Rolling-WAM spreads video-action denoising across cycles for 4.5x faster robot replanning
USC-PSI-Lab · hf · 2026-09-29
World Action Models (WAMs) couple action generation with future visual prediction for robotic manipulation, but completing the joint video-action denoising at each replanning cycle causes latency that limits closed-loop responsiveness.
USC-PSI-Lab's Rolling-WAM maintains a sliding window of video-action chunks at staggered noise levels. Each step, a rolling noise schedule fully denoises the imminent action chunk for execution while partially refining farther-future chunks; as new camera observations arrive, retained future chunks continue denoising. This distributes compute over time while carrying an evolving visual-action context across chunk boundaries.
Evaluations on LIBERO, RoboTwin, and a real Unitree G1 humanoid show competitive manipulation performance, with a 4.5x steady-state replanning speedup over standard joint WAMs by avoiding full-horizon denoising from scratch.
More from Embodied
- Standard Bots Launches No-Code Robot OS, Says Only 6% of US Manufacturers Use Robots — LexiLove · 2026-09-29
- Agility Robotics' Digit bears marks from real work outside the lab — jonstephens85 · 2026-09-29
- Wuji Dexterous Hand at IROS Packs 20 Active DoF to Mimic Human Hand — jonstephens85 · 2026-09-29
- Matt Shumer says Opus 5.5 cleared a 3D printer bed after 16 tries, taking humans out of the loop — mattshumer_ · 2026-09-29
- Pinductor uses LLM priors to learn POMDP world models with 330x fewer episodes than DreamerV3 — burny_tech · 2026-09-29
- Driver swaps FSD subscription for open-source sunnypilot on comma four, 8 months in — SuB8u · 2026-09-29