Improvement of convergence of particle swarm algorithm using learning automata and MPSO model

Particle swarm algorithm is an optimization method based on probability rules. In this method, each particle, in searching space, moves toward the best personal experience and the best group experience. Then, the results are computed on the basis of a competency function. Over time, particles tend to accelerate to the particles which are of higher merit. The main problem of the model is particle trapping in local optimum. In the article, it is proposed a model named MPSO in which a particle is added around the best global particle in order to search more space around the best global, every time the algorithm is executed, then the particle is deleted from the population with the worst expense. In the presented formula for determining the position of the particle, there is a parameter named Alfa that the parameter is selected randomly from the interval of [0,1] in the presented model. In this study, they have been also presented other three models with the names of AMPSO_ LA, AMPSO &LA, and ADMPSO & LA for particle swarm algorithm on the basis of learning automata. The results have shown that the proposed method includes higher convergence velocity and accuracy than similar methods.

Keywords:

Convergence , Particle Swarm Algorithm , Learning Language , Local Optimization