Abstract

We use reinforcement learning (RL) to learn dexterous in-hand manipulation policies that can perform vision-based object reorientation on a physical Shadow Dexterous Hand. The training is performed in a simulated environment in which we randomize many of the physical properties of the system such as friction coefficients and an object’s appearance. Our policies transfer to the physical robot despite being trained entirely in simulation. Our method does not rely on any human demonstrations, but many behaviors found in human manipulation emerge naturally, including finger gaiting, multi-finger coordination, and the controlled use of gravity. Our results were obtained using the same distributed RL system that was used to train OpenAI Five. We also include a video of our results: https://youtu.be/jwSbzNHGflM .

Keywords

Reinforcement learningObject (grammar)Computer scienceArtificial intelligenceShadow (psychology)Computer visionRobotRobotic handHuman–computer interactionPsychology

Affiliated Institutions

Related Publications

Publication Info

Year
2019
Type
article
Volume
39
Issue
1
Pages
3-20
Citations
1502
Access
Closed

External Links

Download PDF (Free) View on DOI.org arXiv Semantic Scholar

Social Impact

Altmetric
PlumX Metrics

Social media, news, blog, policy document mentions

Citation Metrics

1502
OpenAlex
98
Influential
953
CrossRef

Cite This

OpenAI Marcin Andrychowicz, Bowen Baker, Maciek Chociej et al. (2019). Learning dexterous in-hand manipulation. The International Journal of Robotics Research , 39 (1) , 3-20. https://doi.org/10.1177/0278364919887447

Identifiers

DOI
10.1177/0278364919887447
arXiv
1808.00177

Data Quality

Data completeness: 84%