Self Adaptive Acceleration Factor in Particle Swarm Optimization

Abstract

Particle swarm optimization (PSO), which is one of the leading swarm intelligence algorithms, dominates other optimization algorithms in some fields but, it also has the drawbacks like it easily falls into local optima and suffers from slow convergence in the later stages. This paper proposes improved version of PSO called Self Adaptive Acceleration Factors in PSO (SAAFPSO) to balance between exploration and exploitation. We converted the constant acceleration factors used in standard PSO into function of particle’s fitness. If a particle is more fit then it gives more importance to global best particle and less to itself to avoid local convergence. In later stages, particles will be more fitter so all will move towards global best particle, thus achieved the convergence speed. Experiment is performed and compared with standard PSO and Artificial bee colony (ABC) on $14$ unbiased benchmark optimization functions and one real world engineering optimization problem (known as pressure vessel design) and results show that proposed algorithm SAAFPSO dominates others.