An Approach to Enhance the Equilibrium of Search Capabilities for Multi-objective Evolutionary Algorithms Based on Differential Evolution

Applying evolutionary principles in designing algorithms to solve multi-objective optimization problems is a research area that has received attention in recent years. A multi-objective evolutionary algorithm is a population-based algorithm that facilitates a trade-off solution set to provide the decision maker. To ensure that a multi-objective evolutionary algorithm can find a solution set that is both convergent and highly diverse, the issue of ensuring a balance between the ability to explore and exploit during the search process plays a significant role. In this study, we propose a mechanism to maintain this balance by leveraging the relationship between the algorithm’s quality assessment indicators and the trajectory of the search process. This balancing mechanism is then applied to improve the symmetry of multi-objective evolutionary algorithms based on differential evolution. In which the correlation information about the variation in convergence and diversity quality measures was integrated with the step-length of the differential evolution operator to maintain the equilibrium between the exploratory and exploitative nature of evolution. Empirical results on various typical metrics and benchmark problems demonstrate the effectiveness of the proposed method.