Multipopulation cooperative particle swarm optimization with a mixed mutation strategy

• Published in: Information sciences Vol. 529; pp. 179 - 196
• Main Authors: Li, Wei, Meng, Xiang, Huang, Ying, Fu, Zhang-Hua
• Format: Journal Article
• Language: English
• Published: Elsevier Inc 01.08.2020
• Subjects: Differential mutation; Dynamic segment-based mean learning strategy; Multidimensional comprehensive learning; Particle swarm optimization; Dynamic segment-based mean learning strategy; Particle swarm optimization; Multidimensional comprehensive learning; Differential mutation
• ISSN: 0020-0255; 1872-6291
• DOI: 10.1016/j.ins.2020.02.034

•Propose a multipopulation cooperative particle swarm optimization (MPCPSO).•The dynamic segment-based mean learning strategy (DSMLS), which is employed to construct learning exemplars, achieves information sharing and coevolution between populations.•The multidimensional comprehensive learning strategy (MDCLS) is employed to speed up the convergence tendency and solution accuracy of MPCPSO.•A difference mutation operator is introduced to increase the population diversity and enhance the global exploration ability of MPCPSO. The traditional particle swarm optimization algorithm learns from the two best experiences: the best position previously learned by the particle itself and the best position learned by the entire population to date. This learning strategy is simple and ordinary, but when addressing high-dimensional optimization problems, it is unable to quickly find the global optimal solution due to its low efficiency. This paper proposes a multipopulation cooperative particle swarm optimization (MPCPSO) algorithm with a dynamic segment-based mean learning strategy and a multidimensional comprehensive learning strategy. In MPCPSO, the dynamic segment-based mean learning strategy (DSMLS), which is employed to construct learning exemplars, achieves information sharing and coevolution between populations. The multidimensional comprehensive learning strategy (MDCLS) is employed to speed up convergence and improve the accuracy of MPCPSO solutions. Additionally, a differential mutation operator is introduced to increase the population diversity and enhance the global exploration ability of MPCPSO. Sixteen benchmark functions and seven well-known PSO variants are employed to verify the advantages of MPCPSO. The comparison results indicate that MPCPSO has a faster convergence speed, obtains more accurate solutions, and is more robust.