A Tri-Swarm Particle Swarm Optimization Considering the Cooperation and the Fitness Value

In this paper, a new variant of Particle Swarm Optimization (PSO) called Tri-swarm Particle Swarm Optimization (TSPSO) was proposed. Inspired by the cooperation within the animal groups, the whole population of PSO was divided into three sub-swarms according to their responsibility: i) the exploration swarm (ERS) to explore, ii) the exploitation swarm (EIS) to exploit, iii) the convergence swarm (CS) to converge. And according to the fitness value with a tested partition ratio “3/10, 4/10, 3/10” corresponding to three sub-swarms, the particles in the population owned the worst, moderate and best fitness value were respectively divided into ERS, EIS and CS. The results on seven unimodal benchmark functions demonstrated the superiority of the proposed variant compared with other five variants.