The improved particle swarm optimization based on swarm distribution characteristics

Due to the deficiency of characteristics of objective function, such as the function derivative, the solutions can only be iterated according to the evolutionary equations of Particle Swarm Optimization (PSO) with the finite information about current swarm state. But in the evolutionary process of PSO, the distribution characteristics of solutions of the objective function are hidden in the many and many fitness evaluations while the evolutionary equations are iterating. The evolutionary strategies, including the balance strategy between the exploration and exploitation, the re-initialization strategy and the generation strategy of new solution from the elite particles, are designed innovatively according to the distribution characteristics of the swarm solutions extracting statistically from the historical evaluations. The experimental results show that these strategies are effective for the optimization precise and efficiency in the early evolutionary process although the complexity of time and space are increased lightly than that of the standard PSO.