Algorithm Design for Kicking by Bipedal Robots Based on Improved CMA-ES

In response to the issue of slow attack speed and low success rate of football robots' kicking in dynamic environments during RoboCup competitions, an algorithm for bipedal robot kicking was devised using an enhanced Covariance Matrix Adaptation Evolution Strategy (CMA-ES) algorithm. To tackle the problem of CMA-ES getting trapped in local optima, a combined approach of Tent chaotic mapping and Lévy flight random numbers was introduced to widen the search range and increase the population size, thus enhancing the algorithm's ability for global exploration. The refined CMA-ES algorithm was employed to optimize kicking parameters, with parameter feasibility assessed using inverse kinematics. Moreover, a kicking cost function was utilized to determine optimal kicking point coordinates, and trajectory optimization was performed using Bézier curve interpolation. Finally, the correctness and effectiveness of the method were validated through simulation experiments using SimRobot and physical experiments with NAO robots.