A multi-objective evolutionary algorithm based on a grid with adaptive divisions for multi-objective optimization with irregular Pareto fronts

Abstract

The performance degradation of most existing multi-objective optimization evolutionary algorithms (MOEAs) when tackling multi-objective problems (MOPs) with irregular Pareto fronts is a critical challenge in the field of multi-objective optimization. To address this issue, a novel grid-based MOEA is proposed in this paper. This algorithm dynamically adjusts the number of grid divisions during the optimization process, thereby enabling effective partitioning of the objective space and guiding solution distribution across MOPs with varying Pareto front shapes. Additionally, to enhance diversity preservation, a grid stabilization strategy is proposed to maintain a stable environment for diversity, while a boundary solution protection strategy ensures diversity by promoting exploration of the boundaries. Furthermore, a population reselection method is designed to bolster exploration capabilities within the objective space. Experimental results from benchmark test suites, which include a variety of Pareto front types, demonstrate that our proposed algorithm outperforms seven state-of-the-art MOEAs in addressing both irregular and regular Pareto front MOPs.