摘要
为对高速公路尾撞事故进行有效预警,提出一种宏观交通流尾撞事故预警模型。首先考虑不同车辆类型制动反应时间的影响,建立空余安全间距模型,以空余安全间距■=0为尾撞事故临界条件,建立尾撞事故预警模型,依据车辆类型(液压或气压)计算不同车辆比例下制动反应时间的取值范围,确定其置信区间,通过所允许的平均制动反应时间最大值对交通流状态进行预警分级并确定等级概率;然后利用长短时记忆神经网络模型对预警特征参数交通流量q、速度v s进行预测;最后通过实例验证,将预测数据代入尾撞事故预警模型,得到不同车辆类型比例下预警等级的概率,且概率值远大于50%,验证了尾撞预警模型的适用性,对预警等级较低且概率较高的情况及时进行管控,减少多车连环追尾事故的发生。
In order to effectively warn the rear-end collision in expressway,this paper proposed a macro traffic flow rear-end collision accident warning model.Firstly,considering the influence of braking reaction time of different vehicle types,a free safety spacing model was established to establish the rear-end collision accident warning model as the critical condition of rear-end collision accident.According to the vehicle type(hydraulic pressure or air pressure),the value range of braking response time under different vehicle proportions was calculated,and the confidence interval was determined.The traffic flow state was graded and the grade probability was determined by the maximum allowed average braking response time.Then the long and short term memory neural network model was used to predict the characteristic parameters of traffic flow and speed.Finally,through real case verification,data were put into the rear-end collision warning model,then the probability of early warning level under different proportion of vehicle type was obtained,and the probability value was greater than 50%,which verified the applicability of the rear-end collision warning model proposed,realized timely control over cases with low warning level but higher probability,and reduce the happening of serial rear-end collision accidents.
作者
赵雯雯
王晶晶
Zhao Wenwen;Wang Jingjing(School of Traffic&Transportation,Chongqing Jiaotong University,Chongqing 400041,China;Business Department,Chongqing Highway Engineering Testing Center Co.Ltd.,Chongqing 400041,China)
出处
《石家庄铁道大学学报(自然科学版)》
2022年第4期60-66,共7页
Journal of Shijiazhuang Tiedao University(Natural Science Edition)
关键词
公路交通
追尾碰撞
长短时记忆神经网络
预警模型
制动反应时间
highway traffic
rear-end collision
long and short-term memory neural network
early warning model
brake response time