Q-Planning method lets robots self-improve, boosting success rates from 25% to 80%
animesh_garg · x · 2026-08-31
Traditional imitation learning struggles with precise tasks. The new Q-Planning method adds a small "critic" module to score potential next moves.
Core Mechanism:
- Keeps the original Behavior Cloning (BC) policy frozen.
- Updates only the Q-function, learning from both success and failure.
- Uses Q-weighted averaging for action selection during inference.
Results:
- In real-world bimanual robot tests, wallet insertion success rate rose from 25% to 80%.
- Cup-stacking tasks also saw significant gains.
- Performance improves in 30 minutes of autonomous attempts without extra human demos.
Paper: Beyond Imitation: Self-Improving Robot Policies via Off-Policy Q-Planning
More from Embodied
- Berkeley researchers sample humanoid robot policies directly from weight space with diffusion — berkeley_ai · 2026-10-16
- OneWM: unified causal training shows world and action modeling improve each other — IAAI233 · 2026-10-16
- Tesla's hover car isn't a flying car — compressed air can't sustain flight — tinyfool · 2026-10-16
- Scobleizer relays claim that Optimus demo will be even more mind-blowing than the road-flipping Roadster — Scobleizer · 2026-10-16
- Indie dev's AI PASSPORT wearable firmware hits 1,200+ downloads in days — ezshine · 2026-10-16
- 1X Opens California Factory for Mass Production of NEO Humanoid Robots — emmanuelvivier · 2026-10-16