In this paper, we discuss the specification of local search directions in a multi-objective genetic local search (MOGLS) algorithm. The MOGLS algorithm is a GA-based hybrid algorithm for inding a set of Pareto-optimal solutions of a multi-objective optimization problem. Ishibuchi and Murata (1998) have already hybridized a multi-objective genetic algorithm with a local search procedure. In their algorithm, a weighted sum of multiple objectives is used as a fitness function for selecting a pair of parent solutions. The fitness function defined by the weighted sum is also employed in the local search procedure for a new solution generated from the selected parent solutions. That is, the iterative improvement of the newly generated solution is performed in the multi-dimensional objective space based on the local search direction specified by the weight values in the fitness function. In this paper, we propose a specification method of local search directions in the MOGLS algorithm. The proposed method specifies the local search direction of each solution according to its location in the multiobjective space. By computer simulations on flowshop scheduling problems, we demonstrate how the proposed method can improve the search ability of the MOGLS algorithm.
展开▼
