具身开发者社区

外观

论文arXiv 2610.0794622 天前

在执行期间使 Vision-Language-Action 模型适应未知视觉扰动 ​

针对执行中未知视觉扰动，SALT用剩余轨迹自监督测试时适应，并在LIBERO-10和真机上验证。

AI 整理
  • 本体 SDK 与控制
  • 数据与采集
  • 具身模型

原文链接：Adapting Vision-Language-Action Models to Unknown Visual Disruptions During Execution（arXiv 2610.07946 · 发布日期：2026-10-06）

Adapting Vision-Language-Action Models to Unknown Visual Disruptions During Execution

针对执行中未知视觉扰动，SALT用剩余轨迹自监督测试时适应，并在LIBERO-10和真机上验证。

核心要点 ​

  • 执行任务时视觉扰动类型和时机未知，VLA策略需要在线应对。
  • SALT将上一动作块未执行的剩余轨迹作为当前预测的时间对齐监督。
  • 监督完全来自策略自身预测，无需扰动标注、专家动作或目标域演示。
  • LIBERO-10上两种VLA平均成功率提升，真机任务进展由0.49升至0.61。

验证范围与局限：验证范围为LIBERO-10五种持续视觉腐蚀及一台真实机器人上的数字和物理扰动。

英文摘要 (Abstract) ​

Visual disruptions can arise while a robot is executing a task, leaving a vision-language-action (VLA) policy to respond without knowing the disruption type or timing. We introduce Self-supervised Adaptation from Leftover Trajectories (SALT), which uses the leftover trajectory, the unexecuted part of the previous action chunk, as self-supervision for test-time adaptation. Because consecutive chunks overlap in time, the leftover provides a temporally aligned target for the current prediction over the same future control interval. At the onset of a visual shift, the leftover can retain a plan formed before the corruption, so updating the policy toward it anchors the adaptation across the shift (Transition Anchoring). SALT keeps the adapted policy and regenerates the current chunk, whose leftover becomes the target at the next replan, carrying the correction forward along the execution trajectory (Sequential Correction Propagation). Supervision comes entirely from the policy's own predictions, requiring no disruption annotations, expert actions, or target-domain demonstrations, and a lightweight adaptation gate calibrated only on nominal trajectories decides when updates begin. On LIBERO-10, SALT increases average success across five persistent visual corruptions from 43.9% to 53.2% with SmolVLA and from 58.7% to 66.0% with GR00T N1.7, while largely preserving nominal performance. On a real robot, it raises task progress averaged over digital and physical disruptions from 0.49 to 0.61.

作者：Ahin Lee, Jinwoo Seo, Youngsoo Jang, Taesik Gong

来源

本页事实对应的公开页面，以原文为准。

  1. arXiv 2610.07946：Adapting Vision-Language-Action Models to Unknown Visual Disruptions During Execution访问 2026-10-21