Evaluating a reflexive neuromuscular gait model against real-world results

Abstract

The inner workings of human gait are still not well understood. Many models of human gait exist, although no singular model provides insight on human gait behavior and how it alters under disturbances. In an effort to determine if a neuromuscular walking model is capable of predicting human gait under torque disturbances from an ankle-mounted exoskeleton, the results of a prior experiment involving able-bodied subjects with ankle-mounted exoskeletons attached to their feet were recreated in a MATLAB simulation. An optimization-based workflow attempts to reproduce experimental results through selection of optimization objectives based on human gait metrics, producing walking simulations and comparing them with real-world results and establishing a framework which can be applied to future research and analysis. While the optimized simulations do not accurately reproduce human gait as seen in experiments, the gait patterns observed align broadly in some characteristics, and some objective choices produce more convincing results than others. Further testing and more robust statistical analysis is needed to soundly determine whether this model can be predictive of human gait behavior and what minimal set of constraints and optimization objectives is needed to do so.