معرفی
Prof. Dr. Berthold Bäuml serves as a full Professor at the Technical University of Munich (TUM) within the TUM School of Computation, Information and Technology, holding the Chair for Learning AI for Dextrous Robots. His research bridges artificial intelligence and robotics to develop advanced manipulation capabilities in humanoid systems, with particular emphasis on tactile sensing and learning-based control architectures.
His core research interests span Robotics, Artificial Intelligence, Machine Learning, Dextrous Manipulation, Tactile Sensing, and Reinforcement Learning. Bäuml investigates how robots can autonomously learn complex manipulation skills through tactile feedback, focusing on in-hand manipulation, grasping of unknown objects, and sim-to-real transfer for multi-fingered hands. His work integrates state estimation with reinforcement learning to achieve robust performance in unstructured environments.
Analysis of his 2022-2025 publications reveals a concentrated research trajectory in tactile dexterous manipulation, characterized by modular reinforcement learning architectures, shape-conditioned agents, and torque prediction systems. Key trends include the development of time-optimal controllers, self-contained calibration methods, and collision-aware motion planning, all leveraging deep learning for real-time performance in humanoid platforms like Agile Justin.
