Development of an Optimal Control Framework to Predict Human Motion

Motion prediction is a field of great interest in sports and rehabilitation because it can help to determine the best way to perform a sport movement, to choose the optimal treatment for a specific patient, or to personalize an assistive device. The main objective of this bachelor’s thesis is to develop an optimal control framework to predict human motion. The iomechanical simulation software OpenSim has been used to develop the biomechanical model and perform inverse dynamics analysis, and the direct collocation optimal control software Moco has been used to formulate and solve the optimal control simulations. Once the optimal control framework is developed, it has been used to predict human motions, in this thesis, the squat-to-stand and the walking 2D motion. Their cost functions have been modified to predict the most accurate motion.This project includes a theoretical background, a description of what is human motion, the two target motions, the analysis types and the optimal control techniques for human motion prediction. We explain the models used, the reference data and the framework of OpenSim Moco. For every motion, the biomechanical model and the problem formulation are presented in detail, with their results and discussion. Finally, we exposed the environmental, social and economic impact of the project. Regarding the results of this thesis, the most accurate motion is reached by implementing the state tracking goal. It also has been appreciated that the kinetic energy goal improves convergence but it needs some kinematic constraints for a correct motion execution. Moreover, the control goal complemented with tracking does not bring any extra help but it could be a useful cost term for prediction without or little tracking. Regarding the reality of the motion predictions, the squat-to-stand one has been executed as a real one, whereas the gait has not been real since a contact model was missing and, besides, the model used is not a full-body on