A modified group search optimiser for constrained optimisation problems

Abstract

The purpose of this study is to propose a modified group search optimiser (GSO) algorithm for constrained optimisation problems. Three strategies are designed to improve the performance of GSO algorithm. Levy flight strategy is employed for the producer to improve its exploitation ability, and also to simplify the computation. To further improve its exploitation ability, dimension-by-dimension search strategy is used to tackle the dimensions interference problem when producer uses levy flight strategy to find new position. Mutation strategy with adaptive controlling probability is used to improve the exploration ability by keeping the diversity of population. Using the superiority of feasible point scheme and the parameter free penalty scheme to handle constraints, the modified GSO algorithm was tested on 13 well-known benchmark problems. Comparisons are made with the standard GSO algorithm and other algorithms which use penalty function and various constraint handling methods. Results show the modified GSO algorithm is effective and competitive for constrained optimisation problems.