Multi-objective particle swarm optimization algorithm based on leader combination of decomposition and dominance

Article type: Research Article

Authors: Hu, Ziyu^{a; b; *} | Yang, Jingming^{a; b} | Cui, Huihui^{a; b} | Sun, Hao^{a; b} | Wei, Lixin^{a; b}

Affiliations: [a] School of Electrical Engineering, Yanshan University, Qinhuangdao, Hebei, PR China | [b] Key Lab of Industrial Computer Control Engineering of Hebei Province, Yanshan University, Qinhuangdao, Hebei, PR China

Correspondence: [*] Corresponding author. Ziyu Hu, School of Electrical Engineering, Yanshan University, Qinhuangdao, Hebei 066004, PR China. Tel.: +86 18630368118; Fax: +86 0335 8060089; E-mail: sunhao@ysu.edu.cn.

Abstract: Convergence and well-distribution are two basic issues in multi-objective optimization problems(MOPs). However, it is hard to optimize them simultaneously for the selection of leader particle is not always leading the population to the Pareto front. To make a better performance of multi-objective particle swarm optimization algorithm(MOPSO), decomposition and domination leadership particle selection mechanism have been introduced into MOPSO. Decomposition leader particle selection mechanism is used to keep the swarm with diversity, while domination leader particle selection mechanism make the particles move to the Pareto front. The performance of our proposed method is validated based inverted generation distance(IGD) and compared with five state-of-the-art algorithms on a number of unconstrained benchmark problems. Empirical analysis demonstrates the superiority of our proposed method on both proximity and diversity.

Keywords: Intelligence algorithm, multi-objective optimization, particle swarm optimization, soft computing

DOI: 10.3233/JIFS-17336

Journal: Journal of Intelligent & Fuzzy Systems, vol. 33, no. 3, pp. 1577-1588, 2017