自动驾驶安全评价框架对中国测评体系建设的启示

温 泉, 麻 晨, 陈 水源
中国汽车技术研究中心有限公司

摘要


近年来,国际汽车工业界提出了一套基于物理原理的自动驾驶安全评价框架(以下简称“参考框架”),该
框架以感知-决策-操作三层扰动分解、“称职谨慎驾驶员”量化基准以及三层场景抽象体系为核心,实现了从安全
原则到可测试场景的工程化映射,并直接支撑了联合国相关法规的制定。本文基于该框架的公开技术文档及相关独
立研究,系统阐述其核心技术逻辑,包括场景生成的正向推导方法、驾驶员模型的参数化定义及覆盖度验证方法。
其次,结合中国交通环境特征及已有实证研究,分析该框架在面向中国应用时的三大局限性:对复杂混行交通场景
覆盖不足、驾驶员行为参数与中国驾驶人群存在显著偏差、动力学边界假设与中国道路条件不匹配。在此基础上,
论述中国智能驾驶测评体系在场景数据积累、仿真技术及综合评价模型方面的现有基础,并提出基于本土数据的参
数修正、场景库补充及完备性论证方法等改进建议。本文旨在为构建适应中国交通特征的自动驾驶安全测评体系提
供理论依据与技术参考。

关键词


自动驾驶安全评价;物理原理驱动;场景库;称职谨慎驾驶员;中国交通特征;测评体系

全文:

PDF


参考


[1]Kalra N, Paddock S M. Driving to Safety: How Many

Miles of Driving Would It Take to Demonstrate Autonomous

Vehicle Reliability? RAND Corporation, 2016.

[2]JAMA. Automated Driving Safety Evaluation

Framework Ver.4.0. Japan Automobile Manufacturers

Association, 2026.

[3]UN Regulation No. 157 - Automated Lane Keeping

Systems (ALKS). 2021.

[4]中国汽车技术研究中心有限公司.中国智能网联

汽车技术规程(C-ICAP)V1.2. 2024.

[5]Zhang Z. Beyond Scenarios: Why AI-Driven Scenario

Testing Is Incomplete Without Formal Methods. Blog, 2026.

[6]SAKURA Project Annex E: Principle models and

evaluation scenarios for perception disturbances. In [4].

[7]Khastgir S, Birrell S, Dhadyalla G, et al. Calibrating trust

through knowledge: Introducing the concept of informed safety

for automation in vehicles. Transportation Research Part C,

2018.

[8]Makishita, et al. Emergency braking characteristics of

trained drivers. JSAE Review, 2001.

[9]Olleja P, et al. Validation of Human Benchmark Models

for Automated Driving System Approval: How Competent and

Careful Are They Really? Accident Analysis & Prevention, 2025.

[10]Otte D, Krettek C, Brunner H, Zwipp H. German

In-Depth Accident Study (GIDAS). 2003.

[11]ITARDA. Institute for Traffic Accident Research

and Data Analysis annual reports. 2018-2019.

[12]CIDAS. China In-Depth Accident Study database

report. 2023.

[13]张 玉 新.日 本SAKURA自 动 驾 驶 安 全 评 价 框

架V4.0价值与挑战.微信公众号“张玉新-AutoZYX”,

2026-03-21.

[14]Liu R, et al. Comparison Between Cut-In

Performance Model of ECE R157 and Behaviors of Chinese

Drivers. Automotive Innovation, 2025. Springer.

[15]Li Z, et al. Development of Human-Like Automated

Driving Following Rules: A Comparison between China and

Germany. Transportation Planning and Technology, 2025.

Taylor & Francis.


Refbacks

  • 当前没有refback。