Abstracto

Central pattern generator control system research of humanoid robot's smooth walking

Xiaodong Tan, Yanlei Lei, Dongxu Liu

This research is combining Central Pattern Generator approach and advantages of passive dynamics. In this research the smooth walking problem of the large mass torso biped walking robot is solved. The core innovation is a direct, robust and efficient control systemwhich is developed through designing a five-level reflex controller. In this paper we introduce some newmethods:Using a simple sine function as output torquewaveform of CPG controller to drive the motor which links the swing leg and the hip joint; Impose secondary incentives to improve the landing performance of swing leg when it is unbending; Design a pedal foot kinetic energy regulation controller according to the different walking of robot; Design a feedforward discretization torso controller to reduce the swinging amplitude of torso. Through simulation tests in Matlab/Simulink, our strategy for large mass torso biped walking is verified.