Particle swarm optimization with adaptive learning strategy

Population diversity maintenance is a crucial task for preventing a particle swarm optimization (PSO) algorithm from being trapped in local optima. A learning strategy is an effective means of improving population diversity. However, for the canonical PSO algorithm, the learning strategy focuses mainly on the global best particle, which leads to a loss of diversity. To increase the population diversity and strengthen the global search ability in PSO, this paper proposes a PSO algorithm with an adaptive learning strategy (PSO-ALS). To better promote the performance of the learning strategy, the swarm is adaptively grouped into several subswarms. The particles in each subswarm are further classified into ordinary particles and the locally best particle, and two different learning strategies without an explicit velocity are devised for updating the particles to increase the population diversity. Thus, the global optimum is determined by comparing the fitness values of the updated best particles in each subswarm. The proposed algorithm is compared with state-of-the-art PSO variants. The experimental results illustrate that the performance of PSO-ALS is promising and competitive in terms of enhanced population diversity and global search ability.