A collective neurodynamic approach for solving distributed system optimum dynamic traffic assignment problems

Abstract

In this paper, the traditional system optimum dynamic traffic assignment (SO-DTA) problem is solved by distributed multi-agent dynamics. The goal of SO-DTA is to optimally control the route choice of each user to minimize the total travel time by all users over the assignment time period. It is beneficial for reducing the congestion of the traffic network with time-varying demand. Different from the traditional SO-DTA which is formulated and solved in a centralized scheme, we aim at solving it from a multi-agent perspective in a communication network. Based on the cell transmission model, two connector-based relaxed SO-DTA models are provided in a multi-agent system framework for the traffic network with a single destination and multiple destinations, respectively. In the provided models, each cell connector is treated as an agent, which could exchange information with its adjacent connectors to accomplish the system optimization objective. Then, a collective neurodynamic system equipped with the proposed distributed protocol is used to solve general network optimization problems including the above SO-DTA models as special cases. The convergence analysis is further given for the proposed algorithm. Numerical studies over various traffic networks are presented to show the effectiveness of the proposed method.