Solving Constrained Optimization via Dual Particle Swarm Optimization with Stochastic Ranking

Abstract

The genetic particle swarm optimization (GPSO) was derived from the original particle swarm optimization (OPSO), which was incorporated with the genetic reproduction mechanisms, namely crossover and mutation. To combine the characteristics of GPSO and OPSO to solve constrained optimization problems, the paper presents a dual particle swarm optimizations (dual-PSO), where OPSO and GPSO were incorporated, which are continuous and discrete editions PSO, respectively. To deal with the constraints, the stochastic ranking algorithm was employed. Based on which Dual-PSO was introduced, where at each generation GPSO and OPSO generated a new position for the particle synchronously and respectively, with the original position of the particle, and the better one was accepted as the new position. Dual-PSO was experimented with well-known benchmark problems with various stochastic ranking parameters, and by comparison with the evolution strategy, the results have shown robust and consistent effectiveness of Dual-PSO.