A machine learning approach to quantify individual gait responses to ankle exoskeletons
Published in Journal of Biomechanics, 2023
Aim: This study aims to leverage a neural network-based discrepancy modeling framework to quantify complex changes in gait in response to passive ankle exoskeletons in non-disabled adults. It hypothesized that (i) the Nominal model would predict Exo kinematics and EMG less accurately than for the Nominal condition, and (ii) the Augmented (Nominal+Discrepancy) model would capture greater variance in Exo kinematics and EMG than the Nominal model.
Method: This study analyzed gait data for 12 non-disabled adults during treadmill walking in bilateral passive ankle exoskeletons at a self-selected speed, results of which were used in participant-specific continuous-time neural network with discrepancy models to predict gait responses.
Results: Discrepancy modeling successfully quantified individuals’ exoskeleton responses without requiring knowledge about physiological structure or motor control. However, additional measurement modalities and/or improved resolution are needed to characterize Exo gait, as the discrepancy may not comprehensively capture response due to unexplained variance in Exo gait.
Interpretation: These techniques can be used to accelerate the discovery of individual-specific mechanisms driving exoskeleton responses, thus enabling personalized rehabilitation.