An improved memetic algorithm for the vehicle routing problem

This paper proposes an improved memetic algorithm (IMA) for solving the vehicle routing problem (VRP). The IMA uses a hybrid-selection strategy to enhance the crossover operator. Classical local search operators are combined for the route improvement. Besides, the same chromosomes are modified to be different so that the population diversity is preserved and the algorithm kept from premature convergence. The performance of IMA is tested by solving several VRP benchmark instances and compared to two genetic algorithms and the basic memetic algorithm. Experimental results validate that IMA could obtain superior solutions to the three counterparts within reasonable computation time.