Towards Variable Assistance for Lower Body Exoskeletons

Thomas Gurriet, Maegan Tucker, Alexis Duburcq, Guilhem Boeris, Aaron D. Ames

Seminal September 2019

TL;DR A control-barrier-function filter lets an exoskeleton wearer move the swing-leg hip and knee freely inside a tube around the nominal gait and intervenes only at the tube boundary. On the Atalante exoskeleton with eight able-bodied subjects, lowering the assistance (a wider tube) raised the metabolic power the wearer had to contribute.

What they do

Crutch-less exoskeletons normally track a fixed gait rigidly, which is not ideal for patients who are recovering muscle function and should do part of the work themselves. This paper adds a variable assistance layer to the Atalante exoskeleton (12 actuated joints): the wearer controls the sagittal hip and knee of the swing leg, while a safety filter keeps those joints inside a tube, the virtual guide, around the nominal trajectory. An assistance factor sets the tube width, from 1 (the baseline rigid controller) to 0 (lowest assistance).

How it works

Baseline walking. Gaits generated with Partial Hybrid Zero Dynamics over a grid of patient and gait parameters are fitted by a neural network, which outputs joint trajectories tracked by PID. All non-assisted joints stay under this controller.
Transparency. On the assisted joints, joint idealization compensates gravity and friction, and feedforward torques from gait generation partly compensate the exoskeleton’s inertia.
Virtual guide filter. Each assisted joint is treated independently. A control barrier function is evaluated online by integrating the joint dynamics under a backup PD law for the two extreme user torques, so the joint stays in the tube despite user perturbations and the filter acts only when needed.
Haptic feedback. Eight vibration motors on the thighs and shanks vibrate harder as a joint approaches the guide boundary.

Results

On the empty exoskeleton, joint angles stayed inside the bounds over 30 steps for every tube shape tested.
Eight able-bodied subjects (58-91 kg, 1.62-1.93 m) walked five-minute trials at full assistance and at an assistance factor of 0.5 (a tube of plus or minus 4 degrees). Trajectories always stayed within the guides. Metabolic rate rose only when subjects actively tracked the gait under partial assistance; active subjects at full assistance spent about as much as passive ones.
One subject swept assistance factors 1.0, 0.75, 0.5, 0.25, 0.0 and back on three consecutive days: metabolic rate and tracking error both increased as assistance decreased, symmetrically around 0.0, so the increase is not due to exhaustion.

Limitations

Tested only with able-bodied subjects, not in clinical rehabilitation. At low assistance factors an operator standing behind the exoskeleton had to help keep it stable. Subjects found the gaits hard to track because they are not very anthropomorphic, and the haptic feedback was sometimes hard to interpret.

Figures

From Figure 1: The eight able-bodied subjects who took part in the experiments, each wearing the Atalante exoskeleton.

From Figure 3: The baseline walking approach: a patient-exoskeleton model, a library of generated gaits, a neural net... — From Figure 3: The baseline walking approach: a patient-exoskeleton model, a library of generated gaits, a neural network fitted to the library, and joint trajectory tracking.

From Figure 4: Architecture of the variable assistance framework. The gait network and deadbeat filtering feed both t... — From Figure 4: Architecture of the variable assistance framework. The gait network and deadbeat filtering feed both the baseline controller and the variable assistance controller (virtual guide filter, feedforward assistance, joint idealization), whose torques are merged.

From Figure 10: Metabolic rate (left) and tracking error (right) of one subject for assistance factors from 1.00 down... — From Figure 10: Metabolic rate (left) and tracking error (right) of one subject for assistance factors from 1.00 down to 0.00; both rise as assistance decreases.

Abstract

This paper presents and experimentally demonstrates a novel framework for variable assistance on lower body exoskeletons, based upon safety-critical control methods. Existing work has shown that providing some freedom of movement around a nominal gait, instead of rigidly following it, accelerates the spinal learning process of people with a walking impediment when using a lower body exoskeleton. With this as motivation, we present a method to accurately control how much a subject is allowed to deviate from a given gait while ensuring robustness to patient perturbation. This method leverages control barrier functions to force certain joints to remain inside predefined trajectory tubes in a minimally invasive way. The effectiveness of the method is demonstrated experimentally with able-bodied subjects and the Atalante lower body exoskeleton.

Summaries are written automatically from each paper's text. Figures and videos come from the paper and its project page; all credit goes to the authors.

Matt Strong