Papers
13
Total Citations
204
H-Index
8
About
Felix Widmaier is a leading researcher in dexterous robotic manipulation, with a focus on bridging the gap between simulation and real-world application. His primary contributions center on developing and benchmarking reinforcement learning (RL) systems for in-hand object manipulation, particularly using the TriFinger robot—an open-source platform he helped create to democratize dexterity research. Widmaier’s most cited work (45 citations) demonstrates a breakthrough in transferring policies trained entirely in GPU simulation to a remote real-world TriFinger, enabling 6-DoF object pose control. He also pioneered a pixel-wise regression method for robot arm pose estimation (36 citations), improving hand-eye coordination under noisy conditions. His benchmarks for structured policies and policy optimization (24 citations) provide standardized evaluation for real-world dexterous tasks, while his winning entry in Phase 1 of the Real Robot Challenge 2021 showcases practical success using deep RL and knowledge transfer for sparse-reward tasks. With over 180 total citations, Widmaier’s work is instrumental in making dexterous manipulation more accessible, reproducible, and transferable from simulation to reality.
Research Focus
Key Achievements
Top Papers
- 1
- 2Robot arm pose estimation by pixel-wise regression of joint angles36 citations · 2016
- 3
- 4TriFinger: An Open-Source Robot for Learning Dexterity24 citations · 2020
- 5
- 6TriFinger: An Open-Source Robot for Learning Dexterity16 citations · 2020
- 7
- 8Benchmarking Offline Reinforcement Learning on Real-Robot Hardware11 citations · 2023
- 9
- 10