面向自动驾驶运行风险的高风险关键道路辨识

打开文本图片集

中图分类号：F570 文献标志码：ADOI： 10.13714/j.cnki.1002-3100.2025.09.018

Abstract： Autonomous driving technology has been deployed in some open road areas， exposing a series safety risks. These safety risks accidents frequently occur on high-risk critical roads within operational areas. To enhance safety management risk prevention for autonomous driving， it is essential to identify high-risk critical roads. This paper pro posesa method for identifying high-risk roads based on autonomous driving operational risks. Firstly，a structured risk scenario repository is constructed. networks are segmentedinto topological unitsvia feature-based division. Then， LLMs are employed to generate driving scenarios convert m into topological graphs. Finally， sub graph isomorphism matching is implemented between scenarios repository， with criticality ranking enabling accurate identification high-risk road sections. Compared with state---art methods， proposed approach demonstrates substantial improvements in both recognition accuracy computational efficiency.

Keywords： traffic engineering;high-risk critical roads identification; topological matching; autonomous driving; large language model

0引言

随着自动驾驶技术的快速发展，安全问题仍然是制约其规模化和商业化的关键瓶颈。（剩余8583字）