A Multi-objective Multimodal Evolutionary Algorithm Using a Novel Tournament and Environmental Selections

Abstract

Multi-modal multi-objective optimization refers to multi-objective optimization problems that have more than one Pareto set. This paper proposes a new algorithm, modifying both the tournament and the environment selections. The purpose of this paper is to utilize the full potential of the parental population to produce diverse offspring and avoid producing duplicated solutions. In order to achieve the goal, we modify the selection mechanisms, which usually enforce the selection of solutions with high quality. This can lead to duplicates in the mating pool. Using the modified tournament selection, we aim to enable the algorithm to select diverse solutions in sparse regions of the search space. Our experiments agree with our expectations, confirming that our proposed algorithm is more effective than current competitor algorithms at approximating Pareto sets in most multimodal multi-objective optimization problems with different number of decision variables and objective functions.