Erlernen von Repräsentationen menschlichen Motorverhaltens für Robotik und Regelungstechnik
Abstract
Current autonomous systems cannot facilitate a safe human-robot collaboration in unstructured environments. Hence, this thesis develops learning-based human motor behavior models to overcome this. We investigate different learned representations, including Gaussian process dynamics models with online updates and error bounds, stability-certified inverse reinforcement learning, and diffeomorphic function learning with geometric priors. Simulations and experiments validate the proposed methods.
Fields of Study & Subjects
Reinforcement Learning in Robotics
Robot Manipulation and Learning
Motor Control and Adaptation
Publication details
- Journal / source
- mediaTUM (Technical University of Munich)
- Publication date
- 29 January 2026
- Language
- EN
- Document type
- Thesis