Walking pattern generation for Humanoid robots with LQR and feedforward control method

Abstract

This paper proposes new walking pattern method for the humanoid robots. The proposed method based on the inverted pendulum uses the zero moment point(ZMP) and the center of mass(CoM). And our approach consists of feedforward controller and feedback controller. As the feedforward controller, pole-zero cancellation by series approximation(PZCSA) control plays a role of reducing the inherent property of inverted pendulum model, that is, non-minimum phase property due to the unstable zero. As the feedback controller, linear quadratic regulator(LQR) is utilized for tracking effectively the desired ZMP according to footprints of the humanoid robots. The proposed method generates the walking pattern with the combination of LQR and PZCSA. The efficiency of the proposed method is verified by simulation of generating the walking pattern.