基于感知契约原像校正学习型感知以保障安全

AI 摘要

论文提出一种针对基于机器学习感知的安全校正方法，用于自动驾驶等自主系统。该方法先离线计算感知契约的原像以刻画ML状态估计的不确定性，再在运行时用风险启发式从不确定估计中选择状态来驱动控制决策。在45个使用Yolo和LaneNet导致安全违规的ACC场景中，该方法在73%的场景中保持了安全，平均完成时间仅增加2.8%。

正文节选

Correcting Learning-based Perception for Safety Abstract Learning-enabled perception is important in many autonomous systems. Unlike traditional sensors, the boundary where ML perception does or does not work is poorly characterized. Incorrect perception can lead to unsafe or overtly conservative downstream control actions. In this paper, we propose a two-step strategy for correcting ML-based state estimation. First, an offline computation is used to characterize the uncertainties resulting from

发布时间：2026-09-22 12:00

抓取时间：2026-09-22 12:05

来源机构：arXiv