Diversity-enhanced particle swarm optimization algorithm based on the group behaviour of social spiders
Zhao Liu, Han Li, Ping Zhu
Abstract
Particle swarm optimization (PSO) is a representative swarm intelligence algorithm, which has the drawback of being restricted by premature convergence. To make PSO less likely to be restricted by premature convergence and to enhance its exploration and exploitation ability, this article proposes a social spider inspired particle swarm optimization (SSI-PSO). Based on PSO, the proposed algorithm divides the swarm into subgroups to mimic different behaviours of a social spider colony. Particles are classified in each iteration and then a variety of search strategies is adopted. Specifically, dominant male particles aim to search the neighbourhood, while negative female particles tend to search in the opposite direction to enhance the diversity. Meanwhile, positive female particles and non-dominant male particles are designed to balance the potential impact. Various commonly used benchmark functions and truss structural designs are tested for comparison. The results indicate that the proposed SSI-PSO is effective and efficient.