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

📄 arXiv: 2610.42053v1 📥 PDF

作者: Pengfei Zhou, Yixing Ma, Zhengfeiyang Zhang, Haoyuan Chen, Zhenglin Wan, Hexin Wang, Wangbo Zhao, Yang You

分类: cs.RO, cs.CV

发布日期: 2026-10-30

备注: 32 pages, 14 figures


💡 一句话要点

提出REA以解决异步脑小脑控制的协调问题

🎯 匹配领域: 支柱一:机器人控制 (Robot Control) 支柱九:具身大模型 (Embodied Foundation Models)

关键词: 嵌入式代理 异步控制 路由记忆 策略协同 机器人控制 智能系统 高效决策

📋 核心要点

  1. 现有方法在高层推理与快速反应之间存在协调困难,导致执行效率低下。
  2. REA通过路由记忆和判断门的设计,实现了不同控制策略的高效协作,避免了相互等待的问题。
  3. 在多个实验中,REA的成功率达到了95.9%,并在控制步骤和时间上分别减少了25%和17%。

📝 摘要(中文)

基础模型在高层推理方面表现出色,但作为细粒度控制器仍然过于缓慢和昂贵。快速策略如VLA和全身控制器虽然反应高效,但在泛化能力上存在不足。为了解决这两个问题,本文提出了REA(路由嵌入式代理),一个统一的运行时环境,使得深思熟虑的系统与快速反应的系统能够在不相互等待的情况下协同工作。REA通过路由记忆、边界触发的判断门和共享预算升级阶梯来优化策略执行。实验结果显示,REA在多个任务中取得了显著的成功率和效率提升。

🔬 方法详解

问题定义:本文旨在解决现有嵌入式代理系统中,深思熟虑的控制策略与快速反应策略之间的协调问题。现有方法在执行过程中常常出现相互等待和效率低下的情况。

核心思路:REA的核心思路是通过设计一个统一的运行时环境,使得不同层次的控制策略能够在不相互等待的情况下协同工作。具体而言,REA利用路由记忆来提前存储策略的成功率,从而优化决策过程。

技术框架:REA的整体架构包括三个主要模块:深思熟虑的系统(System 2)、嵌入式路由器(System 1)和快速反应策略(System 0)。在执行过程中,边界触发的判断门会根据路由记忆来决定是否需要咨询深思熟虑的系统。

关键创新:REA的主要创新在于其异步协调机制,避免了传统方法中策略间的等待问题。通过路由记忆和共享预算的设计,REA能够实现更高效的策略执行。

关键设计:REA的设计中,路由记忆用于存储每个策略的成功率,判断门根据策略执行情况进行决策,且共享预算的设计确保了在失败时的有界恢复。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

在3000个测试案例中,REA的成功率达到了95.9%,超越了最佳单一代理基线(94.7%),同时控制步骤减少了25%,控制时间减少了17%。在RoboTwin 2.0的研究中,REA与VLA策略的成功率为84%,与VLA+code的成功率为86%,均高于最强单一VLA基线MEDex的80%。

🎯 应用场景

REA的研究成果在机器人控制、自动驾驶和智能制造等领域具有广泛的应用潜力。通过提高不同控制策略之间的协同效率,REA能够在复杂环境中实现更高效的决策和执行,推动智能系统的进一步发展。

📄 摘要(原文)

Foundation models are becoming remarkably reliable at high-level reasoning, yet they remain too slow and costly to serve as fine-grained controllers. Conversely, fast policies like VLAs and whole-body controllers offer more efficient reflexes, but struggle to generalize. While pairing a deliberative brain (foundation models) with a reactive cerebellum (fast policies) seems natural for embodied systems, two obstacles remain: methods are built and scored on incompatible stacks, and their asynchronous coordination is difficult. For example, an agent that calls the VLA in the agentic loop can stall the fast execution, while a VLA that calls the agent on errors tends to re-trigger fallbacks repeatedly or completely miss silent failures. To address this, we propose REA (Routed Embodied Agent), a unified runtime where a deliberative System 2, an embodied router System 1 and reactive System 0 policies cooperate under a single golden rule - no running policy waits on the brain. Instead of querying the brain at each failure, REA employs a route memory that stores each policy's measured success ahead of time, a boundary-triggered judge gate that consults the brain on policy execution, and a shared-budget escalation ladder for bounded recovery. Evaluated on 3,000 test episodes across 60 LIBERO-Pro tasks, REA reaches 95.9% success and beats the best single-agent baseline in its own policy pool (94.7%) while reducing control steps by 25% and control time by 17%. In a separate 100-episode RoboTwin 2.0 study, REA achieves 84% success with VLA policies and 86% with VLA+code, compared with 80% for the strongest single VLA baseline MEDex. These results demonstrate learned and programmatic policies operating within the same runtime. REA also transfers to BEHAVIOR-1K and a real Unitree G1 humanoid, achieving the highest observed success among the compared methods.