FACE: contact-factorized imitation learning lets one policy handle unseen friction and geometry
KyleMorgenstein · x · 2026-11-17
Researchers from Kyung Hee University, Ewha Womans University, and UCF introduce FACE, a contact-factorized imitation learning framework that separates intended task behavior from environment-dependent contact factors. A learned contact-normal estimator and an online friction estimator let a single fixed policy adapt to unseen surface geometry and friction (wrinkled surfaces, PVC tape, etc.) on real robots, without updating policy weights. Paper on arXiv; code coming soon.
More from Embodied
- Connecting AI Agents to the Physical World via a Raspberry Pi MCP Server — jfrankie · 2026-11-17
- OneSteer: one-step diffusion controller lets Unitree G1 duck under 1.1m bars — Yifei Huang · 2026-11-17
- Samsung-Google AI glasses sell out within hours of pre-order across all 16 models — Scobleizer · 2026-11-17
- comma_ai's World Model Simulator Had a Hidden Bug—Fixed Without Retraining — minchoi · 2026-11-17
- A $250 robot hands you tools, fly-around World Series home runs: 10 wild AI visual demos this week — minchoi · 2026-11-17
- NYU team open-sources MABEL, a mobile bimanual robot built for ~$8K in parts — chris_j_paxton · 2026-11-17