Adaptive control of biped robots based on Gennet

Abstract

The dynamic walking stability of robots is a key to biped robots study. Due to the uncertainty of outside environment, adaptability has become a priority in the research of dynamic walking of robots. In order to realize the dynamic walking of robots, a Genetic Neural Network is used to develop an Adaptive Control Neural Network. A better walking result is realized by the Neural Network trained by abnormal walking samples and self-adaptive fitness formulation.