Verified80% confidenceFactExact time
In 2019, OpenAI used domain randomization techniques to train a robotic hand to solve a Rubik's Cube, randomizing more than 100 physical parameters (including friction coefficients, object mass, and joint damping), and ultimately succeeded in transferring the policy to a physical robot.
6
Sources
80%
Confidence
Long-term
Relevance
8/8/2026
First Seen
Sources
Related Claims
UnverifiedLandmark works such as OpenAI's training of a robotic hand to solve a Rubik's Cube and ETH Zurich's training of quadruped robots both use domain randomization as a core technique.75% similarVerifiedWhen training a dexterous hand to solve the Rubik's Cube, OpenAI used Automatic Domain Randomization, which maintains training efficiency by dynamically adjusting environment difficulty.75% similarUnverifiedA 2023 study by a Berkeley team demonstrated that bipedal robot policies trained through large-scale domain randomization can achieve zero-shot transfer to real hardware.73% similarUnverifiedOpenAI's early robotic hand project Dactyl adopted a paradigm combining imitation learning and reinforcement learning, training in a simulated environment before transferring to a real robotic hand to achieve Rubik's Cube solving.68% similarVerifiedDomain randomization technology randomly perturbs physical parameters (such as mass, friction coefficient, motor delay, etc.) during training, forcing the policy to learn to operate robustly under various conditions.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/711055API
curl https://kongchang.com/api/v1/knowledge/claims/711055MCP
get_claim(id=711055)