Rapid musculoskeletal simulations currently use optimization techniques dealing with automatic differentiation (AD) and gradient-based methods. Those simulations typically estimate contact forces based on point contacts, calculating the resulting contact wrench. Mesh-based contact models would provide the details of the contact pressures at the contact interface. However, collision-detection algorithms present some drawbacks when using AD methods and gradient-based optimizations, making the convergence difficult. We developed a smooth mesh-based contact model to be used within musculoskeletal simulations in a reasonable computational time.

Two perspectives are used to estimate the outcome of surgical interventions. Musculoskeletal simulations are developed to estimate and predict the gait patterns with people who undergone with knee arthroplasties. Finite element methods (FEM) are used to estimate the stresses and strains at the knee contact in different types of osteotomies.

Optimal control algorithms are developed to predict muscle and contact forces during human movement. The research lines are focused on clinical and sports biomechanics. Neural networks and other machine learning algorithms are also used to predict dynamic biomechanical variables using only kinematics information.

A telerehabilitation system based on a depth camera, a regular PC and a computational application is developed to facilitate the in-home rehabilitation of post-stroke subjects after the acute phase. The physiotherapist can plan the treatment remotely and the user interacts with the exercises and games through his/her body motion.