A computational study of flexible routing strategies for the VRP with stochastic demands

Abstract

We develop and numerically test a new strategy for the vehicle routing problem with stochastic customer demands. In our proposed approach, drivers are assigned to predetermined delivery routes in which adjacent routes share some customers. This overlapping assignment structure, which is inspired by the open chain design from the field of manufacturing process flexibility, enables drivers to adapt to variable customer demands while still maintaining largely consistent routes. Through an extensive computational study and scenario analysis, we show that relative to a system without customer sharing, such flexible routing strategies partly mitigate the transportation costs of filling unexpected customer demands, and the relative savings grow with the number of customers in the network. We also find that much of the cost savings is gained with just the first customer that is shared between adjacent routes. Thus, the overlapped routing model forms the basis for a practical and efficient strategy to manage costs from demand uncertainty.