This chapter focuses on the definition of an efficient parallel metaheuristic which takes advantage of the multi-core design of recent processors. The approach is designed as a Greedy Randomized Adaptive Search Procedure (GRASP) hybridized with a multi-threaded version of an Evolutionary Local Search (ELS) metaheuristic scheme. Our approach is evaluated on an extension of the Vehicle Routing Problem where a heterogeneous fleet of vehicles is available to service a set of customers. The objective consists in designing a set of trips for a limited heterogeneous fleet of vehicles located at a depot node which minimizes the total transportation cost. Each type of vehicles is defined by a capacity and by the number of available vehicles. The efficiency of the parallel approach is evaluated on a new set of real-life instances built out of data from the French districts. A fair comparative study, using a same implementation, is done to evaluate the impact of the number of threads on the convergence rate. Thus, a better trade-off between solution quality and computational time can be reached. The numerical experiments show that the hybrid GRASPxparallel ELS outperforms the classical iterative version and provides numerous new best solutions. © 2013 Springer-Verlag Berlin Heidelberg.
CITATION STYLE
Duhamel, C., Gouinaud, C., Lacomme, P., & Prodhon, C. (2013). A Multi-thread GRASPxELS for the heterogeneous capacitated vehicle routing problem. Studies in Computational Intelligence, 434, 237–269. https://doi.org/10.1007/978-3-642-30671-6_9
Mendeley helps you to discover research relevant for your work.