NVIDIA & Berkeley open source tactile robotics method T-Rex
DrJimFan · x · 2026-08-22
NVIDIA and UC Berkeley open-sourced T-Rex, a tactile learning methodology for robotics addressing the lack of touch sensing. Key contributions include: (1) Architecture: A mixture-of-transformer running async clocks—a slow visuomotor expert plans motion while a fast tactile expert refines it at 4 touch ticks per vision tick. (2) Dataset: The largest tactile dataset released (50 hours, 5.5k episodes) collected on SOTA 22-DOF hardware. (3) Training: Extends EgoScale, using egocentric video for pretraining and tactile robot play for mid-training, bridging the gap to contact-rich manipulation. Tactile sensing is emphasized as critical for the "last mile" of manipulation.
Related event: NVIDIA and UC Berkeley Open-Source T-Rex Tactile Robot Learning(2 posts)→
More from Embodied
- Tevel's robotic system harvests fruit around the clock using computer vision — LexiLove · 2026-10-16
- SHOES: 10 minutes of real-robot RL lets quadrupeds walk in shoes, CoRL 2026 paper — kscottz · 2026-10-16
- Tesla Cybercab robotaxis logged zero crashes in first two weeks of public rides, NHTSA data shows — elonmusk · 2026-10-16
- Robotics LLMs tested on 16 tasks: GPT-6 Astra hits 17.2% success vs Fable's 9.1% — pabbeel · 2026-10-16
- WEDGE paper generates training environments conditioned on navigation agent failures — _akhaliq · 2026-10-16
- Viral meme nails AI's irony: it can write symphonies but can't do the dishes — briannekimmel · 2026-10-16