Mode-free Control of Prosthetic Lower Limbs

Abstract

Intuitive control of powered prosthetic lower limbs is still an open-ended research goal. Current controllers employ discrete locomotion modes for well-defined scenarios such as stair ascent, stair descent, or ramps. General human locomotion, however, is a continuous motion, fluidly adapting to the environment, and not always categorizable into modes. A prominent feature of normal movement is that it exhibits strong inter-joint coordination, and the movement of a single joint, such as the ankle, can be largely predicted based on the movement of the rest of the body. We show that using body motion from the intact limbs and trunk, a reference trajectory can be generated for a prosthetic joint for every instant in time.A wearable motion capture suit was worn by 11 healthy subjects to record full body kinematics during flat ground walking with and without random stops and stair ascent and descent. Three machine learning techniques were employed to predict right ankle kinematic trajectory using other joint kinematics as inputs. We found that a Recurrent Neural Network (RNN) was the best performing model, robust to subject-specific variations such as walking speed and step length. Performance of the network evaluated on a test subject (not included in training) showed that ankle angle can be estimated with a root mean square error of fewer than 7 degrees. The change in performance when using a smaller array of sensors, gathering partial body kinematics instead of full-body motion, is also evaluated. This approach demonstrates the potential for the application of data-driven models to prosthetic control without explicit featurization of terrains or gait phases.