城乡快速干道车-人冲突时间窗预测模型

Prediction Model for Vehicle-pedestrian Conflict Time Window on Urban and Rural Expressways

为了解决城乡快速干道车-人冲突和事故严重的问题,将车-人冲突点的分析方法扩展到车-人冲突时间窗的分析方法,构建一种行人运动轨迹实时监测和穿越时间预测相结合的车-人冲突时间窗组合预测模型。首先,分析行人违规穿越的实测数据,确定不同类别行人对应的穿越时间置信区间以及松弛时间;其次,根据运动学模型的预测结果判断行人的所属类别并初步确定行人的穿越时间,同时通过卡尔曼滤波算法对行人穿越过程进行实时监测;再次,融合运动学模型预测结果和卡尔曼滤波监测结果,确定最终的车-人冲突时间窗;最后,对所提出的组合预测模型进行标定和验证,并通过VISSIM仿真平台进行安全性能测试。模型验证结果表明:在正常情况下,该模型能够保障行人的安全且能兼顾松弛时间重置次数;在行人初始穿越速度过低或穿越前、中期存在持续低速的情况下,该模型可以通过多次松弛时间重置来解决模型的适用性问题。安全性能测试结果表明,在车辆行驶时间均值增加4.7%的情况下,安全车辆数占比增加了37.3%,车辆的后侵占时间(PET)测试值则增加53.8%。因此,与无松弛时间的预测模型相比,所提出的有松弛时间的车-人冲突时间窗预测模型能够在对交通效率影响较小的前提下,较大程度地提高车-人冲突的安全性。

To solve the problem of serious vehicle-pedestrian conflicts and accidents on urban and rural expressways, the analysis method of the vehicle-pedestrian conflict point was extended to the analysis method of the vehicle-pedestrian conflict time window, and then a prediction model combining real-time monitoring of pedestrian trajectory and crossing time was constructed. First, the measured data of illegal pedestrian crossings were analyzed to determine the confidence interval of crossing time and relaxation time corresponding to different types of pedestrians. Second, according to the outputs of the kinematic model, the category of pedestrians, and the pedestrian crossing time was preliminarily determined. Simultaneously, the pedestrian crossing process was monitored in real time using the Kalman filter algorithm. Third, the outputs of the kinematics and Kalman filter monitoring models were fused to determine the final vehicle-pedestrian conflict time window. Finally, the proposed combined prediction model was calibrated and verified, and its safety performance was tested using the VISSIM simulation platform. The results of model verification show that, under normal circumstances, this model can not only guarantee pedestrian safety but also balance times of relaxation reset. The applicability problem of the model by resetting the relaxation time several times when the initial pedestrian crossing speed is slow or when there is a continuous low speed in the early and middle of the crossing, is also solved. The results of the safety performance tests show that when the average running time increases by 4.7%, the proportion of safe vehicles increases by 37.3%, and the value of PET increases by 53.8%. Therefore, compared with the prediction model without relaxation time, the prediction model with relaxation time proposed in this paper can greatly enhance the security of vehicle-pedestrian conflicts with less effect on traffic efficiency.