| 基于车辆行驶轨迹的道路不良驾驶行为谱构建与特征值计算方法 |
| |
| 引用本文： | 王可,陆键,蒋愚明.基于车辆行驶轨迹的道路不良驾驶行为谱构建与特征值计算方法[J].交通运输工程学报,2020,20(6):236-249. |
| |
| 作者姓名： | 王可 陆键 蒋愚明 |
| |
| 作者单位： | 同济大学 交通运输工程学院，上海 201804 |
| |
| 基金项目： | 国家自然科学基金;国家重点研发计划 |
| |
| 摘 要： | 为了量化描述不同道路驾驶场景下驾驶行为的动态变化过程与不良驾驶程度, 研究了不良驾驶行为谱的构建与分析方法; 基于车辆行驶轨迹关键参数建立驾驶行为谱; 应用风险度量方法量化4种不良驾驶行为, 包括不良跟驰、蛇形驾驶、速度不稳与不良换道; 基于驾驶行为谱建立了不良驾驶行为谱; 基于交通流量-密度关系与驾驶行为统计参数的差异对交通流状态进行划分; 在不同交通流状态下, 使用四分位差法确定了不良驾驶行为特征参数阈值; 基于特征参数阈值计算每个驾驶人的不良驾驶行为得分; 使用CRITIC赋权法确定了不良驾驶行为的权重, 为每个驾驶人计算不良驾驶行为谱特征值; 为了验证方法的有效性, 使用无人机交通视频采集了上海市的车辆行驶轨迹数据, 分析了小汽车不良驾驶行为特征; 通过专家打分的方法对不良驾驶行为谱特征值进行验证。分析结果表明: 基于驾驶行为参数的交通流状态聚类方法将数据中的交通流状态分为自由流、饱和流、拥堵流3类; 聚类方法比基于基本图的交通流状态划分方法更适合驾驶行为分析; 不同交通流状态下的不良跟驰、蛇形驾驶、速度不稳特征参数分布明显不同, 拥堵流状态下的不良跟驰、蛇形驾驶、速度不稳极端值出现更频繁, 而不良换道特征参数在各交通流状态下有相似的分布; 蛇形驾驶、速度不稳、不良换道的特征参数阈值随交通流密度上升而上升; 使用CRITIC赋权法计算的不良跟驰、蛇形驾驶、速度不稳、不良换道的权重分别为0.19、0.33、0.37、0.11;自由流、饱和流、拥堵流的不良驾驶行为谱特征值的分布范围相近, 均处于0与0.4之间; 专家的不良驾驶行为评价与不良驾驶行为谱特征值一致。可见, 不良驾驶行为谱的构建与特征值计算方法能够使用车辆行驶轨迹数据自动辨识不良驾驶人, 具有客观性、适应性以及可靠性, 能及时发现不良驾驶人, 给驾驶人提供安全提示, 为交通管理部门提供交通安全预警的技术支持。
|
| 关 键 词： | 交通安全 驾驶安全评价 不良驾驶行为 驾驶行为谱 风险度量 阈值计算 交通流状态 |
| 收稿时间： | 2020-09-11 |
Abnormal road driving behavior spectrum establishment and characteristic value calculation method based on vehicle driving trajectory |
| |
| WANG Ke,LU Jian,JIANG Yu-ming.Abnormal road driving behavior spectrum establishment and characteristic value calculation method based on vehicle driving trajectory[J].Journal of Traffic and Transportation Engineering,2020,20(6):236-249. |
| |
| Authors: | WANG Ke LU Jian JIANG Yu-ming |
| |
| Affiliation: | College of Transportation Engineering, Tongji University, Shanghai 201804, China |
| |
| Abstract: | To quantitatively describe the dynamic change process of driving behavior and the degree of abnormal driving under different road driving scenarios, the establishment and analysis methods of abnormal driving behavior spectrum were studied. The driving behavior spectrum based on the key parameters of vehicle driving trajectory was established. The measurement of risk method was applied to quantify four types of abnormal driving behaviors, including abnormal car-following, serpentine driving, speed instability, and abnormal lane-changing. An abnormal driving behavior spectrum was established based on the driving behavior spectrum. Traffic flow conditions were divided based on the traffic volume-density relationship and the differences among the statistical parameters of driving behavior. Under different traffic flow conditions, the thresholds of the characteristic parameters of abnormal driving behaviors were determined by using the interquartile range method. The abnormal driving behavior scores of each driver were calculated based on the characteristic parameter thresholds. The weights of abnormal driving behaviors were determined by using the CRITIC weighting method and the characteristic values of the abnormal driving behavior spectrum for each driver were calculated. To verify the effectiveness of the method, vehicle driving trajectory data were collected by an unmanned aerial vehicle traffic video in Shanghai and the characteristics of car abnormal driving behavior were analyzed. The characteristic values of the abnormal driving behavior spectrum were verified by the expert scoring method. Analysis result shows that the traffic flow condition clustering method based on driving behavior parameters divides the traffic flow condition of data into three categories: free flow, saturated flow, and congested flow. The clustering method is more suitable for driving behavior analysis than the traffic flow condition division method based on the fundamental diagram. The characteristic parameter distributions of abnormal car-following, serpentine driving, and speed instability under different traffic flow conditions are significantly different. The occurrence of extreme values of abnormal car-following, serpentine driving, and speed instability under congested flow conditions is more frequent, while the abnormal lane-changing characteristic parameter has a similar distribution under each traffic flow condition. The thresholds of serpentine driving, speed instability, and abnormal lane-changing increase with the increase of traffic flow density. The weights of abnormal car-following, serpentine driving, speed instability, and abnormal lane-changing calculated by the CRITIC weighting method are 0.19, 0.33, 0.37, and 0.11, respectively. The distribution ranges of the abnormal driving behavior spectrum characteristic values under free flow, saturated flow, and congested flow are similar, all between 0 and 0.4. The expert's abnormal driving behavior evaluation is consistent with the abnormal driving behavior spectrum characteristic values. The establishment of an abnormal driving behavior spectrum and the calculation method of characteristic values can automatically identify abnormal drivers by using the vehicle driving trajectory data. The method is objective, adaptable, and reliable, and can detect abnormal drivers in time, providing drivers with safety tips and technical support for traffic safety early warning to traffic management departments. |
| |
| Keywords: | |
| 本文献已被 万方数据 等数据库收录！ |
| 点击此处可从《交通运输工程学报》浏览原始摘要信息 |
|
点击此处可从《交通运输工程学报》下载全文 |
|
