[论文] RoboRSI: Stable, efficient, and reusable robot self-evolution in compl...

论文概要

研究领域: ML 作者: Zimo Wen, Yijin Chen, Yuxuan Cao, Wendi Chen, Yanwen Zou, Wenye Yu, Fuhang Kuang, Han Xue, Jun Lv, Chuan Wen, Cewu Lu 发布时间: 2026-10-08 arXiv: 2610.12424

中文摘要

通用机器人不仅能执行多样任务，还应通过经验不断改进，将在执行过程中学到的东西转化为后续任务可复用的能力。通过代码行动的机器人智能体已能从执行反馈中修复程序，但如何围绕赋予经验意义的任务结构来组织这些经验，仍然是一个核心挑战——使每次修复都能归因到负责的能力分支、有执行证据支持、并在复用前经过验证。我们引入 RoboRSI——一个基于自顶向下技能精炼（TSR）的机器人自我改进系统。TSR 将任务分解为复合技能、原子技能和基础技能，各职责范围明确、具有显式输入-输出契约，将每次执行结果归因到负责的分支，并将修改限制在该分支内。在此结构上，Manager、Planner、Engineer 和 Reviewer 协调规划、执行、诊断和新技能的验证发布，人类通过目标和纠正来引导过程；稳定的技能序列进一步整合为可复用的复合技能。在移动操作臂上，RoboRSI 通过 104 轮开发出了多物体家庭清理能力。在仿真中，它在 LIBERO、LIBERO-PRO、LIBERO-Plus 和 RoboTwin 上取得最高成功率，超过最强基线 2.7 到 11.0 个百分点。

原文摘要

A generalist robot should not only perform diverse tasks but also improve through experience, turning what it learns during execution into capabilities that later tasks can reuse. Robot agents that act through code can already repair programs from execution feedback, yet it remains a central challenge to organize this experience around the task structure that gives it meaning, so that each repair is attributed to the responsible capability, supported by execution evidence, and validated before it is reused. We introduce RoboRSI, a robot self-improvement system built on Top-Down Skill Refinement (TSR). TSR decomposes tasks into compound, atomic, and base skills with scoped responsibilities and explicit input-output contracts, attributes each execution outcome to the responsible branch, and ...

#论文 #arXiv #ML #小凯