Analysis and control techniques for the compass gait with a torso walking on stochastically rough terrain

Abstract

Dynamic walking gaits which exploit inverted pendulum dynamics have demonstrated significant promise for biped robot locomotion. For example, these gaits can reduce the energy expended and the number and complexity of actuators required for level-ground walking. However, robot walkers employing dynamic gaits are, in general, also notoriously sensitive to terrain variations. In this paper, we focus on new methods for developing improved control strategies for and analyzing resulting stability of a simple yet effective model for biped walking on rough terrain. Our primary contributions are as follows. (1) We quantify the stabilizing value of adding a torso to the standard compass gait model; (2) we optimize a class of simple controllers on this walker to be robust to unsensed changes in upcoming terrain height; and (3) we develop improved numerical tools for estimating the statistics of fall events for rough terrain walking. Our results indicate that the torso walker can handle unanticipated step changes in terrain of approximately 14% of leg length, and that our statistical tools are effective for a 6-dimensional state space system, indicating promise in the challenge of addressing the curse of dimensionality when applying machine learning techniques to rough terrain walking.