A path-based cross entropy approach for solving dynamic multimodal transit assignment problem

This paper proposes an effective-path-based cross entropy multiagent approach for solving the dynamic simulation-based transit assignment problem. A multiagent-based simulation model is proposed for dynamic transit system simulation. Based on the minimization of expected travel cost, the passengers learn iteratively optimal departure time and path choice towards user equilibrium. For the transit network modeling, a multilevel transit network structure is proposed for the dynamic assignment of passenger flow. The passengers are assumed to use a set of attractive paths to destinations generated by a modified k-shortest path algorithm. The proposed solution algorithm is implemented on the Sioux Falls network with transit lines. The results show that the proposed approach can quickly find the approximates of dynamic user equilibrium in dynamic congestion transit network.