The Vehicle Routing Problem with Stochastic Demands (VRPSD) is a popular stochastic variant of the standard multiple vehicle routing problem with vehicle load capacity constraints. In the VRPSD, it is assumed that customer locations are known when planning, but that customer load demand is known only in distribution before vehicle arrival. The VRPSD is generally modeled as a stochastic optimization problem with recourse, where the recourse actions determine how vehicles respond to capacity failures during the operation of preplanned tours. In this study, we present a locally-coordinated recourse strategy where vehicles pool capacity by operating in pairs. We develop a tabu search heuristic to configure systems under this strategy. Focusing on the case with homogeneous customer demand distributions, we compare the approach to the popular "detour-to-depot" single vehicle recourse strategy.
Computational results on practical instances indicate that required fleet sizes and expected travel costs can be substantially reduced using the locally-coordinated strategy.

Students and faculty are invited to remain after the seminar for a coffee and snacks reception sponsored by the The Logistics Institute.