Neighborhood structures for genetic local search algorithms

We examine the performance of a genetic local search (GLS) algorithm for flowshop scheduling problems. The GLS is a hybrid algorithm of a local search and a genetic algorithm. We have already modified the local search procedure in order to improve the performance of the GLS. In the modified local search procedure, all the neighborhood solutions are not examined. The performance of the GLS is not sensitive to the choice of parameter values such as the crossover probability and the mutation probability. That is the main advantage of the GLS. In this paper, we examine the relation between a mutation operator and a local search procedure. By computer simulations on flowshop scheduling problems, we find that a shift change is appropriate for the local search procedure in the GLS.