Restructuring Particle Swarm Optimization algorithm based on linear system theory

Abstract

The original Particle Swarm Optimization (PSO) used two formulas to describe updating of particle's position and velocity, respectively, based on simulating the foraging behavior of bird swarm. The general improving methods on PSO are to adjust and optimize its parameters or combine new learning strategy to update velocity formula for the better performance. But these methods lack of theoretical analysis and make the algorithm more complex. This paper proposes a new formulation to restructure the particles' position updating behaviors based on linear system theory, and obtain a Restructuring PSO algorithm (RPSO). Compared with the conventional PSO algorithm, RPSO only uses one particle position updating formula, without velocity updating formula, and takes fewer parameters. In order to verify the effectiveness of RPSO, experiments on the CEC 2013 benchmark functions have been conducted to compare with four algorithms, and the final results show that proposed algorithm has a certain degree of competition.