Realizing Torque Controllers for Underactuated Bipedal Walking Using the Ideal Model Resolved Motion Method

Abstract

This thesis presents an application of hybrid zero dynamics to realize underactuated bipedal walking on DURUS, a testbed designed and built by SRI International. The main contribution of this work is the ideal model resolved motion method (IMRMM), which is a simple method to convert ideal torque controllers to PD controllers to implement on hardware. Walking was first achieved using the proven method of the hybrid zero dynamics (HZD) reconstruction, followed by the Input-Output Feedback Linearization (IO) and Rapidly Exponentially Stabilizing Control Lyapunov Function Quadratic Programs (CLF-QPs) torque controllers implemented via IMRMM. The simulation and experimental results are presented and compared, and the best resulting specific cost of electrical transport on hardware was computed as 0.63 for the CLF-QP IM-RMM controller, and the record for walking was achieved on a separate occasion with the same CLF-QP IM-RMM controller, which yielded walking for 2 hours and 53 minutes, covering 7 km.