考虑交通运行条件影响的驾驶员特征聚类

Driver Characteristics Clustering Under Impact of Varying Traffic Operation Conditions

为提升驾驶员特征聚类方法的适用性与可靠性,本文基于机动车运行轨迹分析提出考虑交通运行条件影响的驾驶员特征聚类改进方法。首先,经过对车辆运行轨迹数据的分析发现,不同道路类型和平均速度条件会显著影响驾驶行为的集计特征;其次,提出改进的驾驶员特征聚类方法,第1步设计考虑道路类型与平均速度因素的车辆轨迹的切片和分类方法,从而稳定提取典型交通条件下的驾驶行为特征参数,第2步选用高斯混合模型聚类驾驶员特征。聚类案例表明,在相同的道路类型和平均速度条件下,驾驶员类型越激进,其速度变异系数、加速度标准差和平均减速度等参数均值越高。不同聚类方法的对比表明,改进方法在驾驶员聚类的类内聚集度和类间分离度方面均表现更好,能有效提升驾驶员聚类的适用性与可靠性。

To improve the stability and reliability of driver characteristics clustering,this paper proposes an improved driver clustering method considering the impact of varying traffic conditions using motor vehicle trajectory data.First,the analysis of vehicle trajectories indicates that different road types and average speeds significantly affect the aggregate characteristics of driving behavior.Considering the road types and average speed,a slicing and classification method of vehicle trajectory is designed to steadily extract the driving behavior characteristics under typical traffic conditions.Then,the driver characteristics are clustered using the Gaussian Mixture Model(GMM).Data analysis demonstrates that road types and average speed can significantly affect driving behavior characteristics.The clustering results show that the improved clustering method shows better performance on intra-class aggregation and inter-class separation,and this method can improve the applicability and reliability of driver clustering.