Fleet organization models for online vehicle routing problems

Abstract

Online vehicle routing problems with time windows are highly complex problems for which different artificial intelligence techniques have been used. In these problems, the exclusive optimization of the conventional criteria (number of vehicles and total traveled distance) leads to the appearance of geographic areas and/or time periods that are not covered by any vehicle because of their low population density. The transportation demands in these zones either cannot be satisfied or need to mobilize new vehicles. We propose two agent-oriented models that propose a particular dynamic organization of the vehicles, with the objective to minimize the appearance of such areas. The first model relies on a spatial representation of the agents' action zones, and the second model is grounded on the space-time representation of these zones. These representations are capable of maintaining an equilibrated distribution of the vehicles on the transportation network. In this paper, we experimentally show that these two means of distributing vehicles over the network provide better results than traditional insertion heuristics. They allow the agents to take their decisions while anticipating future changes in the environment. © 2012 Springer-Verlag Berlin Heidelberg.