Authors: Hai-ying Lan; Gang Xu; Yu-qun Yang
Addresses: School of Mathematics and Statistics, Jiangxi Normal University, Nan Chang, Jiang Xi, 330022, China ' Department of Mathematics, Nanchang University, Nan Chang, Jiang Xi, 330031, China ' The Middle School attached to Nanchang University, Nan Chang, Jiang Xi, 330047, China
Abstract: In the scope of multi-objective particle swarm optimisation (MOPSO) research, avoiding premature convergence remains a challenge. To address this issue, the article develops an enhanced multi-objective particle swarm optimisation with Levy flight (LF-MOPSO). In LF-MOPSO, swarm is made to evolve based on the original MOPSO to accelerate convergence. Then, Levy flight is adaptively activated to maintain diversity, so as to deal with the premature convergence when Pareto frontier is stagnant. It realises the transformation between shrinkage and divergence of population diversity by self-adaptive conversion mechanism, which further improves the search ability of MOPSO. LF-MOPSO has been contrasted with some recently improved MOPSOs, the experimental outcomes indicate that LF-MOPSO ensures the better approximation to the Pareto optimal frontier, and gains the non-dominated solutions with good diversity and distribution.
Keywords: MOP; multi-objective optimisation; PSO; particle swarm optimisation; Pareto solution; external archive; Levy flight; conversion mechanism; population diversity.
International Journal of Computing Science and Mathematics, 2023 Vol.17 No.1, pp.79 - 94
Received: 06 Feb 2021
Accepted: 20 Jul 2021
Published online: 20 Apr 2023 *