Adaptive Particle Swarm Optimizer: Response to Dynamic Systems through Rank-based Selection

All Papers

A response method to dynamic changes based on evolutionary computation is proposed for the particle swarm optimizer. The method uses rank-based selection to replace half of the lower fitness population with the higher fitness population, when changes are detected. Time-varying values for the acceleration coefficients are proposed to keep a higher degree of global search and a lower degree of local search at the beginning stages of the search. Performance is compared with two previous response methods using the parabolic De Jong benchmark function. Experimental results on the function with varying severity and dynamic change frequency is analyzed.