The Quick Time Dependent Quickest Flow Problem: A Lesson in Zero-Sum Cycles

Abstract

A quick solution technique for the integral time-dependent quickest flow problem with no waiting is presented. The proposed technique is based on the successive shortest path approach and modifies an existing algorithm to improve its average performance. At each iteration, a reoptimization procedure is employed to determine the augmenting path given updates to the residual graph. The residual graph, by construction, almost always contains zero-sum cycles when employed in this context. These zero-sum cycles pose a unique problem for the reoptimization technique. A heuristic that can be embedded in the reoptimization algorithm to provide path solutions in the presence of zero-sum cycles has been proposed. In the computational experiments, the heuristic provided an optimal solution nearly 100% of the times. Further, a modified implementation of an existing path-finding algorithm has been used to solve the time-dependent quickest flow problem with source waiting.