Particle Swarm Optimization Based on Self-adaptive Acceleration Factors
Gai-yun Wang, Dong-xue Han · 2009
The particle swarm optimization (PSO), which goes right after Ant Colony Algorithm, is another new swarm intelligence algorithm. PSO has the same drawbacks as other optimization algorithms in spite of its predominance in some fields. That is easily falling into local optimization solution and low convergence velocity in the final stage. An improved algorithm called acceleration factors self-adaptive PSO (ASAPSO) was proposed for the drawbacks. The constant acceleration coefficients in the standard PSO were changed into self-adaptive acceleration factors in the progress of evolution. By controlling the acceleration factors, the particles have stronger global search capability in the early stage and are less likely to be impacted by the current global optimum position and the particles fly to global optimum position more rapidly in the final stage, thus achieved enhanced the convergence velocity. From the numerous experimental results on 4 widely used benchmark functions, we can show that ASAPSO outperforms other three improved PSO.