Experimental Evaluation of Robust Swing-Leg Placement Controls in Robotic Limb Testbeds - Robotics Institute Carnegie Mellon University

Experimental Evaluation of Robust Swing-Leg Placement Controls in Robotic Limb Testbeds

Tech. Report, CMU-RI-TR-18-62, Robotics Institute, Carnegie Mellon University, August, 2018

Abstract

Variable swing-leg placement is crucial for robust and stable locomotion. While appropriate leg placement is generated in humanoid robotics by tracking pre-planned swing motions, this approach is not suitable for controlling powered limbs like prostheses, which form only part of the human-robot locomotor system. In recent work, we proposed swing controls for a double pendulum leg model that predicts variable and robust leg placement in the presence of large gait disturbances. Here we evaluate the performance of these novel controls on robotic leg testbeds. In hardware experiments we find that the control achieves robust foot placement of anthropomorphic robotic limbs within a range of landing leg angles observed in human gait. We further observe that a version of this control that conforms with the constraints of neuromuscular dynamics generates a more human-like leg behavior at the cost of placement precision. The results suggest that the proposed controls transfer well to robotic limb systems; in particular, they may enable improved balance recovery in amputee locomotion with powered leg prostheses.

BibTeX

@techreport{Schepelmann-2018-107386,
author = {Alexander Schepelmann and Yin Zhong and Jessica Austin and Kathryn A. Geberth and Hartmut Geyer},
title = {Experimental Evaluation of Robust Swing-Leg Placement Controls in Robotic Limb Testbeds},
year = {2018},
month = {August},
institute = {Carnegie Mellon University},
address = {Pittsburgh, PA},
number = {CMU-RI-TR-18-62},
keywords = {Legged Locomotion, Swing-Leg, Humanoid, Prosthesis, Neuromuscular Control},
}