Orthogonal Latin Square Based Particle Swarm Optimization: A Dynamic Approach for Continues Functions Optimization

Particle Swarm Optimization (PSO) is one of the best solutions that can find the optimal solution to an optimization problem. While, the canonical PSO may suffer from the problems that it can trap in a local optima during some successive generations from the beginning to the end of the stopping criteria. To make the population of PSO escape from the local optima quickly, an Orthogonal Latin Square (OLS) based experimental design method is applied and can significantly improve the convergence performance. In this article, an OLS based PSO algorithm is proposed which can be used if the PSO algorithm traps into a local optima during the convergence time. In the iteration, the OLS based design method will be implemented to help the population escape to better positions only when the population falls into local optima. And the best so-far position can be replaced by a better one around it. In this way, a shortened convergence time is shown compared with the canonical PSO algorithm as well as a better final best fitness value. When applying our proposed algorithm in the test functions, the result shows that our proposed algorithm can find the optimal solutions with an excellent convergence time.