让机器人在模仿学习中出错时自动回退到安全状态，提升长序列操作可靠性。
Rewind-IL: Online Failure Detection and State Respawning for Imitation Learning
- 用时间片段间差异检测失败，无需训练数据。
- 发现错误后回退到已验证的安全中间状态重新执行。
- 适合需要高可靠性的机器人长期任务部署场景。
模仿学习使机器人能从示范中习得复杂视觉-运动操控技能，但部署中的失败仍是主要障碍，尤其对长时程分段动作策略而言。一旦执行偏离示范流形，这些策略常继续生成局部合理动作却无法恢复。现有运行时监控器或需失败数据、易在正常特征漂移下误触发，或仅在检测失败后停止，缺乏恢复机制。本文提出Rewind-IL，一种针对生成式分段动作模仿策略的免训练在线保护框架。该框架结合基于时间片段间差异估计（TIDE）的零样本失败检测器（经分割置信区间校准），以及状态重生机制，将机器人返回至语义验证的安全中间状态。离线阶段，通过视觉-语言模型识别示范中的恢复检查点，并利用冻结的策略编码器构建紧凑检查点特征库；在线阶段，监测重叠动作段的自一致性，追踪与检查点库的相似性，一旦检测到失败，则回溯至最新验证的安全状态，从干净策略状态重启推理。真实世界与模拟环境中的长时程操控任务实验（包括迁移至流匹配分段策略）表明，策略内部一致性与语义基础的重生相结合，为提升模仿学习的可靠性提供了实用路径。补充材料见 https://sjay05.github.io/rewind-il
原文摘要 · Abstract (English)
Imitation learning has enabled robots to acquire complex visuomotor manipulation skills from demonstrations, but deployment failures remain a major obstacle, especially for long-horizon action-chunked policies. Once execution drifts off the demonstration manifold, these policies often continue producing locally plausible actions without recovering from the failure. Existing runtime monitors either require failure data, over-trigger under benign feature drift, or stop at failure detection without providing a recovery mechanism. We present Rewind-IL, a training-free online safeguard framework for generative action-chunked imitation policies. Rewind-IL combines a zero-shot failure detector based on Temporal Inter-chunk Discrepancy Estimate (TIDE), calibrated with split conformal prediction, with a state-respawning mechanism that returns the robot to a semantically verified safe intermediate state. Offline, a vision-language model identifies recovery checkpoints in demonstrations, and the frozen policy encoder is used to construct a compact checkpoint feature database. Online, Rewind-IL monitors self-consistency in overlapping action chunks, tracks similarity to the checkpoint library, and, upon failure, rewinds execution to the latest verified safe state before restarting inference from a clean policy state. Experiments on real-world and simulated long-horizon manipulation tasks, including transfer to flow-matching action-chunked policies, demonstrate that policy-internal consistency coupled with semantically grounded respawning offers a practical route to improved reliability in imitation learning. Supplemental materials are available at https://sjay05.github.io/rewind-il
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。