PSO with dynamic acceleration coefficient based on mutiple constraint satisfaction: Implementing Fuzzy Inference System

Particle Swarm Optimization (PSO) parameters like the Inertia weight and acceleration coefficients are generally kept constant in classical PSO. But it has been found that changing these parameters dynamically makes the PSO more efficient. In this paper we propose a modified PSO algorithm where we change the value of the acceleration coefficient dynamically over iterations. We have used a Fuzzy Inference system (FIS) to obtain a new coefficient value for the PSO for each round. The coefficient depends on satisfaction of certain constraints given as inputs to the FIS. This dynamic modification of the coefficient has been found to increase the efficiency of PSO and also improve its convergence speed.