A Multi-Niching Multi-Objective Genetic Algorithm for Solving Complex Multimodal Problems

Abstract. In this work, a Multi-Niching Multi-Objective Genetic Algorithm is presented for solving multimodal optimization problems. The originality of this algorithm resides in its niching procedure, which maintains population diversity in both objective and design variable spaces. In particular, the clearing of non-dominated individuals in the archive update is carried out using a global density estimator computed from distances between individuals in objective and design variable spaces. The efficiency of this algorithm is shown on mathematical test functions with multiple equivalent Pareto-optimal fronts and on electromagnetic design problems