Emulation of Dynamic Equilibrium in Traffic Networks

Abstract

Some notions of equilibria, both static and dynamic, are discussed relative to traffic assignment. Existing types of assignment techniques are examined. Their relative merits and deficiencies in predicting the self-assignment of drivers to a road network according to the principle where each individual minimizes his own travel time are discussed. A technique is proposed, which combines the incremental, iterative and queueing procedures employed in previous techniques. For each time period with its assumed time-stationary demands, a sequence of iterations is performed, each consisting of a complete incremental assignment. In each iteration the final link costs from the previous iteration are given a relative weight W, with the costs normally used in the incremental technique having a weight of 1-W. This trades off some of the multipath capability of incremental techniques in order to avoid irreversible assignments to ultimately irrational paths in the early increments.