Efficient Neighborhood Search for the Probabilistic Multi-Vehicle Pickup and Delivery Problem

This paper deals with the probabilistic multi-vehicle pickup and delivery problem. We develop an efficient neighborhood evaluation procedure which allows to reduce the computational complexity by two orders of magnitude with respect to a straightforward approach. The numerical experiments indicate that, if incorporated in a local search strategy, our neighborhood evaluation technique, provides very good results in terms of computation time reduction and equity of the workload distribution among the available vehicles.