Turning a Corner in Optimal Control Simulations for Running: Predicting and Reconstructing 3D Full-Body Biomechanics from Optical and Inertial Motion Capture

Many sports disciplines performed for recreation or competition involve change-of-direction running, such as running on a curved path or cutting. Change-of-direction running can induce high demands on the musculoskeletal system due to rapid directional changes and fast decelerations. Thus, the execution is fundamental for injury risk but, at the same time, determines an athlete’s performance. Reconstruction of a measured motion can give detailed insights into biomechanical variables that are not measured directly, such as joint moments. Additionally, predicting human responses allows studying conditions that cannot be measured easily or could harm the participant. Optimal control enables simulations of a human model for movement reconstruction and prediction to analyze kinematics and kinetics comprehensively. However, optimal control simulations were limited to 2D models of the lower body or straight walking and running. Therefore, this thesis aims to develop and investigate novel methods to generate optimal control simulations, especially for change-of-direction running. The foundation of this cumulative thesis comes from three publications. In the first publication, I, together with my co-authors, created efficient optimal control simulations of a 3D full-body musculoskeletal model using implicit dynamics and trajectory optimization. We demonstrated that curved running could be predicted from a random initial guess in 1 h 17 min by prescribing a curved motion path. While the predicted kinematics and kinetics matched reference data well, our work revealed that interlimb collisions should be modeled to predict change-of-direction movements robustly. In the second publication, we developed reconstructive simulations to estimate 3D kinematics and kinetics from optical motion capture for arbitrary running movements. We tracked measured marker data directly in addition to ground reaction forces in the simulations to avoid error propagation and accumulation associated with tracking estimated generalized coordinates. We showed that tracking marker positions instead of generalized coordinates reduced the marker error by around half but required longer to solve. In the third publication, my co-authors and I designed reconstructive simulations tracking linear acceleration and angular velocity from inertial motion capture to facilitate the assessment of 3D kinematics and kinetics outside the motion lab. Our evaluation with virtual inertial data proved the feasibility and potential accuracy of our proposed optimal control problem formulation, as the reconstructed simulations were able to distinguish between individual movement repetitions. Furthermore, this cumulative thesis presents an extensive literature overview and provides broader perspectives on the contributions of the included publications. In conclusion, in this thesis, I addressed the challenges of obtaining a comprehensive biomechanical analysis for change-of-direction running by researching optimal control simulations. Future work should evaluate the proposed methods in the context of applications, such as studying cutting maneuvers on the sports field before and after injury prevention training. Additionally, further technological advancements are possible when investigating other data sources, such as radar, or by combining the physics-based optimal control simulations with machine learning. By contributing to predicting human responses and reconstructing human movement, my work lays a foundation for a better understanding of change-of-direction running movements in and outside the lab.