  Home

  Research

    Principles

    Motor Control

  * Robotics

    Projects

  Publications

  Students

  Teaching

Application to
Rehabilitation Robotics

For robots and rehabilitation devices alike, leg control often relies on matching motion patterns that have been recorded from human subjects performing the target tasks. While sufficient for these tasks, such a control approach denies artificial legs the agility and adaptiveness that human legs display in normal locomotion.

The understanding of the interaction between legged dynamics and motor control could trigger more human-like artificial leg behavior. The strength of the neuromuscular human model introduced in the section on motor control is its reflexive control which auto-adapts to environmental changes by taking advantage of principles underlying legged system dynamics. Applying this control to artificial legs that interact with amputees or patients should allow these persons a greater agility and adaptiveness in locomotion than current artificial legs provide.

Impact of Human-like neurocontrol on Rehab Robotics

Figure 1

Control of a powered ankle-foot prosthesis that inherits auto-adaptiveness of neuromuscular model

To test this hypothesis, we implemented part of the human model's ankle control on a powered ankle-foot prosthesis. The prosthesis is one in development by iWalk, LLC, and is a successor to a series of prototypes developed in the Biomechatronics Group at MIT. For more details on the prosthesis, please follow this link to the MIT Biomechatronics Group.

Figure 2

We found in clinical trials with a transtibial amputee walking on level ground, ramp ascent and ramp descent, that an adaptation in prosthetic ankle work occurred in response to the ground slope, in a manner comparable to intact subjects, without the difficulty of explicit terrain sensing.

auto-adaptation to level and ramp walking

Figure 3

The results demonstrate the potential that applying neuromuscular control strategies has to the control of legs in rehabilitation robotics. But the actuated ankle presents just a first step in that direction. Future prosthetic and orthotic devices, as well as the legs of spinal cord injured patients, will have to be controlled concerting several joints. I am interested in developing such articulated leg controls based on neuromuscular control strategies.

More details about this research can be found in

MF Eilenberg, H Geyer, HM Herr. Control of a Powered Ankle-Foot Prosthesis Based on a Neuromuscular Model. IEEE Trans Neural Syst Rehabil Eng. 18(2): 164-173, 2010. [PDF]
HM Herr, H Geyer, MF Eilenberg. Model-Based Neuromechanical Controller for a Robotic Leg. US Patent Application 20100324699. [Ask me for details]

[back to top]

Extension to Robotic Limbs with Ankle, Knee and Hip

Our recent focus is on generalizing the reflex-like control approach to multi-joint robotic limbs. We have developed a robotic gait testbed to evaluate the aproach for robotic limbs with hip, knee and ankle joints. Figure 4 shows a single leg which is about half the size of a human leg. In tests with this leg, we have found that the robust swing leg placement control explored for the neuromuscular model work translates well to the robotic leg hardware.

Powered Robotic Limb with Human-like Actuation Morphology

Figure 4

This observation has led to our current work on the design and control of a highly powered knee-and-ankle prosthesis for balance recovery in amputee gait after large disturbances. Figure 5 shows simulation examples of the prosthesis prototype using different controllers in an amputee gait model. The reflex-like control (A) outperforms impedance control (B) in stumble and trip recovery. The figure also shows our current steps toward hardware realization.

Figure 5