Fei-Fei Li's WorldLabs Unveils R2S2R for Robot Sim-to-Real Training
智东西 · wechat · 2026-07-29
Fei-Fei Li's startup WorldLabs released new results for its robot training engine R2S2R (Real-to-Sim-to-Real). The technology aims to unlock the Scaling Law in robotics by building interactive simulated worlds highly aligned with reality, solving the high costs and data bottlenecks of real-world testing.
R2S2R operates in two phases: Real-to-Sim converts physical robot tasks into simulated models preserving physical dynamics, while Sim-to-Real trains and evaluates policies in the virtual environment before deploying them to real robots. WorldLabs demonstrated robotic arms successfully executing complex tasks like cable manipulation purely on simulated data before real-world transfer.
Furthermore, the system serves as a high-throughput evaluation layer to proactively identify model failure modes before hardware testing. WorldLabs believes high-fidelity simulation will become critical infrastructure for scalable robot training.
More from Embodied
- Origin Autonomy emerges with $20M to automate interior construction, robots running 3x human speed — lukas_m_ziegler · 2026-10-17
- One-piece 4-meter PETG bar counter shows how far large-format 3D printing has come — lukas_m_ziegler · 2026-10-17
- PayBox open-source SDK gives robots a wallet — a hexapod paid real USDC on-chain — R4nsack · 2026-10-17
- JR West to deploy giant humanoid robots this fall, cutting high-altitude rail crews by 30% — CyberRobooo · 2026-10-17
- Hobbyist models humanoid robot, hits battery space and servo torque limits — _Stocko_ · 2026-10-17
- Android robots learn to switch emotions in real time, crossing the uncanny valley — TinfoilTricorn · 2026-10-17