A Comparative Study of the Cross Entropy Approach with the State–of-the-art Simulation-based Traffic Assignment Algorithms

Abstract

This paper presents a path-based cross entropy algorithm for solving simulation-based dynamic traffic assignment problem. The performance of the cross entropy algorithm is compared with two state-of-the-art algorithms: method of successive averages and gap function based projection algorithm. The dynamic network loading model is based on a mesoscopic queue model complying with generic first order macroscopic node model. The computational study implemented on four realistic networks shows the cross entropy method provides satisfactory convergence accuracy to user equilibrium.