Ropedia Xperience 2: 1,589 real-scene scans and 4,230 egocentric recordings to scale physical AI
liuziwei7 · x · 2026-11-20
The Ropedia Xperience 2 paper (arXiv) proposes a gold standard for scaling physical AI data and environments:
- Scale: 1,589 metric-scale 3D scene scans (homes, offices, restaurants, stores, outdoors across 596 locations) plus 4,230 egocentric recordings covering 74 everyday tasks and 131 hours.
- Reuse pipeline: scenes are converted into physically interactive simulation environments (3DGS etc.), human motion is retargeted to robots, enabling real-to-sim, virtual teleoperation, and long-horizon trajectory generation.
- Motivation: foundation models like GPT-6 Astra and Claude Fable 5.1 can control robots from visual observations but remain brittle on long-horizon, contact-rich tasks; physical AI needs its own scaling laws, which first requires knowing which data to scale and how to produce and reuse it continuously.
- Benchmarks: bundled benchmarks for depth/geometry, 3D object detection, and spatial intelligence (20 scenes, 3 tracks, 30 baselines).
Related event: Ropedia Xperience 2 Ships Large-Scale Physical AI Dataset(3 posts)→
More from Embodied
- Maker Builds a Flying Probe Machine from Two Broken Ender3 Printers — kscottz · 2026-11-20
- Robots will proliferate far faster than new homes get built, argues roboticist — KyleMorgenstein · 2026-11-20
- Codex-branded Raspberry Pi drives open-source AgOpenGPS tractor autosteer in the field — 4310sy · 2026-11-20
- Robotics engineer dismantles the 'humanoids fit human spaces' debate: it's about safety — KyleMorgenstein · 2026-11-20
- NVIDIA Cosmos world models hit 20 million downloads on Hugging Face — heghbalz · 2026-11-20
- Framework Desktop 192GB config ships this week; current pricing may not last — film_girl · 2026-11-20