变道切入场景下ADAS系统测试与评价研究

为了满足变道切入场景下的ADAS系统测试评价需求，提出一种考虑场景风险系数的变道切入场景生成方法和客观综合评价模型。通过采集自然驾驶数据，采用阈值法自动提取变道切入功能场景并深入分析变道切入行为特征。使用单因素方差分析法与皮尔逊相关性检验法共同分析场景风险系数与场景要素的相关性来确定关键场景要素。结合Kmeans算法对离散逻辑场景参数进行聚类，从而得到5个典型测试场景。基于场景风险系数，采用AHP与CRITIC法构建多层次综合评价模型，采用灰色关联理论对ADAS系统进行客观评价。借助VTD仿真软件构建变道切入虚拟测试场景库，进行仿真试验验证。结果表明，相关性分析使场景要素维度降低了60%，生成的测试场景可以有效验证ADAS系统的综合性能，综合评价模型可对ADAS系统表现进行客观有效的评价，为智能驾驶系统开发提供有效参考。

Research on Testing and Evaluation of ADAS Systems in Lane Change Cut-in Scenarios

To meet the requirements for testing and evaluating ADAS systems in lane change cut-in scenarios, the paper proposes a method for generating such scenarios and an objective, comprehensive evaluation model considering the scenario risk coefficients. By collecting natural driving data, the threshold method is used to automatically extract lane change cut-in function scenarios and deeply analyze the lane change cut-in behavior characteristics. The correlation between scenario risk coefficients and scenario elements is jointly analyzed using one-way ANOVA and Pearson''s correlation test to identify key scenario elements. Furthermore, by applying the K-means clustering method to the parameters of discrete logic scenarios, five typical test scenarios are obtained. Based on the scenario risk coefficient, the AHP and CRITICmethods are used to construct a multi-level comprehensive evaluation model. The ADAS system is objectively evaluated using the gray correlation theory. Finally, the VTD simulation software is used to create a virtual test scene library for lane change cut-in scenarios for simulation testing and validation. The results show that correlation analysis reduces the dimensionality of scenario elements by 60%. The generated test scenarios can effectively validate the comprehensive performance of the ADAS system. Moreover, the comprehensive evaluation model can objectively and effectively evaluate the performance of the ADAS system, providing a valuable reference for the development of intelligent driving systems.