Improved particle swarm optimization through orthogonal experimental design

In the particle swarm optimization (PSO), each particle is enhanced based on its own best experience and one of the local or global best particle in local or global particle swarm optimization (LPSO or GPSO). In this paper, an orthogonal learning (OL) technique is proposed that mixes these experiences as a new combined algorithm that is named MOLPSO. MOLPSO is the result of mixed two algorithms OLPSO-L and OLPSO-G through orthogonal experimental design (OED). This technique can construct a more effective leadership vector to lead particles toward the best area by selecting better dimensions of these experiences. This technique is tested on a set of some benchmark functions that the results of tests confirm that the strategy significantly enhances the performance of PSO.