针对高速匝道交通场景提出了一种基于自然驾驶数据的自动驾驶汽车安全性评价方法。通过分析合流区冲突特征，建立碰撞时间(Time to Collision，TTC)模型、后侵占时间(Post Encroachment Time，PET)模型、最小安全换道距离(Minimum Safe Spacing，MSS)交通冲突指标计算模型来确定安全性评价指标，利用模糊聚类自然驾驶指标数据确定指标阈值范围，搭建自动驾驶汽车仿真测试试验，应用层间相关性的重要性准则权重分配方法和灰色关联评分模型，计算得到不同控制算法下自动驾驶汽车安全性的综合评价得分。结果表明，被试车辆驾驶行为与理想驾驶行为在各安全性指标的关联度分布明显，计算总体关联度，得分可以直观说明不同自动驾驶系统的综合安全性能。
Research on Safety Testing and Evaluation Methods for Autonomous Vehicles in Ramp Merging Traffic Scenarios
In order to promote the development of autonomous vehicle applications,conducting accurate and reliable safety testing and evaluation is essential.This paper proposes a safety evaluation method for autonomous vehicles tailored to high-speed ramp traffic scenarios using natural driving data.By analyzing the conflict characteristics in the confluence area,the models for calculating traffic conflict indicators such as TTC,PET and MSS are established to determine the safety evaluation indicators.The fuzzy clustering of natural driving indicator data is used to obtain the threshold ranges for these indicators.The autonomous vehicle simulation test has been built.The importance criterion weight distribution method based on interlayer correlation and the gray correlation scoring model are applied.The comprehensive evaluation scores regarding the safety of autonomous vehicles are calculated under different control algorithms.The results show a distinct correlation in the distribution of safety indices between the test vehicle's driving behavior and ideal driving behavior.By calculating the overall correlation degree,the scores can directly reflect the comprehensive safety performance of different autonomous driving systems.
国家科技期刊平台
NSTL
万方数据
维普
