SentryVLA：基于不确定性引导的动作选择与状态监测的视觉-语言-动作机器人控制
SentryVLA: Uncertainty-Guided Action Selection and State Monitoring for Vision-Language-Action Robotic Control
AI 导读
SentryVLA 是一个面向冻结 VLA 策略的轻量级运行时封装框架，无需重训练或修改底层模型架构。它在动作层面用共形化分位数回归估计动作偏差上界，据此对多个随机候选动作排序选择，并在执行监测层面用 Mahalanobis 距离异常分数检测异常执行状态。仿真与实体机器人实验表明该框架可改善动作选择并提供有效的运行时异常警示信号，成果将发表于 CoRL 2026。
正文 · 原文
Abstract:Vision-language-action (VLA) models have emerged as generalist robotic controllers that map visual observations and language instructions to continuous actions. However, pretrained VLAs typically provide limited information about the reliability of their generated actions, making deployment challenging when execution failures must be anticipated. We address this problem by introducing SentryVLA, a lightweight runtime wrapper for frozen VLA policies. At the action level, SentryVLA uses conformalized quantile regression to estimate a calibrated upper bound on the deviation between sampled VLA actions and expert-like behavior under the calibration distribution. The predicted deviation is then used to rank multiple stochastic candidates for action selection. At the execution-monitoring level, SentryVLA applies a Mahalanobis-distance-based anomaly score to the contextualized latent execution state extracted before final action prediction, providing a complementary signal for detecting abnormal execution states during closed-loop control. The framework does not require retraining or architectural modification of the underlying VLA policy. Experiments in simulation and on a physical robot show that SentryVLA improves action selection and provides useful runtime warning signals for abnormal execution states. These findings demonstrate the practical value of post-hoc reliability scoring for improving the runtime behavior of pretrained VLA policies. Project webpage: this https URL.
| Comments: | To be published in the 2026 Conference on Robot Learning (CoRL) |
| Subjects: | Robotics (cs.RO); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2604.16677 [cs.RO] |
| (or arXiv:2604.16677v2 [cs.RO] for this version) | |
| https://doi.org/10.48550/arXiv.2604.16677 arXiv-issued DOI via DataCite |
Submission history
From: William Beksi [view email]
[v1]
Fri, 17 Apr 2026 20:20:43 UTC (15,396 KB)
[v2]
Sun, 25 Oct 2026 02:55:19 UTC (22,852 KB)
来源：arXiv cs.AI · arxiv.org