Sam Holt @ NeurIPS 2025 (@samianholt)

2025-01-17 | โค๏ธ 76 | ๐Ÿ” 11


Thanks for the shout-out, @m_wulfmeier! Thrilled to open-source the non-prehensile block reorientation setup from my @GoogleDeepMind internship, now part of MuJoCo Playground. By leveraging domain randomization and curriculum learning, we achieve smooth, compliant high-frequency torque control (200Hz)โ€”trained in 10 minutes for robust zero-shot transfer to real hardware!

Weโ€™re also excited to share MuJoCo Playground itselfโ€”a broader open-source sim-to-real pipeline harnessing GPU-accelerated reinforcement learning and environment simulation with JAX, tackling complex real-world tasks with training in as little as 10 minutes. It was a delight to contribute during my internship at @GoogleDeepMind, collaborating with incredible folks from @berkeley, @utoronto, @stanford, @ludaprojects, and the whole @GoogleDeepMind team.

Check out @kevin_zakkaโ€™s excellent thread for more details, and dive in here: https://playground.mujoco.org/

๐Ÿ”— ์›๋ณธ ๋งํฌ

๋ฏธ๋””์–ด

video


Auto-generated - needs manual review

์ธ์šฉ ํŠธ์œ—

Markus Wulfmeier (@m_wulfmeier)

Super exciting to have @samianholt s @GoogleDeepMind internship setup accessible to everyone!

์›๋ณธ ํŠธ์œ—

Tags

domain-simulation domain-visionos