David Rosin

The core topic of my work is surrogate modelling. More specifically, the project I am working on involves using deep learning models as a means to predict muscle deformation - taken from 3D finite element simulations - based on a set of muscle activation levels, or even motion capture data, for the purpose of on-body visualization in VR/AR environments. This holds the potential to open up 3D continuum-biomechanical simulations to a variety of different application fields in the context of forward simulation of the biomechanical system. This area has historically been reserved for either phenomenological animation models, or so-called Hill-type muscle models in combination with multi-body simulations. These have considerably lower evaluation times, than a continuum-mechanical model, but do not provide any 3D information regarding the simulated muscles, or in the case of animation models, at least not based on any established biomechanical material models for our skeletal muscles. Current modelling efforts are focused on the human upper arm, but the methods developed for this project could be equally applicable for e.g. other extremities or other regions of the musculo-skeletal apparatus.