Multiobjective Evolutionary Algorithm based on Fast Elite Sampling Strategy and Difference-based Local Search for VRPTW

This paper addresses the vehicle routing problem with time windows (VRPTW), aiming to reduce the vehicles number and to minimize the time-wasting during the delivery process caused by early arrival. For solving VRPTW, a multiobjective evolutionary algorithm based on fast elite sampling strategy and difference-based local search (MOEAFESS/DLS) is proposed. The special strategy in MOEAFESS/DLS is fast elite sampling strategy (FESS) which consists of two parts. First, using Pareto dominating and dominated relationship-based fitness function (PDDR-FF) to evaluate individuals, which can easily select nondominated individuals and the individuals which have larger domination area. Then, mixing with the sampling strategy of vector evaluated genetic algorithm (VEGA) can achieve fast convergence and sufficient diversity. The evolution according to FESS will be used as a global search strategy for MOEA-FESS/DLS. In addition, a local search based on differences between individuals called difference-based local search (DLS) is used in MOEA-FESS/DLS. In this way, the individuals with poor performance in the population generated by global search are guided to move closer to those who perform better. So it can further enhance the search ability of MOEA-FESS/DLS. Experimental results on Solomon benchmark demonstrate that the proposed MOEA-FESS/DLS is effective, which the performance outperforms NSGA-II, SPEA2 and MOEA/D.