A dynamic neighborhood based multi-objective particle swarm optimizer for multimodal multi-objective problems

This paper presents a dynamic neighborhood based multi-objective particle swarm optimizer (DN-MOPSO) for solving multimodal multi-objective optimization problems. In DN-MOPSO, the whole evolutionary process is consisted of two parts. Firstly, the entire population is divided into multiple dynamic sub-swarms and the number of particles is randomly distributed in each sub-swarm. This is conducive to keeping population diversity and improving the speed of information diffusion among the population. Secondly, after the population is divided into multiple sub-swarms, there can be isolated particles that do not belong to any sub-swarm. By sorting the position information of isolated particles in space, we form a new sub-swarm of three nearby isolated particles to improve the convergence of obtained solutions. The comparison results between DN-MOPSO and other well-known algorithms on eight test problems suggest that DN-MOPSO demonstrates superior performance for solving different types of multimodal multi-objective functions.