Non-directional Learning Strategy Particle Swarm Optimization Algorithm

Abstract

In conventional particle swarm optimization (PSO) algorithm, each particle adjusts their position and velocity to achieve an optimal solution by iteration, but it has the tendency to fall into the local optimum. In order to avoid the classic PSO problem, a new variant of PSO exemplar based on non-directional learning strategy (NLS) is introduced in this paper, which uses random information of partial dimension of personal best experience in every iteration. Initially, the above method randomly extracts dimensional experience from all dimensions of personal best position of particles. Then, the non-directional position is generated by information of random-dimension. Based on the above mechanism, particles are set to obtain information from personal, population and non-directional position, which can enhance particles search ability. Non-directional learning strategy PSO is tested by several benchmark functions, along with some novel PSO algorithms, and the results illustrate that the convergence accuracy is improved significantly.