Researchers manage to perform biomechanical simulations up to 20 times faster

+
Download

The method developed has been used to simulate pendulum movement, human gait and subject-exoskeleton collaborative movement.

A team of researchers from the Katholieke Universiteit Leuven (Belgium), Stanford University (United States) and the UPC have improved the efficiency of biomechanical simulation software systems to perform simulations up to 20 times faster. The results of this research have been published in the journal ‘PlosOne’.

Feb 11, 2020

The researchers Gil Serrancolí, from the Department of Mechanical Engineering of the Universitat Politècnica de Catalunya · BarcelonaTech (UPC), and Antoine Falisse, from the Katholieke Universiteit Leuven (KU Leuven, Belgium), have implemented automatic differentiation tools to the OpenSim and Simbody biomechanical software systems and thus managed to perform biomechanical simulations up to 20 times faster than using finite difference methods.

Automatic differentiation tools calculate derivatives of mathematical expressions more efficiently than finite difference tools. They are very useful for solving optimisation problems that require calculating thousands or millions of derivative evaluations. Biomechanical simulations that aim to calculate optimal trajectories are usually based on solving an optimal control problem (dynamic optimisation). The authors have applied the method developed to simulate pendulum movement, human gait and subject-exoskeleton collaborative movement.

Collaborating with Stanford
In 2017, Falisse and Serrancolí were chosen as
OpenSim Visiting Scholars to visit Stanford University for a period of a month to develop their project. The results of this work, led by Friedl De Groote (KU Leuven), have been published in the journal PlosOne. The paper compares the efficiency of automatic differentiation to that of finite difference in biomechanical simulation optimisations.

Gil Serrancolí, a researcher from the Simulation and Movement Analysis Lab (SIMMA Lab) and a professor at the Barcelona East School of Engineering (EEBE), has used this method to estimate the parameters of subject-exoskeleton and foot-ground contact models, and subsequently predict new subject-exoskeleton collaborative movements. This study is part of his postdoctoral at KU Leuven, the results of which have been published in the journal IEEE Transactions on Neural Systems and Rehabilitation Engineering.

Antoine Falisse, a researcher from KU Leuven, has used this new method to predict human gait and study control strategies. The results have been published in the Journal of the Royal Society Interface.