EMG Control of Powered Prostheses

Powered knee prostheses need to coordinate their motion with the user.  Most powered prosthesis controllers use mechanical sensors to estimate the user's intention, but with the wide range of activities performed in daily life, estimating the desired motion in every scenario becomes very complex. The EMG control project aims to simplify this problem by enabling users to control their knee prosthesis directly using the muscles of their residual limb. Publication in progress. This work builds on our previous EMG control work, Shared Neural Control: Sit-to-Stand, Squat, Lunge.

Unified Controller for Ankle Prostheses

The Unified Controller for powered foot and ankle prostheses is built upon intrinsic biomechanical behaviors. Primarily, the virtual “stiffness” of the ankle joint is commanded as a function of the shank angle, enabling natural adaptation to uneven terrain and variable pushoff power based on step length. Furthermore, the virtual “damping” of the joint is commanded as a function of walking speed, enabling increased pushoff the faster you walk. Lastly, the position of the ankle is controlled synergistically with the knee, enabling biomimetic coordination between the joints of the lower leg. In combination, these control elements enable walking at variable speed and inclines, climbing stairs, sit-to-stand, and other activities of daily living. Check out our journal paper here, and two related conference papers here and here.

Unified Control for Powered Knee Prostheses

The Unified Controller for Powered knee prostheses is a novel control strategy that enables walking, stair climbing, sit-to-stand, and other activities of daily living without classifying the intended activity or ambulation mode. Instead, the controller commands desired torques and positions based on the user’s movement of their residual thigh. We achieve this by combining a series of concurrently acting control elements that modulate the knee control as a function of the device’s sensor measurements. These control elements, such as knee-thigh synergy, minimum-jerk trajectory planning, and biarticular torque are biomimetic behaviors that enable natural ambulation across multiple activities without classification. Check out our journal paper here, and two related conference papers here and here.

Shared Neural Control for Powered Prostheses: Sit-to-Stand, Squat, Lunge

We developed a novel control approach for powered prostheses that combines the neural commands from the user's residual hamstring muscle with the mechanical sensors in the powered leg. The knee extension torque is directly controlled by the user's residual hamstring contraction, enabling controlled lifting and lowering in a way that is both safe and intuitive. The direct control was combined with our minimum-jerk controller for swing to enable walking. Using this approach, two above-knee amputees were able to perform stand-up, sit-down, squatting, lunging, walking, and transitions between activities. Read the full open-access paper here. This controller was extended to perform stair ascent and descent - check out the conference paper here. Our EMG control work continues, read about our newest project: EMG Control of Powered Prostheses

Adaptive Swing Control and Obstacle Crossing - Powered Knee-Ankle Prostheses

We developed a novel controller for the swing phase of gait. The controller uses the minimum-jerk principle, which creates a smooth and adaptable joint angle trajectory that minimizes the jerk (the derivative of acceleration) of the angle. The controller also enables natural obstacle crossing, where the prosthesis flexes more when the user raises their residual thigh to a higher angle. Using this method, above-knee amputees were able to cross obstacles of different sizes, and their compensatory movements were reduced. This paper was published in Science Robotics! Check out the full paper here.