Orthogonal Permutation Particle Swarm Optimizer with Switching Learning Strategy for Global Optimization

摘要

This paper aims to improve the performance of original particle swarm optimization (PSO) so that the consequent method can be more robust and statistically sound for global optimization. A variation of PSO called the orthogonal permutation particle swarm optimization (OPPSO) is presented. An orthogonal permutation strategy, based on the orthogonal experimental design, is developed as a metabolic mechanism to enhance the diversity of the whole population, where the energetic offspring generated from the superior group will replace the inferior individuals. In addition, a switching learning strategy is introduced to exploit the particles' historical experience and drive individuals more efficiently. Seven state-of-the-art PSO variants were adopted for comparison on fifteen benchmark functions. Experimental results and statistical analyses demonstrate a significant improvement of the proposed algorithm.