SOLVING AN INSTANTANEOUS DYNAMIC USER-OPTIMAL ROUTE CHOICE MODEL

Dynamic models of the behavior of automobile drivers in choosing routes through urban transportation networks are required for the design and operation of intelligent transportation systems generally, and traveler information systems in particular. In this paper, drivers' route choice behavior based on current, or instantaneous, information on network conditions is described using an optimal control theory formulation. The time-dependent origin-destination vehicle trip pattern is assumed to be known. The instantaneous dynamic user-optimal route choice problem is to allocate vehicle flours to the current minimal-cost routes, defined as the routes that minimize the route travel cost between each decision node (any node on the route including the origin) and the destination node based on the currently prevailing travel times. The continuous time formulation of the problem is transformed into a discrete time nonlinear programming (NLP) formulation. Each of a sequence of NLPs is then solved by an algorithm based on the Frank-Wolfe technique. In this sequence, approximate link travel times are updated iteratively in order to represent flow propagation over routes. Computational results from applying the algorithm to two test networks conclude the paper.