Efficient musculoskeletal simulations including mesh-based contact pressure models
Abstract: Full-body musculoskeletal simulations are essential for understanding human movement dynamics, solving for kinematics, muscle forces, and external forces, like ground reaction forces, concurrently. While these simulations yield valuable insights, this information might be limited if one needs to predict full-body movements with a criterion involving detailed joint pressures. This could be crucial to study the effect of a surgery or a rehabilitation treatment on the full-body forces in silico. Optimal control methods have emerged as powerful tools for solving musculoskeletal simulations. These could be used as tracking simulations, to track a measured movement and estimate model parameters, or as predictive simulations, to predict new movements and their involved forces. Direct collocation, a common optimization technique, formulates the problem as a nonlinear program problem by discretizing states and controls over time. Typically, gradient-based optimizations are employed, necessitating continuous derivatives for proper convergence. Automatic differentiation has proven invaluable in accelerating these computations, enabling predictive simulations to be completed within 30 minutes on a standard laptop.
In this presentation, we will explore the potential of musculoskeletal simulations and the utilization of a mesh-based contact model to estimate joint contact pressures within these simulations. By incorporating detailed joint pressure information into the cost functions or constraints of the optimization problem, we can enhance the accuracy of predictions. Furthermore, we will discuss the potential application of this approach in simulating treatments for conditions like osteoarthritis.
Biosketch: Gil Serrancolí is Associate Professor, Serra Húnter Fellow, at the Universitat Politècnica de Catalunya (UPC). He is coordinating the Simulation and Movement Analysis Lab within the Department of Mechanical Engineering. He did the Degree in Industrial Engineering (especialized in Mechanics, 2010), and Msc in Biomedical Engineering (2011). He obtained his PhD in Biomedical Engineering in 2015 in UPC. He worked as postdoc in KU Leuven (Belgium, 2015-2016). He also did other research stays at University of Florida (USA, 2012-13), Stanford University (USA, 2017) and Northwestern University (USA, 2022). He was OpenSim Visiting Scholar in 2017, and is OpenSim Fellow since 2020.
His main interests are in developing more efficient musculoskeletal simulations, enhancing their capabilities and improving computational speed. He contributed in developing a smooth foot-ground contact model and in speeding musculoskeletal simulations with the use of automatic differentiation. Currently, he is focused on developing a mesh-based contact pressure model to incorporate joint pressures within the decision criteria to predict movements. In parallel, he is also contributing integrating biomechanical simulations into Muvity, a telerehabilitation system for people with mobility impairements.
