An improved PSO with detecting and local-learning strategy (Q2221570)

Summary: Particle swarm optimisation (PSO) has been applied to a variety of problems due to its simplicity of implement. However, the standard PSO suffers from premature convergence and slow global optimisation. This paper presents a novel PSO algorithm, in which detecting strategy and local-learning strategy are adopted to improve PSO's performance. In the new PSO algorithm, which is called DLPSO in this paper, search space of each dimension is divided into many equal subregions. According to statistical information of all particles' historical best position, the globally best particle can detect some inferior (or superior) subregions. In the local-learning strategy, the global best particle can carry out a local search during the later evolution process. The results of experiments show that the detecting strategy can act on the globally best particle to jump out of the likely local optimal solutions while local-learning strategy can help DLPSO obtain more accurate solutions. In addition, experimental results also demonstrate that DLPSO is more suitable for multimodal function optimisation while it has a comprehensive ability for function optimisation.