I am an ELLIS PhD student supervised by Dieter Büchler (JKU Linz, CIFAR, Amii), Ingmar Posner (University of Oxford), and Bernhard Schölkopf (ELLIS, MPI-IS). My research interests include reinforcement learning, sim-to-real transfer, learned dynamics models, tendon-driven robots, and artificial muscles.
My research focuses on learning-based control for musculoskeletal robots. Driven by tendons and artificial muscles, these robots are fast, lightweight, and inherently compliant, making them well suited for dynamic tasks and safe interaction with humans. However, nonlinear actuator dynamics, hysteresis, and tendon friction make them notoriously hard to model and control. I address these challenges by learning accurate actuator models from real robot data and combining them with physics simulators to train policies entirely in simulation. This approach enabled the first zero-shot sim-to-real transfer for multi-joint musculoskeletal robot arms, solving tasks that require dynamic and precise motions, such as ball-in-a-cup and table tennis.
For more information on my research projects, check out my Google Scholar.