Skip to Main Content
Abstract-This article presents an original adaptation of particle swarm for solving high dimensional optimization problems with discrete variables. The proposed method combines Particle Swarm Optimization (PSO) theory with both Variable Neighbourhood Search (VNS) and Stretching Technique (ST) principles. A dedicated mathematical formalism used to handle real-coded discrete variables is defined previously to the theoretical background section which ends with a full description of the developed hybrid local/global optimization algorithm (Jumping PSO + VNS local heuristic + ST). Its performances for solving low and high combinatorial unconstrained optimization problems are evaluated. These latter are obtained from a specific benchmark composed of several reference multimodal functions whose search domains have been disadvantageously discretized. Some comments and perspectives dealing with the future works conclude this paper.