Predictive modeling of human locomotor response to ankle exoskeletons

Ankle exoskeletons (exos) can augment human locomotion, but modeling subject-specific responses to exo torques is essential to optimize their design. Existing subject-specific modeling approaches are experimentally cumbersome or contain assumptions that limit model accuracy. We investigated the potential of non-physiological data-driven models to predict response to exo torques during walking in healthy adults. Phase-varying and linear phase-varying models predicted kinematics similar to, and muscle activity better than, a neural network. Phase-varying models of gait may be useful in optimizing exos for individuals with diverse abilities.