GPSOM: group-based particle swarm optimization with multiple strategies for engineering applications

摘要

Particle Swarm Optimization (PSO) is a classic optimization algorithm; however, it has issues when solving high-dimensional complex optimization problems, such as sensitivity to initial parameters, insufficient population diversity, and susceptibility to becoming stuck in local optima. To this end, we propose a particle swarm optimization algorithm with multiple swarm strategies (GPSOM) that significantly improves the algorithm’s optimization performance by designing grouping and diversification strategies for the population. GPSOM divides the population into three subgroups and adopts targeted strategies based on different phases: sub-swarm 1 focuses on exploration and improves population diversity and local search ability by dynamically adjusting operators; sub-swarm 2 focuses on exploitation, introducing sine cosine factors and adaptive mechanisms to traverse the solution space efficiently; sub-swarm 3 achieves global and local balance, generating high-quality solutions through adaptive factors and position adjustment operators. The experimental results show that GPSOM performs significantly better than traditional algorithms on the Combinatorial Evolutionary Computing (CEC) 2020 benchmark test set (hereinafter referred to as 20TS), and its reliability and efficiency have been demonstrated in practical applications, such as 15 engineering examples, vehicle cruise control, and robot path optimization. GPSOM is an efficient, stable, practical optimization algorithm that effectively solves high-dimensional complex optimization problems.