How Reflex trains humanoids: 3-stage RL pipeline from control to RGB-D grounding
DJiafei · x · 2026-10-09
Reflex trains in three stages: (1) reinforcement learning for whole-body catching, (2) inferring box dynamics from delayed, incomplete observations, and (3) learning to recover that dynamics representation from RGB-D history while keeping the controller fixed. This decomposition gives each capability a direct training signal and makes visual learning substantially more scalable.
Related event: Reflex Lets Unitree G1 Catch Tossed Boxes in About a Second(8 posts)→
More from Embodied
- Reflex: Humanoid Robot Catches Thrown Boxes in Under a Second Using Onboard Vision — DJiafei · 2026-10-09
- Face P1: A Screen-Faced Robot Prototype Explores How Humans Relate to Agents — genmon · 2026-10-09
- GeneralistAI ships GEN-1.5: the packing task that stumped them for two years is now routine — E0M · 2026-10-09
- AI Smart Cap With Neural Sensing, Camera and Voice AI Previewed for CES 2027 — ThePeterMick · 2026-10-09
- Robot Girl Gang: A New Blog of Sharp Takes from Top Female Roboticists — mattbeane · 2026-10-09
- Watch: Robots folding hotel laundry end to end — LexiLove · 2026-10-09