Interactive Dynamic Walking: Learning Gait Switching Policies With Generalization Guarantees

摘要

In this letter, we consider the problem of adapting a dynamically walking bipedal robot to follow a leading co-worker based on physical interaction. Our approach relies on switching among a family of Dynamic Movement Primitives (DMPs) as governed by a supervisor. We train the supervisor to orchestrate the switching among the DMPs in order to adapt to the leader&#x2019;s intentions, which are only <i>implicitly</i> available in the form of interaction forces. The primary contribution of our approach is that it furnishes <i>certificates of generalization</i> to novel leader intentions for the trained supervisor. This is achieved by leveraging the Probably Approximately Correct (PAC)-Bayes bounds from generalization theory. We demonstrate the efficacy of our approach by training a neural-network supervisor to adapt the gait of a dynamically walking biped to a leading collaborator whose intended trajectory is not known explicitly.