Perturbation Analysis Approach to the Evaluation of Simulated Path Travel Time Marginals: Application to System Optimal Dynamic Network Assignment

摘要

Algorithms for finding time-dependent system optimal flow patterns in a network require theevaluation of marginal travel times/costs, which is difficult to do in a dynamic context, especiallywhen a traffic simulator is used to capture flow propagation for network loading. Trafficsimulators that track vehicle movement through the network produce trajectories that define theanatomy of vehicular flow movement in the network. The present paper builds on the microlevel analysis of the anatomy of flow movement to evaluate marginals that can guide the searchfor system optimal traffic strategies. A perturbation analysis method designed for a generaldiscrete event traffic simulator is introduced in order to quantify the sensitivity of the path traveltimes to the path flows. The method is applicable as long as trajectories are available. Thisframework relies on the general rules of traffic flow such as moving forward along a path, andsharing common sources of capacity at the merge points. This general technique is implementedin connection with a specific mesoscopic traffic simulator. The performance of the technique isevaluated by comparing its derivative values with the values obtained from a brute forcenumerical method. Finally application of the perturbation analysis in solving the system-optimaldynamic traffic assignment (SO_DTA) problem is described, in conjunction with a new solutionmethod called SO_ordered method. Executing this method requires the calculation of the particlemarginal travel times which is made possible through use of the perturbation analysis method.The approach presented in this paper has the potential to considerable improve our ability tocompute system optimal traffic strategies in practice.