Guidelines for Parameter Selection in Particle Swarm Optimization According to Control Theory

Taking account of the inherent property of standard particle swarm optimization algorithm, the dynamic characteristics of position expectations on z-plane are studied and the effects of real and complex eigenvalues on the dynamic characteristic are discussed based on control theory. And then a conclusion is drawn that the maximum overshoot and the angular frequency of damped oscillation affect the optimization behavior greatly. The new guidelines for parameter selection in particle swarm optimization are given based on the theoretical qualitative analysis and the massive experimental results. The results which show that the product of the maximum overshoot and the angular frequency of damped oscillation approximately equal to 1 when the angular frequency belongs to (0.65¿, 0.35¿) are the promising guidelines for parameter selection in PSO. Moreover, comparisons are made between research results both in opened literatures and this paper in order to show the validity of the new guidelines.