发表机构
College of Connected Computing, Vanderbilt University; School of Computer Science, The University of Sydney; Australian Centre for Robotics, The University of Sydney（范德堡大学连接计算学院; 悉尼大学计算机学院; 悉尼大学澳大利亚机器人中心）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
ContactGuard是基于动作条件潜世界模型的接触前执行监控器，通过预测接触前动作的潜态后果判断是否失败，能准确预测接触型操纵任务失败，可迁移至真实机器人且无需修改底层策略。
AI 中文摘要
接触丰富型操纵任务的失败往往仅在机器人已完成接触后才被检测到，这在腕部相机设置中尤为受限：夹爪与物体的近距离视图有助于观察接触，但常规检测器做出反应前，不佳的接近策略可能已推动、错过、滑动或干扰物体。我们提出ContactGuard，一种针对分块视觉运动策略的接触前执行监控器。给定策略规划的动作分块，ContactGuard在潜视觉空间中预测其短程后果，若预测的未来潜态表明可能失败则中止。其潜世界模型从未标记的机器人轨迹中训练，以在规划动作下预测紧凑的多视图视觉嵌入，避免像素级视频预测。随后从少量带标签的接触前片段中训练轻量型失败探测模型。部署时，ContactGuard在即将发生的接触事件前锚定预测，在策略自身动作下将模型前向滚动，并验证预测的接触后潜态。在真实世界接触丰富型操纵任务中，ContactGuard比直接和受损动作的 ablation 更准确地预测失败，且作为接触前弃权（不执行）信号可迁移到真实机器人，无需修改底层策略。
英文摘要
Contact-rich manipulation failures are often detected only after the robot has committed to contact. This is especially limiting in wrist-camera setups: close gripper--object views help observe contact, but a poor approach may already push, miss, slip, or disturb the object before conventional detectors react. We introduce \emph{ContactGuard}, a pre-contact execution monitor for chunked visuomotor policies. Given the policy's planned action chunk, ContactGuard predicts its short-horizon consequence in latent visual space and aborts if the predicted future latent indicates likely failure. Its latent world model is trained from unlabelled robot trajectories to predict compact multi-view visual embeddings under planned actions, avoiding pixel-level video prediction. A lightweight failure probe is then trained from a small labelled set of pre-contact clips. At deployment, ContactGuard anchors prediction before an imminent contact event, rolls the model forward under the policy's own actions, and verifies the predicted post-contact latent. Across real-world contact-rich manipulation tasks, ContactGuard predicts failure more accurately than direct and corrupted-action ablations, and transfers to live robot as a pre-contact abort signal without modifying the underlying policy.
Comments14 pages, 5 figures, 8 tables
