Papers
9
Total Citations
105
H-Index
5
About
Leon Sievers is a roboticist at the forefront of **dextrous, purely tactile in-hand manipulation**, a field that aims to give robotic hands the human-like ability to reorient objects using only touch, without relying on vision. His core contributions lie in developing **reinforcement learning (RL) frameworks** that enable multi-fingered, torque-controlled hands to perform continuous regrasping and object rotation while maintaining permanent force closure—even with the hand oriented upside down. His most cited work (2022, 35 citations) demonstrated that such complex tasks can be learned from scratch and executed robustly using only tactile feedback, a significant step beyond vision-dependent approaches. Sievers has further advanced the field by tackling the critical challenge of **estimator-coupled RL**, creating novel differentiable particle filters that allow a hand to infer an object’s pose solely from proprioceptive signals (2022, 10 citations). His recent papers extend these capabilities to **shape-conditioned manipulation of diverse objects** (2024) and **composing grasping with in-hand manipulation** (2025), pushing toward real-world applicability. With a growing body of work that systematically addresses robustness, calibration, and state estimation, Sievers is shaping a future where robots can handle objects with the dexterity and sensory reliance of the human hand.
Research Focus
Key Achievements
Top Papers
- 1Learning Purely Tactile In-Hand Manipulation with a Torque-Controlled Hand35 citations · 2022
- 2
- 3
- 4Learning a State Estimator for Tactile In-Hand Manipulation10 citations · 2022
- 5
- 6
- 7
- 8
- 9