A MODEL OF TIME-VARYING FLOWS ON A CONGESTED MULTIDESTINATION NETWORK

In designing an improved highway network, it is useful to predict traffic congestion under various settings. In particular, a dynamic flow model is suitable for a highly congested road network during the morning rush-hour. In the literature on dynamic traffic assignment models, however, attention has been paid for the most part to a single-destination network so far. The purpose of this paper is to propose a multicommodity dynamic flow model for predicting traffic congestion on a multidestination network. The model is formulated as an equilibrium programming problem, the solution of which represents a 'commodity optimizing' flow pattern in the network. The problem consists of a set of convex programming subproblems and each subproblem amounts to minimizing the total cost for the users having a common destination. Existence of an equilibrium flow pattern is established under appropriate conditions and an algorithm for computing it is described.