Robert Marzilger (Frauenhofer IIS), Prof. Dr. Anne Koelewijn
06/2021 – 12/2021
A Musculoskeletal model used for simulation is complex and has to be preselected and
adapted to the subject to reflect the subject’s body properties. Models are typically scaled
using motion capture systems  and simulation software like OpenSim . Based on
capturing optical reference markers on specific landmarks, the segment lengths of the
subject can be calculated in these software. However, muscle parameters are usually not
adapted individually to the subjects. An estimate of individual muscle forces can be
achieved by measuring ground reaction forces and the application of invers kinematics
. The current research project aims to combine personalized segment and muscle
parameters based on marker data and ground reaction forces using optimal control
simulation [3, 4], with the assistance of advanced machine learning (ML) methods. This
will allow for personalized simulations in healthcare, sports science, and industrial practice
[5, 6]. However, the whole personalization process is currently a time-consuming and
expensive process requiring specialized equipment and expertise. The final goal of the
project is to apply the selection of individual model input parameters automatically with
low computational effort and to make additional expensive measurements unnecessary.
As a part of the research project, this master thesis aims to improve parameter selection
and processing time of the model parameter optimization / personalization. In a first step,
this thesis should identify meaningful features of human walk and running gait and use
those for clustering subjects of already existing data according to their meta data (age,
gender, body height, body and mass) as well as marker and ground reaction force data.
It can be expected, that persons with similar meta, marker and ground reaction force data
can have similar walking and running biomechanics thus resulting in a similar model
output. In a following step, the identified cluster should be used to define the range /
borders for the optimal control simulation parameters for specific persons. Thus, using a
pre-clustering method the range for the input parameters of the optimal control
simulation can be reduced, resulting in an overall shorter time for the hyper-parameter
search for the optimal model parameters. In a last working package a leave-one-out cross
validation should help to validate the clustering quality.
1. M. E. Lund, M. S. Andersen, M. de Zee and J. Rasmussen, “Scaling of musculoskeletal
models from static and dynamic trials,” International Biomechanics, vol. 2, no. 1, 2015.
2. S. L. Delp, F. C. Anderson, A. S. Arnold, P. Loan, A. Habib, C. T. John, E. Guendelman and
D. G. Thelen, “OpenSim: Open-Source Software to Create and Analyze Dynamic Simulations
of Movement,” IEEE Transactions on Biomedical Engineering, vol. 54, no. 11, p. 1940–1950,
3. Gail, T., Hoffmann, R., Miezal, M., Bleser, G., & Leyendecker, S. (2015). Towards bridging
the gap between motion capturing and biomechanical optimal control simulations.
Proceedings of the ECCOMAS Thematic Conference on Multibody Dynamics 2015,
Multibody Dynamics 2015, 1080–1091.
4. Felis, M. L., & Mombaur, K. (2016). Synthesis of full-body 3-D human gait using optimal
control methods. Proceedings – IEEE International Conference on Robotics and Automation,
5. Dorschky, E., Krüger, D., Kurfess, N., Schlarb, H., Wartzack, S., Eskofier, B. M., & van den
Bogert, A. J. (2019). Optimal control simulation predicts effects of midsole materials on
energy cost of running. Computer Methods in Biomechanics and Biomedical Engineering,
6. D. Koelewijn. (2018). Predictive simulations of gait and their application in prosthesis design,
7. A. Seth, M. Sherman, J.A. Reinbolt, S.L. Delp. (2011) OpenSim: a musculoskeletal modeling
and simulation framework for in silico investigations and exchange, Procedia IUTAM,
Symposium on Human Body Dynamics, 2, 212-232. 10.1016./j.piutam.2011.04.021