Push recovery for humanoid robots with passive damped ankles

In this paper a double inverted pendulum model is proposed for humanoid robots with passive damped ankle joints, which considers the mass of leg and body, and the mass is distributed on the whole link, thus it is more accurate than the traditional Linear Inverted Pendulum Plus Flywheel Model. Based on the new dynamic model, a push recovery controller for standing balance is designed. In this method, the standing balance problem of passive damped humanoid robots is converted into a stabilization problem of linear systems by linearizing the double invert pendulum model, then the linear quadratic regulator (LQR) is used as a standing balance controller and the model´s parameters are estimated by a least square algorithm. Experimental results demonstrate the robustness and effectiveness of the proposed approach on the position controlled humanoid robot "Kong" by exerting different kinds of external force disturbance under different damping coefficients.