A GRASPxELS with Depth First Search Split Procedure for the HVRP

Split procedures have been proved to be efficient within global framework optimization for routing problems by splitting giant tour into trips. This is done by generating optimal shortest path within an auxiliary graph built from the giant tour. An efficient application has been introduced for the first time by Lacomme et al. (2001) within a metaheuristic approach to solve the Capacitated Arc Routing Problem (CARP) and second for the Vehicle Routing Problem (VRP) by Prins (2004). In a further step, the Split procedure embedded in metaheuristics has been extended to address more complex routing problems thanks to a heuristic splitting of the giant tour using the generation of labels on the nodes of the auxiliary graph linked to resource management. Lately, Duhamel et al. (2010) defined a new Split family based on a depth first search approach during labels generation in graph. The efficiency of the new split method has been first evaluated in location routing problem with a GRASP metaheuristic. Duhamel et al. (2010) provided full numerical experiments on this topic. This research report focuses on the application of Depth First Search Split strategy for Heterogeneous VRP (HVRP) trying to extend the Split proposals introduced on the LRP to a new routing problem. The numerical experiments use classical HVRP instances and a set of new real life instances matching to the French districts.