A novel multi-objective immunization algorithm based on dynamic variation distance

Abstract

Multi-objective immunization algorithms have received wide attention in solving multi-objective optimization problems. However, most multi-objective immune algorithms use clone selection strategies to accelerate convergence and focus on only the local information of clone solutions, which damages the diversity of the population and leads the algorithms to fall into local optimum easily. To address the above problems, this paper proposes a novel multi-objective immunization algorithm based on dynamic variation distance. First, the algorithm ensures the uniform distribution of the optimal solution by the established dynamic variation distance model. Second, we divide the population into mainpop and auxpop by quality, and design a new evolutionary strategy to achieve specific variants at different evolutionary stages under different constraints. Third, the mating selection strategy and the population selection operator are further explored to handle infeasible solutions adaptively in the late evolutionary stage. And we have conducted extensive experiments on the proposed algorithm. The results prove that our proposed algorithm outperforms 14 existing algorithms in most cases of 16 test problems, which confirms the superiority of the algorithm.