A hybrid particle swarm optimization approach with prior crossover differential evolution

Particle swarm optimization (PSO) is population-based heuristic searching algorithm. PSO has excellent ability of global optimization. However, there are some shortcomings of prematurity, low convergence accuracy and speed, similarly to other evolutionary algorithms (EA). To improve its performance, a hybrid particle swarm optimization is proposed in the paper. Firstly, the average position and velocity of particles are incorporated into basic PSO for concerning with the effect of the evolution of the whole swarm. Then a differential evolution (DE) computation, which introduces an extra population for prior crossover, is hybridized with the improved PSO to form a novel optimization algorithm, PSOPDE. The role of prior crossover is to appropriately diversify the population and increase the probability of reaching better solutions. DE component takes into account the stochastic differential variation, and enhances the exploitation in the neighborhoods of current solutions. PSOPDE is implemented on five typical benchmark functions, and compared with six other algorithms. The results indicate that PSOPDE behaves better, and greatly improve the searching efficiency and quality.