2L-CVRP: A GRASP resolution scheme based on RCPSP

This paper addresses an extension of the well known capacitated vehicle routing problem (CVRP) where the customer demands represented by two dimensional weighted items have to be packed into the vehicles. This problem, denoted 2L-CVRP, is NP-hard and it is particularly difficult to solve in practice. We propose a GRASP algorithm in which the loading constraint is handled by a resource constraint project scheduling problem (RCPSP) heuristic specially tuned to address the bin packing part of the problem. The effectiveness of our approach is demonstrated through computational experiments on both CVRP and non-sequential 2L-CVRP instances. Three new best average costs for class 2 to 5 are provided for 2L-CVRP instances.