PredSim: A Framework for Rapid Predictive Simulations of Locomotion

Predictive simulations are a powerful tool for advancing our understanding of the neuromechanics of movement. By generating gait patterns using a (neuro-) musculoskeletal model, without relying on motion capture data, we can study the influence of neuro-musculoskeletal features on gait mechanics and energetics, and accelerate the design of orthopedic devices. However, computational challenges have historically limited the complexity of models used for predictive simulations. Advancements in computational approaches have enhanced the numerical stability of simulations by using direct collocation instead of direct shooting, implicit formulation of dynamics, and algorithmic differentiation to calculate derivatives required by gradient-based solvers exactly, rather than finite differences. We leveraged these advancements to predict gait based on complex neuro-musculoskeletal models. Although all code was made public, adapting the simulations for different research needs often required significant implementation efforts. We have created PredSim based on the simulation framework previously developed in our research group. PredSim was designed to be user-friendly for new users and efficient and convenient for experienced users. The PredSim framework allows for adjusting various simulation aspects without the need for direct involvement with the underlying code. To demonstrate the capability of adjusting muscle-tendon parameters in PredSim, we simulated walking with varying Achilles tendon stiffness. Additionally, we utilized the Orthosis interface, a novel feature, to simulate walking with different exoskeletons that support ankle plantarflexion. PredSim is hosted on GitHub to encourage researchers to explore and contribute to its development.