RNN-Based On-Line Continuous Gait Phase Estimation from Shank-Mounted IMUs to Control Ankle Exoskeletons.

Several research groups have developed and studied powered ankle exoskeletons to improve energetics of healthy subjects and the mobility of elderly subjects, or to reduce asymmetry in gaits induced by strokes. To achieve optimal effect, the timing of assistive torque has been proved to be of crucial importance. Previous studies estimated the onset timings mostly by extrapolating the time horizon from past gait events observed with sensors. Such methods have inherently limited performance when subjects are not walking at steady frequencies. To overcome such limitation and allow the use of exoskeletons in various scenarios in a daily life, we propose to estimate the gait phase as a continuous variable progressing over a gait cycle, hence allowing immediate response to frequency changes rather than iteratively correcting it after each cycle. Our method uses recurrent neural networks to estimate gait phases out of an inertial measurement unit (IMU) every 10 ms. By replacing foot sensors with an IMU we can obtain rich enough information to estimate gait phase continuously as well as avoid physical damage in sensors from ground impacts. Our preliminary tests with 2 healthy subjects showed qualitatively positive outcomes regarding the gait phase estimation and the assistive torque control.