Metastable legged locomotion: Methods to quantify and optimize reliability

Abstract

Measuring the stability of highly-dynamic bipedal locomotion is a challenging but essential task for more capable human-like walking. By discretizing the walking dynamics, we treat the system as a Markov chain, which lends itself to an easy quantication of failure rates by the expected number of steps before falling. This meaningful and intuitive metric is then used for optimizing and benchmarking given controllers. While this method is applicable to any controller scheme, we illustrate the results with two case demonstrations. One scheme is the now-familiar hybrid zero dynamics approach and the other is a method using piece-wise reference trajectories with a sliding mode control. We optimize low-level controllers, to minimize failure rates for any one gait, and we adopt a hierarchical control structure to switch among low-level gaits, providing even more dramatic improvements on the system performance.