Learning Dexterous Manipulation Skills from Imperfect Simulations
Elvis Hsieh, Wen-Han Hsieh, Yen-Jen Wang, Toru Lin, Jitendra Malik, Koushil Sreenath, Haozhi Qi, UC Berkeley · 2025
Read Learning Dexterous Manipulation Skills from Imperfect Simulations by Elvis Hsieh, Wen-Han Hsieh, Yen-Jen Wang, Toru Lin, Jitendra Malik, Koushil Sreenath, Haozhi Qi, UC Berkeley (2025) 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-2512-02011.txt.
Description
Robots with multi-fingered hands struggle to learn delicate tasks in simulation, because contact forces and touch signals are so hard to imitate faithfully. Elvis Hsieh and co-authors present DexScrew in 2025, a framework that carries skills learned in imperfect simulations onto real hardware.
They demonstrate the transfer on two demanding jobs: fastening a nut onto a bolt and driving a screw with a multi-fingered hand. Both tasks depend on rich contact and tactile feedback, exactly what simulators model worst. Showing that learning can survive those flaws widens the range of manipulation skills robots can practice safely in virtual worlds first.
Catalog record
| CREATOR | Elvis Hsieh, Wen-Han Hsieh, Yen-Jen Wang, Toru Lin, Jitendra Malik, Koushil Sreenath, Haozhi Qi, UC Berkeley |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 42 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-02011