Predicting the mechanics and energetics of a variety of human gaits based on complex musculoskeletal models

Predictive simulations have the potential to support personalized medicine, but detailed musculoskeletal models come with large computational costs that limit their use. We have developed a computationally efficient framework to predict human gaits based on optimization of a performance criterion. Our three-dimensional muscle-driven predictions converge in about 36 minutes—more than 20 times faster than existing simulations—using direct collocation, implicit differential equations, and algorithmic differentiation. Our framework predicts physiologically realistic gaits with varied gait speed, as well as gait changes due to muscle strength deficits or lower leg prosthesis use. The ability to predict gait mechanics and energetics with realistically complex models of the musculoskeletal system will allow testing novel hypotheses about gait control and hasten the development of optimal treatments for neuro-musculoskeletal disorders.