摘要
近年,基于网联车辆轨迹数据的交通管控与服务研究方兴未艾。其中,信号控制交叉口排队长度估计备受关注。然而,在低渗透率条件下,单个周期内轨迹稀少且提供的交通信息十分有限。现有研究仅以当前周期内网联车辆轨迹数据为输入,难以获得准确且可靠的周期级排队长度估计结果。因此,融合利用历史网联车辆轨迹数据提供的车辆到达和停车位置信息以及当前周期内实时观测的网联车辆排队信息,提出一种基于最大后验概率的周期最大排队长度估计方法。首先,依据历史轨迹数据的停车位置信息,估计排队长度的先验分布;其次,依据历史轨迹数据的车辆到达信息,估计周期内车辆的历史到达分布,并结合周期内最后1辆排队网联车辆的到达时刻与停车位置,构建排队长度似然函数;最后,基于贝叶斯理论,结合前述先验分布与似然函数,推导周期排队长度的后验分布,并采用最大后验概率方法实现周期最大排队长度的估计。仿真结果表明:所提方法在不同饱和度和渗透率条件下,均优于现有的方法;即使在车辆轨迹数不超过1 veh·周期^(-1)的低渗透率条件下,所提方法的平均绝对估计误差也不超过2 veh·周期^(-1)。实证结果表明:在渗透率仅为8.96%的条件下,所提方法的平均绝对误差为2.12 veh·周期^(-1),平均相对估计误差为12.4%,同样优于现有同类方法。
In recent years,studies on traffic control and services based on connected vehicles(CVs)have been carried out.Among these studies,queue length estimation at signalized intersections has attracted much attention.However,under the condition of a low penetration rate,CV trajectories observed in a single cycle are scarce,and the traffic information provided in the cycle is very limited.Using only CV trajectories in the cycle as input,existing studies have difficulties in obtaining accurate and reliable cycle-based queue length estimation results.Therefore,by fusing the vehicle arrival distribution and queuing position information provided by historical CV trajectories and the queuing information of real-time observed CV in the cycle,this study proposes a cycle-based maximum queue length estimation method based on the Maximum A Posteriori(MAP).First,the prior distribution of the queue length was estimated according to the queuing position information of the historical trajectories.Then,based on the arrival information of historical trajectories,the historical cycle arrival distribution of vehicles was obtained.Given the arrival and queue information of the last queued connected vehicle during the cycle,the likelihood function of the queue length was derived.Finally,based on Bayes’theorem,the posterior distribution of queue length was derived by integrating the aforementioned prior distribution and likelihood function of the queue length,and the MAP estimation was adopted to achieve cycle-based queue length estimation.The simulation results show that the proposed method outperforms the existing method with different degrees of saturation and penetration rates.Even under the condition of low penetration rates,that is,there is no more than one vehicle trajectory per cycle on average,the mean absolute error of the proposed method is no more than two vehicles per cycle.The empirical results show that when the penetration rate is only 8.96%in the real world,the mean absolute error of the proposed method is 2.12 vehicles per cycle,and the mean absolute percentage error is 10.4%,which also outperforms the existing method.
作者
谈超鹏
姚佳蓉
曹喻旻
唐克双
TAN Chao-peng;YAO Jia-rong;CAO Yu-min;TANG Ke-shuang(Key Laboratory of Road and Traffic Engineering,Ministry of Education,Tongji University,Shanghai 201804,China;College of Transportation Engineering,Tongji University,Shanghai 201804,China)
出处
《中国公路学报》
EI
CAS
CSCD
北大核心
2021年第7期140-151,共12页
China Journal of Highway and Transport
基金
国家自然科学基金项目(61673302)。
关键词
交通工程
排队长度估计
最大后验概率估计
网联车辆轨迹
历史轨迹数据
先验分布
traffic engineering
queue length estimation
maximum a posteriori estimation
connected vehicle trajectory
historical trajectory data
prior distribution
