A Priori Stochastic Dynamic Traffic Assignment for On-Line Operations Under Information Systems

Abstract

Existing dynamic traffic assignment (DTA) formulations predominantly assume the time-dependent origin-destination (O-D) trip matrix to be known a priori for the planning horizon or require robust a priori O-D predictions. Such assumptions are unduly optimistic for general networks in an on-line operational framework. Also, the associated solution approaches are computationally intensive in a centralized architecture. To address these key on-line issues, a stochastic DTA approach is proposed in which O-D desires are explicitly recognized as random variables. To ensure that the associated solution methodology is feasible for on-line implementation, a combination of on-line and off-line strategies is proposed. In this paper, an a priori optimization strategy is used to address the off-line problem. Experimental results indicate that the off-line solution serves as an effective and robust initial on-line solution.