced network. This approach results in a highly accurate algorithm which is faster than the straightforward approach by a factor of 4. (C) 2001 Elsevier Science B.V. All rights reserved.