Kosmo and Ropedia release Physical AI dataset: 1,589 3D-scanned environments, 131 hours of egocentric video
liuziwei7 · x · 2026-11-15
Kosmo, a 3D world data engine, has partnered with Singapore-based startup Ropedia to build scalable real-world data infrastructure for Physical AI — an integrated pipeline covering spatial data capture, structuring, standardized delivery, and model validation, targeting leading AI labs including OpenAI, Meta, DeepMind, and NVIDIA.
The first concrete output is Ropedia Xperience 2, released November 12, featuring:
- 1,589 metric-scale 3D-scanned environments
- 74 tasks
- 4,230 egocentric recordings totaling 131 hours
All environment scans were captured with Kosmo One, natively aligning static spatial assets and dynamic human experiences in a shared coordinate system and timeline. The project demonstrates four data-reuse pathways — real-to-sim conversion, motion retargeting, virtual teleoperation, and long-horizon trajectory generation — to help train foundation, embodied mobile manipulation, and world models while bridging the Sim-to-Real gap.
More from Embodied
- Berkeley Legend Patterson Predicts Superintelligent Chef Robots Within Two Years — davidpattersonx · 2026-11-15
- Robot dog spotted walking a real dog on a leash in San Francisco — Polymarket · 2026-11-15
- NVIDIA showcases startups using vision AI and robots to recover recyclables from waste — nvidia · 2026-11-15
- Google robotics team denies disbanding: refocused around Astra, same roadmap and goals — carlosdponx · 2026-11-15
- NVIDIA GR00T 2 demo: two Unitree G1 robots team up to fill and deliver popcorn — CyberRobooo · 2026-11-15
- Polymarket gives Tesla California robotaxi launch by Dec 31 just 6% odds — Polymarket · 2026-11-15