A hybrid multi-objective algorithm for dynamic vehicle routing problems

Abstract

This paper analyzes firstly the limitation of traditional methods when used to solve Dynamic Vehicle Routing Problem (DVRP), and then constructs an adapted DVRP model named DVRPTW based on Multi-objective Optimization. In this model, we consider two sub-objectives such as vehicle number and time cost as an independent objective respectively and simultaneously to coordinate the inherent conflicts between them. Also, a hybrid Multi-objective ant colony algorithm named MOEvo-Ant is proposed and some crucial techniques used by MOEvo-Ant algorithm are discussed too. In our ant colony algorithm, an EA is introduced into our ant colony algorithm to increase pheromone update. The main reason of the introduction is that we try to take advantage of the outstanding global searching capability of EA to speed up the convergence of our algorithm. Simulating experiments demonstrate that no matter when compared with the known best solutions developed by previous papers or when use it to solve dynamic vehicle routing problems generated randomly, our algorithm illustrates pretty good performance. © 2008 Springer-Verlag Berlin Heidelberg.