A dynamic-niching-based Pareto domination for multimodal multiobjective optimization

Maintaining the diversity of the decision space is of great significance in multimodal multiobjective optimization problems (MMOPs). Since the traditional Pareto-dominance-based algorithms prioritize the convergence of individuals by the Pareto-dominated sorting, it will face a phenomenon that a large number of well-distributed individuals could be dominated by other well-converged individuals during the optimization of MMOPs. To solve this problem, we propose a dynamic-niching-based Pareto domination, called DNPD, which adds a dynamic niche to constrain the tranditional Pareto dominantion to achieve a balance of convergence and diversity of population in the decision space. In the early stage of the algorithm, the smaller niche makes the algorithm retain a large number of well-distributed individuals. In the later stage of the algorithm, the dynamically increased niche accelerates the convergence of the population. DNPD can be integrated into the Pareto-dominance-based algorithms to solve MMOPs. Experimental results show that the DNPD performs well on MMF and IDMP series benchmark functions after comparing the original algorithm with the original algorithm combined with the DNPD.