Particle swarm optimization with a new update mechanism

Abstract

Particle swarm optimization (PSO) has been invented by inspiring social behaviors of fish or birds to solve nonlinear global optimization problems. Since its invention, many PSO variants have been proposed by modifying its solution update rule to improve its performance. The social component of update rule of PSO is based on subtraction between current position of particle and global best information. Similarly, the cognitive component works by using subtraction between the current position of particle and personal best information. The subtraction-based solution update mechanism has caused premature convergence and stagnation in particle population during the iterations. To overcome these issues, this study presents a distribution-based update rule for PSO algorithm. The performance of proposed approach named as PSOd is investigated on solving 13 nonlinear global optimization benchmark functions and three constrained engineering optimization problems. Obtained results are compared with standard PSO algorithm, its classical variants and some state-of-art swarm intelligence algorithms. The experimental results and comparisons show that PSOd outperforms PSO and its variants on solving the numerical benchmark functions in terms of solution quality and robustness.