Dynamic multi-swarm global particle swarm optimization

Abstract

To satisfy the distinct requirements of different evolutionary stages, a dynamic multi-swarm global particle swarm optimization (DMS-GPSO) is proposed in this paper. In DMS-GPSO, the entire evolutionary process is segmented as an initial stage and a later stage. In the initial stage, the entire population is divided into a global sub-swarm and multiple dynamic multiple sub-swarms. During the evolutionary process, the global sub-swarm focuses on the exploitation under the guidance of the optimal particle in the entire population, while the dynamic multiple sub-swarms pour more attention on the exploration under the guidance of the neighbor’s best-so-far position. Moreover, a store operator and a reset operator applied in the global sub-swarm are used to save computational resource and increase the population diversity, respectively. At the later stage, some elite particles stored in an archive are combined with the DMS sub-swarms as a single population to search for optimal solutions, intending to enhance the exploitation ability. The effect of the new introduced strategies is verified by extensive experiments. Besides, the comparison results among DMS-GPSO and other 9 peer algorithms on CEC2013 and CEC2017 test suites demonstrate that DMS-GPSO can effectively avoid the premature convergence when solving multimodal problems, and yield more favorable performance in complex problems.