Title: Parameter settings in particle swarm optimisation algorithms: a survey

Authors: Jing Li; Shi Cheng

Addresses: Department of Basic Course, Shaanxi Railway Institute, Weinan 714000, China ' School of Computer Science, Shaanxi Normal University, Xi'an 710119, China

Abstract: In swarm intelligence, 'fair comparison' is critical for the performance evaluation of algorithms. In this paper, the setting of parameters in particle swarm optimisation (PSO) algorithms, which include the population size S, topology structure (number of neighbours k), inertia weight w, acceleration coefficient c1, c2, velocity constraint Vmax, and the boundary constraint strategy, are reviewed and analysed. Based on the analysis and discussion of parameters and the variants of PSO algorithms, a list of parameter settings of PSO algorithms and a recommendation of PSO comparison are given. To compare variants of PSO algorithms, a recommended solution maybe that all compared algorithms have the same number of population size and the maximum number of fitness evaluations, and the inertia weight w, acceleration coefficient c1, c2 are the same settings as its original version.

Keywords: swarm intelligence; particle swarm optimisation; PSO; parameter investigation; performance comparison.

DOI: 10.1504/IJAAC.2022.121124

International Journal of Automation and Control, 2022 Vol.16 No.2, pp.164 - 182

Received: 24 Dec 2019
Accepted: 15 May 2020

Published online: 28 Feb 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article