汽车工程 ›› 2024, Vol. 46 ›› Issue (3): 375-382.doi: 10.19562/j.chinasae.qcgc.2024.03.001

• •    

自动驾驶仿真多逻辑场景综合评价方法

张培兴1,秦孔建2,朱冰1(),赵健1,范天昕1,赵文博3   

  1. 1.吉林大学,汽车底盘集成与仿生全国重点实验室,长春 130025
    2.中国汽车技术研究中心有限公司,中汽研汽车检验中心,天津 300300
    3.国汽(北京)智能网联汽车研究院有限公司仿真及传感器综合实验室,北京 102607
  • 收稿日期:2023-08-09 修回日期:2023-08-22 出版日期:2024-03-25 发布日期:2024-03-18
  • 通讯作者: 朱冰 E-mail:zhubing@jlu.edu.cn
  • 基金资助:
    国家重点研发计划(2022YFB2503401);国家自然科学基金(U22A20247);吉林省科技发展计划(212551GX010390619);吉林省长春市重大科技专项(20220301009GX)

Comprehensive Evaluation Method for Automated Vehicle in Multiple Virtual Logical Scenarios

Peixing Zhang1,Kongjian Qin2,Bing Zhu1(),Jian Zhao1,Tianxin Fan1,Wenbo Zhao3   

  1. 1.Jilin University,National Key Laboratory of Automobile Chassis Integration and Bionics,Changchun  130025
    2.China Automotive Technology & Research Center Co. ,Ltd. ,China Automotive Research Automotive Inspection Center,Tianjin  300300
    3.Integrated Simulation and Sensor Laboratory,China Intelligent and Connected Vehicles (Beijing) Research Institute Co. ,Ltd. ,Beijing  102607
  • Received:2023-08-09 Revised:2023-08-22 Online:2024-03-25 Published:2024-03-18
  • Contact: Bing Zhu E-mail:zhubing@jlu.edu.cn

摘要:

基于场景的仿真测试方法已成为解决自动驾驶汽车性能验证的核心思路,该方法将车辆连续驾驶过程进行分割获取非重复的独立场景片段,并在仿真环境中进行测试。与测试过程相贴合,本文提出了一种自动驾驶仿真多逻辑场景综合评价方法。首先建立自动驾驶汽车多逻辑场景综合评价方法,明确兼顾场景自身特征信息及仿真过程信息的场景权重分析流程;通过暴露度、失控度、危害度建立逻辑场景自身特征信息权重;建立包含仿真精度信息、要素种类信息、参数空间信息、离散步长信息的仿真过程信息权重;基于HighD数据集抽取前车制动、前车左侧切入、前车右侧切入场景信息,并基于本文提出的方法及基准算法测试结果计算场景权重,最终获取两种被测算法在多逻辑场景中的综合评价结果。

关键词: 自动驾驶汽车, 仿真测试, 综合评价, 逻辑场景, 权重分析

Abstract:

The scenario-based simulation testing method has become the core idea for automated vehicle performance verification, which splits the continuous vehicle driving process to obtain non-repetitive and independent scenario segments and tests them in virtual environment. Fitting in with the test process, a comprehensive evaluation method for automated vehicle simulation testing in multi-logical scenarios is proposed in this paper. Firstly, a comprehensive evaluation method for automated vehicle in multi-logical scenarios is established and the scenarios weighting analysis process considering both the scenario's own characteristic information and simulation test process information is defined. Then, the logical scenario's own characteristic information weighting is built by exposure degree, control loss degree and hazard degree. The simulation test process information weighting is built by simulation accuracy information, element type information, parameter space information and discrete step information. Finally, the information of front braking, front left cut-in and front right cut-in scenarios is extracted based on HighD dataset and the scenario weight is calculated through the method proposed in this paper and base algorithm test results, and the two tested algorithms’ comprehensive evaluation results are obtained in multi-logical scenarios.

Key words: automated vehicle, simulation test, comprehension evaluation, logical scenario, weighting analysis