Learning a fast walk based on ZMP control and hip height movement

Nima Shafii, Nuno Lau, Luís Paulo Reis

Abstract: The Linear inverted pendulum model is widely used in biped walking approaches. This model assumes that the hip height is fixed while the robot walks. In this paper, the hip height movement, or vertical Center of Mass (CoM) trajectory, is used by a robot to achieve a faster and more stable walk. For the first time, the hip height movement is modeled in a formal way and its parameters are learned. The inverted pendulum model and a numerical approach are used to control the Zero Moment Point (ZMP) for generating a balanced walk. Covariance Matrix Adaptation Evolution Strategy (CMA-ES) is applied to optimize the hip height trajectory and walking parameters with respect to walking speed and stability. Experimental results are achieved on a simulated NAO robot. A comparison of the results of the proposed gait model (and development approach) with those obtained using fixed hip height shows that fixed height walking is slower than variable height walking.