arXiv 2610.196002026-10-15cs.RO

破解移动操作难题：从故障诊断到智能体机器人系统

Cracking Mobile Manipulation: From Failure Diagnosis to Agentic Robot System

School of Intelligence Science and Technology, Peking University（北京大学智能科学与技术学院）；State Key Laboratory of General Artificial Intelligence, Peking University（北京大学通用人工智能国家重点实验室）；Alibaba DAMO Academy, Alibaba Group（阿里巴巴集团达摩院）；Institute of Automation, Chinese Academy of Sciences（中国科学院自动化研究所）；Tsinghua University（清华大学）；University of Technology Sydney（悉尼科技大学）；University of Liverpool（利物浦大学）

Zixuan Wang, Jia-Ning Li, Yinzhou Tang, Yanzhao Yu, Biao Wu, Meng Fang, Xin Li, Yiwu Zhong

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

该研究针对VLA模型在长程移动操作中性能骤降的问题，通过故障诊断确定三类瓶颈，开发故障引导的智能体干预方法，在RoboCasa365等数据集及真实机器人任务中显著提升了移动操作成功率。

摘要：

近期，视觉-语言-动作（VLA）模型在局部机器人操作任务中展现出较强性能，但其成功难以直接迁移至长程移动操作任务。在实验中，预训练的子任务VLA在原子级操作任务上达到80%的成功率，但当这些技能组合为移动操作时，成功率仅为19%。研究人员认为，该差距源于执行过程中的故障，而非仅局部操作能力不足。为揭示这些瓶颈，研究人员系统诊断了长程移动操作中的故障，确定了三类主导场景：任务规划、导航与基座定位、局部操作。基于该诊断，研究人员开发了故障引导的智能体干预方法，通过基于演示的规划引导、自适应视觉记忆、可访问性感知的基座调整以及局部末端执行器调整，增强了固定的子任务VLA。在RoboCasa365数据集上，这些干预措施逐步将任务成功率从19%提升至47%，最终系统在保留的场景种子上达到49%，而基线方法仅为18%；在BEHAVIOR-1K数据集上，相同设计将性能从14%提升至37%；在真实机器人任务中，从5/20提升至13/20。这些结果表明，可靠的移动操作可通过诊断并干预当前瓶颈来推进，而非单纯扩展底层策略，代码将在指定网址发布。

英文摘要：

Recent vision-language-action (VLA) models have shown strong performance on local robot manipulation, yet their success does not readily translate to long-horizon mobile manipulation. In our experiments, a pretrained subtask VLA achieves 80% success rate on atomic manipulation tasks but only 19% when these skills are composed into mobile manipulation. We argue that this gap stems from failures across the execution process rather than from local manipulation capability alone. To uncover these bottlenecks, we systematically diagnose failures in long-horizon mobile manipulation and identify three dominant regimes: task planning, navigation and base positioning, and local manipulation. Based on this diagnosis, we develop failure-guided agentic intervention that augments a fixed subtask VLA with demonstration-based planning guidance, adaptive visual memory, accessibility-aware base adjustment, and local end-effector adjustment. On RoboCasa365, these interventions progressively improve task success from 19% to 47%, with the finalized system reaching 49% on held-out episode seeds, compared with 18% for the baseline. The same design improves performance from 14% to 37% on BEHAVIOR-1K and from 5/20 to 13/20 on real-robot tasks. These results suggest that reliable mobile manipulation can be advanced by diagnosing and intervening on the current bottleneck, rather than uniformly scaling the underlying policy. Our code will be released at https://github.com/PKU-VaLuE-Lab/Cracking-Mobile-Manipulation.git.

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。