An evolution strategy with competing subpopulations
Keisuke Izumi, M. M. A. Hashem, Keigo Watanabe · 2002
Evolution strategies, based on the natural evolution, are the algorithms to solve the parameter optimization problems numerically. In this paper, a new evolution strategy (ES) is proposed to solve optimal control problems. With a view to making a balance between exploration and exploitation, a competing subpopulation based arithmetical crossover technique is proposed. The effectiveness of the proposed ES is illustrated by some simulations for the push-cart (a discrete-time optimal control model) control problem.