基于车-车事故的道路测试场景风险评价方法

Risk Assessment Method of Road Test Scenario Based on Vehicle-to-Vehicle Crashes

为建立适用于中国自动驾驶车辆测试场景库的风险评价体系,统计分析国家车辆事故深度调查体系数据库的车-车事故数据,得到6类典型车-车事故场景;根据事故造成的损失提出车-车事故致损模型,并利用信息熵理论构建包含静态和动态场景要素的综合复杂度模型;结合致损和复杂度2个维度建立自动驾驶测试场景的风险评价模型,依据风险指数的分布,为车-车事故场景风险等级界定阈值.以6类典型事故场景为例,采用提出的评价方法评价其风险等级.研究表明:该方法综合考虑了经济损失及场景复杂程度,能合理划分其风险程度,有助于为自动驾驶车辆测试选取合适的场景、推进自动驾驶道路测试场景评价进程.

In order to establish a risk assessment system applicable to China autonomous driving vehicle test scenario database,the vehicle-to-vehicle crash data of national Vehicle Crash In-depth Investigation System database were statistically analyzed,and six typical vehicle-to-vehicle crash scenarios were obtained.According to the loss caused by the crash,the vehicle-to-vehicle crash damage model is proposed,and the comprehensive complexity model including static and dynamic scene elements is constructed by using the information entropy theory.The risk assessment model of automatic driving test scenarios was established by combining the two dimensions of loss and complexity.According to the distribution of risk index,the risk level threshold of vehicle-to-vehicle crash scenarios was defined.The risk level of six typical crash scenarios was evaluated by the proposed method.The results show that this method comprehensively considers the economic loss and the complexity of the scene,and can reasonably divide the risk degree,which is helpful to select the appropriate scene for the autonomous vehicle test and promote the evaluation process of the autonomous road test scenario.