Skip to content

Event

REA: Routed Embodied Agent Runtime

1 reports · 1 sources

Pengfei Zhou et al. propose REA, a routed embodied agent runtime that lets a reasoning-focused System 2, an embodied routing System 1, and a reactive System 0 policy collaborate within the same runtime; the core rule is that running policies do not wait for the brain. Across 3000 test episodes on 60 LIBERO-Pro tasks, REA achieved a 95.9% success rate, higher than the best single-agent baseline in its policy pool at 94.7%, while reducing control steps by 25% and control time by 17%. In a separate 100-episode RoboTwin 2.0 study, REA with a VLA policy achieved an 84% success rate, and with VLA+code 86%, versus 80% for the strongest single-VLA baseline, MEDex. The authors state that REA also transfers to BEHAVIOR-1K and to a real Unitree G1 humanoid robot, achieving the highest observed success rate among compared methods.

Report timeline

  1. REA: A Routed Embodied Agent Runtime for Asynchronous Brain-Cerebellum Control

    The authors propose REA, a routed embodied agent runtime that lets a reasoning-focused System 2, an embodied router System 1, and reactive System 0 policies cooperate within a single runtime, with the core rule that a running policy never waits on the brain.

    arXiv cs.RO ·

Related topics

Unitree RoboticsPlanning and decision makingRobot models and learningSoftware and compute systemsResearch papers

Interest trend

Current
3
Peak
4Nov 4 01:00
Past 24 hours
–

The trend compares participants observed throughout the same period, so its range may be smaller than the current interest count. Move or tap to inspect each hour; use the arrow keys on a keyboard.