Acceleration of neighborhood evaluation for a multi-objective vehicle routing

Abstract

In this paper a multi-objective vehicle routing problem (MOVRP) with the criteria being the total distance and the utilization of the vehicle space is considered. Two methods were developed to decrease the execution time of the main part of the algorithm - The neighborhood search procedure. First method, an accelerator, is defined in order to reduce the computational complexity of the algorithm from O(n3) to O(n2). The second method utilizes multiple threads of execution to speedup the neighborhood search. Both methods were applied to tabu search metaheuristic and tested against the basic version of the algorithm. In result, we concluded that the enhanced version allows for significant reduction of execution time (2500 times for 5000 clients) that scales well with the number of clients. Moreover, this allows the enhanced algorithm to find significantly better approximations of the Pareto front in the same time as the original algorithm.