EmbodiedSkills: A Unified Framework for Orchestrating, Training, and Deploying VLA Agents

📄 arXiv: 2609.01281v1 📥 PDF

作者: Wei Wang, Wenqiao Zhang, Yutong Lin, Yuqian Yuan, Tianwei Lin, Jinhao Mao, Zhenxuan Fan, Mingjian Gao, Yang Dai, Wentong Li, Zheqi Lv, Zheng Dong, Yingjie Niu, Jiaqi Zhu, Jun Xiao, Chao Li, Yueting Zhuang

分类: cs.RO, cs.AI

发布日期: 2026-09-01

备注: 20 pages, 4 figures, 5 tables

💡 一句话要点

提出EmbodiedSkills框架以解决长时间任务中的VLA代理协调问题

🎯 匹配领域: 支柱九：具身大模型 (Embodied Foundation Models)

关键词: 视觉-语言-动作 机器人协调 执行验证 任务适应 智能代理 长时间任务 多模态融合

📋 核心要点

  1. 现有的VLA模型在处理长时间任务时，缺乏有效的协调机制，无法确保动作的有效性和结果的验证。
  2. EmbodiedSkills框架通过将技能决策视为执行提案，提供了一个统一的接口，确保在执行前后进行必要的检查和验证。
  3. 在实验中，任务适应的低层VLA策略在多个任务中表现出色，成功率显著提升，证明了该框架的有效性。

📝 摘要（中文）

视觉-语言-动作（VLA）模型将视觉观察和语言指令直接映射到机器人动作，但长时间任务需要的不仅仅是动作预测。代理必须协调感知、规划、执行、进度验证和恢复。我们提出了EmbodiedSkills，一个统一框架，将每个技能决策视为执行提案：在执行前检查其前提条件，并在执行后验证结果。该框架通过共享的可执行技能接口连接高层技能选择、低层VLA执行和后续验证，支持低层策略的替换或适应而无需更改代理循环。我们在RoboTwin 2.0和LIBERO上实例化了EmbodiedSkills，任务适应的低层VLA策略在50个RoboTwin 2.0任务中平均成功率为86.20%，在四个LIBERO套件中为97.40%。

🔬 方法详解

问题定义：本论文旨在解决长时间任务中VLA代理在执行过程中的协调问题。现有方法往往只关注动作预测，缺乏对执行有效性和结果验证的全面考虑。

核心思路：EmbodiedSkills框架将每个技能决策视为执行提案，设计了一个固定的可执行技能接口，以便在执行前检查前提条件，并在执行后验证结果，从而增强了代理的执行能力。

技术框架：该框架包括高层技能选择、低层VLA执行和后续验证三个主要模块。通过共享的接口，这些模块在单一代理循环中协同工作，确保了执行过程的连贯性和有效性。

关键创新：EmbodiedSkills的最大创新在于其统一的可执行技能接口，使得低层VLA策略可以灵活替换或适应，而不需要更改整个代理的执行循环。这种设计显著提高了系统的可维护性和扩展性。

关键设计：在实现中，框架记录了规划、执行、验证和恢复事件，形成结构化轨迹，为各个组件提供监督，并支持在交互反馈可用时进行在线适应。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

在实验中，任务适应的低层VLA策略在50个RoboTwin 2.0任务中实现了86.20%的平均成功率，在四个LIBERO套件中达到了97.40%。这些结果表明，EmbodiedSkills框架显著提升了VLA代理在复杂任务中的执行性能。

🎯 应用场景

EmbodiedSkills框架在机器人领域具有广泛的应用潜力，尤其是在复杂的长时间任务中，如自动化制造、服务机器人和人机交互等场景。其设计理念可以为未来的智能代理系统提供更高效的执行和适应能力，推动机器人技术的进一步发展。

📄 摘要（原文）

Vision-language-action (VLA) models map visual observations and language instructions directly to robot actions, but long-horizon tasks require more than action prediction. An agent must coordinate perception, planning, execution, progress verification, and recovery as the physical state evolves. An action prediction or a model-generated skill decision does not, by itself, guarantee that the proposed operation is valid in the current state or that its outcome will be verified. We propose EmbodiedSkills, a unified framework that treats each skill decision as an execution proposal: the runtime checks its prerequisites before execution and verifies the outcome afterward. A shared executable-skill interface connects high-level skill selection, bounded low-level VLA execution, and post-action verification within a single agent loop. Because this interface remains fixed, low-level VLA policies can be replaced or adapted without changing the agent loop. The interface also records planning, execution, verification, and recovery events as structured trajectories, which provide supervision for individual components and can support optional online adaptation when interactive feedback is available. We instantiate EmbodiedSkills with Qwen3-VL and OpenPI/pi0.5 on RoboTwin 2.0 and LIBERO. Task-adapted low-level VLA policies achieve an average success rate of 86.20% across 50 RoboTwin 2.0 tasks and 97.40% across the four LIBERO suites. These results establish the execution performance of the task-adapted low-level VLA policies used in EmbodiedSkills. On four memory-dependent RMBench tasks, the same task-adapted execution approach achieves 12.5% average success. The framework provides a trainable and inspectable agent layer for turning these policies into closed-loop embodied systems.