Abstract: In order to prevent and reduce the occurrences of longitudinal rear-end collision accidents, a hierarchical early warning model of vehicle longitudinal collision based on internet of vehicles was proposed. Firstly, the framework of internet of vehicles based on next generation mobile communication network was constructed. The location information of the preceding vehicle and the target vehicle was collected by the internet of vehicles platform, and the location information was fused based on Kalman filter. The optimal value of vehicle position information at a certain time was obtained, and the real-time distance between the target vehicle and the preceding vehicle was obtained. Secondly, based on the real-time distance between the target vehicle and the preceding vehicle, and the critical early warning distance and critical braking distance obtained by analyzing the braking process of the vehicle, a hierarchical early warning model of vehicle longitudinal collision was established. Moreover, different warning information prompts were designed according to different warning levels, early warning parameters and hazard warning sensitivity parameters. Considering the high cost and safety aspects of vehicle early warning test in the field, the simulation platform of internet of vehicle based on VISSIM and OMNeT ++ was established. The simulation test results show that: compared with the vehicle with traditional fixed threshold warning algorithm, the false alarm rate of the hierarchical early warning model based on the internet of vehicles is reduced by 14.4% and the missing alarm rate is reduced by 3.5%. On the whole, the hierarchical early warning model of vehicle longitudinal collision based on internet of vehicles can reduce the occurrence of longitudinal traffic rear-end collisions and improve the safety of driving more effectively.

Key words: traffic and transportation engineering, internet of vehicles, longitudinal collision, Kalman filter, hierarchical early warning, simulation platform

摘要： 为了预防和减少车辆纵向追尾事故的发生，提出了一种基于车联网的车辆纵向碰撞分级预警模型。首先，构建了基于下一代移动通信网络的车联网框架，通过车联网平台采集前车和自车的位置信息，基于卡尔曼滤波进行位置信息的融合，得到某一时刻车辆位置信息的最优值，得到自车和前车的实时距离。其次，基于自车和前车的实时距离以及通过对车辆制动过程分析得到的临界预警距离和临界制动距离建立了车辆纵向碰撞分级预警模型，并且根据不同的预警等级以及预警参数和危险警告灵敏度参数设计了不同的预警信息提示。考虑到实地进行车辆预警试验存在着成本高和安全性等方面问题，搭建了基于VISSIM和OMNeT++的车联网仿真平台。仿真实验结果表明：相比较于传统固定阈值预警算法的车辆，基于车联网的分级预警模型的误警率降低了14.4%，预警漏报率降低了3.5%。综合来看，基于车联网的分级预警模型可以更加有效的减少纵向交通追尾事故的发生，提高了行车的安全性。

关键词: 交通运输工程, 车联网, 纵向碰撞, 卡尔曼滤波, 分级预警, 仿真平台