Reflex: Humanoid Robot Catches Thrown Boxes in Under a Second Using Onboard Vision
DJiafei · x · 2026-10-09
A UW/UNC/NUS team introduced Reflex, a framework enabling a Unitree G1 humanoid to catch human-thrown boxes using only onboard RGB-D egocentric vision — perceiving, predicting and moving its whole body within 1 second. The skill was trained robustly via large-scale simulation, with more humanoid work promised.
More from Embodied
- Danu Robotics' six-year fight to build a better recycling robot — TechCrunch AI · 2026-10-10
- NVIDIA releases GR00T-based Agile One S SSD Pick robot deployment model on Hugging Face — _akhaliq · 2026-10-10
- Success-Guided Sampling: sim-to-real RL nails dexterous assembly with zero demos — kevin_zakka · 2026-10-10
- Stanford's Mulligan targets failure states for on-robot learning, beats uniform collection across 2,550 blind evals — ZhaoMandi · 2026-10-10
- Build Your Own Robot Dog With This ESP32-Powered Board — natesiggard · 2026-10-10
- Nvidia's ARC boosts robot reasoning success 5x with no new demos or retraining — arankomatsuzaki · 2026-10-10