Dynamic traffic modelling and dynamic stochastic user equilibrium assignment for general road networks

This paper investigates the requirements of dynamic traffic modelling, and proposes the deterministic queuing model as a plausible link performance function to describe the relationship between inflows, outflows, and link travel costs in time-varying condition. Then, it explains how we can perform logit-based stochastic network loading for general road networks in the dynamic case. In particular, this paper shows how to perform dynamic stochastic network loadings for many-to-many origin-destination pairs, and what should be considered to maintain correct flow propagation in the network loading process. Next, this paper shows how the stochastic dynamic user equilibrium (SDUE) assignment problem can be solved without direct evaluation of the objective function. For this purpose, a quadratic interpolation, the method of successive average, and the pure network loading method are adopted at the line-search step in the solution algorithm. Numerical examples show that the present SDUE assignment model with a quadratic interpolation gives rise to a convergent solution with good quality whilst needing less computation time. Furthermore, it is found that the predictive cost-flow association (Proceedings of the European Transport Conferences, Seminar F, P434, 1999, p. 79) is preferable to the reactive one because the former can produce consistent assignment patterns regardless of the size of dispersion parameter theta in the logit model for route choice. (C) 2002 Elsevier Science Ltd. All rights reserved.