Optimal and Robust Walking using Intrinsic Properties of a Series-Elastic Robot

Abstract

Series-Elastic Actuators (SEA) have been proposed as a technology to build robust humanoid robots. The aim of this work is to generate efficient and robust walking for such robots. We present a combined approach which exploits the system dynamics through optimization based trajectory generation and a robust control scheme. The compliant actuator dynamics are explicitly modeled in the optimal control problem. For local stabilization, a passivity based tracking controller distributes the required control forces onto the available contacts. Additionally, a predictive control scheme for step adaptation is presented, which provides feasible contact points in the future. Using a reduced model, this combines efficient walking with robustness against model or environment uncertainties and external disturbances.