Rewind-IL: Online Failure Detection and State Respawning for Imitation Learning

TL;DR

Rewind-IL is a training-free online safeguard framework that detects failures in imitation learning policies and respawns the robot at safe states, improving reliability in long-horizon manipulation tasks.

Contribution

It introduces a novel zero-shot failure detector and a state-respawning mechanism using vision-language models and policy checkpoints, enhancing imitation learning robustness.

Findings

Rewind-IL improves failure detection accuracy in real-world tasks.

The framework enables recovery and continued operation after failures.

Experiments show increased reliability in long-horizon manipulation tasks.

Abstract

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