arXiv 2505.234502025-06-11cs.RO

Agentic Robot：面向具身智能体中视觉-语言-动作模型的脑启发框架

Agentic Robot: A Brain-Inspired Framework for Vision-Language-Action Models in Embodied Agents

Jilin University（吉林大学）；Harvard University（哈佛大学）；Massachusetts Institute of Technology（麻省理工学院）；Huazhong University of Science and Technology（华中科技大学）；Southern University of Science and Technology（南方科技大学）；Lehigh University（莱斯大学）；Shanghai Jiao Tong University（上海交通大学）

Zhejian Yang, Yongchao Chen, Xueyang Zhou, Jiangyue Yan, Dingjie Song, Yinuo Liu, Yuting Li, Yu Zhang, Pan Zhou, Hechang Chen, Lichao Sun

备注：: 20 pages, 8 figures

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出脑启发框架 Agentic Robot，利用 SAP 协调推理、VLA 执行与时序验证形成闭环，在 LIBERO 长程操控任务上达到 79.6% 的最先进成功率。

摘要：

长程机器人操控对自主系统构成重大挑战，需要在复杂序列任务中具备扩展推理、精确执行和稳健错误恢复能力。现有方法无论基于静态规划还是端到端视觉运动策略，都存在误差累积问题，并且在执行过程中缺乏有效验证机制，限制了其在真实场景中的可靠性。我们提出 Agentic Robot，这是一个脑启发框架，通过 Standardized Action Procedure（SAP，标准化动作流程）来解决这些局限；SAP 是一种新颖的协调协议，用于管控整个操控任务中的组件交互。该方法从人类组织中的 Standardized Operating Procedures（SOP，标准作业程序）汲取灵感，为规划、执行和验证阶段建立结构化工作流。我们的架构包含三个专用组件：（1）大型推理模型，将高层指令分解为语义连贯的子目标；（2）视觉-语言-动作执行器，根据实时视觉输入生成连续控制命令；（3）时序验证器，通过内省式评估实现自主推进和错误恢复。这种由 SAP 驱动的闭环设计支持无需外部监督的动态自验证。在 LIBERO 基准上，Agentic Robot 取得最先进性能，平均成功率为 79.6%，在长程任务上分别比 SpatialVLA 和 OpenVLA 高出 6.1% 和 7.4%。这些结果表明，专用组件之间由 SAP 驱动的协调提升了序列操控的性能和可解释性，显示出构建可靠自主系统的巨大潜力。项目 GitHub：https://agentic-robot.github.io。

英文摘要：

Long-horizon robotic manipulation poses significant challenges for autonomous systems, requiring extended reasoning, precise execution, and robust error recovery across complex sequential tasks. Current approaches, whether based on static planning or end-to-end visuomotor policies, suffer from error accumulation and lack effective verification mechanisms during execution, limiting their reliability in real-world scenarios. We present Agentic Robot, a brain-inspired framework that addresses these limitations through Standardized Action Procedure (SAP)--a novel coordination protocol governing component interactions throughout manipulation tasks. Drawing inspiration from Standardized Operating Procedures (SOPs) in human organizations, SAP establishes structured workflows for planning, execution, and verification phases. Our architecture comprises three specialized components: (1) a large reasoning model that decomposes high-level instructions into semantically coherent subgoals, (2) a vision-language-action executor that generates continuous control commands from real-time visual inputs, and (3) a temporal verifier that enables autonomous progression and error recovery through introspective assessment. This SAP-driven closed-loop design supports dynamic self-verification without external supervision. On the LIBERO benchmark, Agentic Robot achieves state-of-the-art performance with an average success rate of 79.6%, outperforming SpatialVLA by 6.1% and OpenVLA by 7.4% on long-horizon tasks. These results demonstrate that SAP-driven coordination between specialized components enhances both performance and interpretability in sequential manipulation, suggesting significant potential for reliable autonomous systems. Project Github: https://agentic-robot.github.io.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。