Genetic algorithm with adaptive diversification and intensification for the vehicle routing problem

Abstract

Hybridization is a common theme employed to improve metaheuristics. Having been known as a less effective metaheuristic for the vehicle routing problem, genetic algorithm (GA) has received attention from researchers for modification and improvement by means of hybridization, for example by adopting a local search technique for the mutation operator. In this paper, we propose another hybridization idea by using an adaptive threshold in population management, whereby in earlier stages of the GA iterations, a larger threshold is used to open up the search space, and in later stages the threshold will be reduced to intensify the search in a smaller neighborhood area. This idea is similar to diversification and intensification processes used in the Tabu Search. The main GA engine follows good principles found from the literature. Two crossover operators, the partially mapped crossover (PMX) and the order crossover (OX), were also tested to see if the adaptive threshold has complication with the other concepts. The experiment results based on Solomon benchmark instance point out that the adaptive threshold favors the PMX but produces worse fitness and longer run time with the OX. More fine-tuning of parameters is invited to further enhance the GA performance from this research.