Learning Reactive Dexterous Grasping via Hierarchical Task-Space RL Planning and Joint-Space QP Control
Ho Jae Lee, Yonghyeon Lee, Alexander Alexiev, Tzu-Yuan Lin, Se Hwan Jeon, Sangbae Kim · 2026
Read Learning Reactive Dexterous Grasping via Hierarchical Task-Space RL Planning and Joint-Space QP Control by Ho Jae Lee, Yonghyeon Lee, Alexander Alexiev, Tzu-Yuan Lin, Se Hwan Jeon, Sangbae Kim (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-03363.txt.
Description
Ho Jae Lee, Yonghyeon Lee, Alexander Alexiev, Tzu-Yuan Lin, Se Hwan Jeon, and Sangbae Kim build a layered controller for reactive dexterous grasping that splits spatial intent from joint execution. Separate arm and hand learning agents act as a high-level planner, issuing task-space velocity commands.
A fast quadratic programming controller then turns those commands into joint velocities that respect limits and avoid collisions, keeping hardware safety strict. Operators can adjust safety margins live with no retraining. On a seven-axis arm with a twenty-joint hand, the system transferred from simulation to real grasping of unseen objects and recovered from disturbances.
Catalog record
| CREATOR | Ho Jae Lee, Yonghyeon Lee, Alexander Alexiev, Tzu-Yuan Lin, Se Hwan Jeon, Sangbae Kim |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 86 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.03363