摘要
路径流量是精细化交通规划和管控的基础,对识别路网的关键通道、路径、流向和节点具有重要作用。为了解决现有路径流量估计方法在模型假设和方法适用性上存在的局限性,提出一种以抽样车辆轨迹数据作为唯一输入数据源的信号控制路网路径流量估计方法。该方法对基于经典广义最小二乘法的OD估计模型进行改进拓展,以路网路径流量和交叉口流向流量的估计误差的加权和最小化为优化目标,建立一个基于广义最小二乘法的路径流量估计基本框架。首先,基于奇异值阈值算法和各交叉口的抽样车辆轨迹数据估计路网中所有交叉口受控流向的到达流量,从而计算得到交叉口全流向流量,以及不同流向的抽样车辆轨迹渗透率的先验估计值;其次,基于路网拓扑特征得到路径和流向之间的关联矩阵,并通过不同路径捕获到的样本轨迹数量和全样流向流量的估计值计算得到路径流量的先验估计值;最后,将先验路径流量和流向流量输入到广义最小二乘框架中,通过梯度搜索算法迭代求解即可得到路径流量。基于青岛市市南区路网建立了VISSIM仿真模型,选取渗透率、抽样方式、数据上传间隔和权重系数4个因素对不同参数组合的仿真场景下的路径流量估计精度和敏感性进行了验证。结果表明:在渗透率为0.1的分层随机抽样情况下,路网路径流量的估计精度达92.8%,即使在渗透率为0.05的稀疏数据场景下,估计精度仍可保持在85%以上;同时,提出的路径流量估计模型对渗透率和抽样方式较敏感;在数据上传间隔不大于15 s或路径流量误差项权重占主导的情况下,模型鲁棒性较好。
Path flow plays an important role in elaborative traffic control and management because it helps identify critical network elements,such as critical links,critical corridors,and critical paths.To overcome the drawbacks in terms of prerequisite conditions and applicability in existing path flow estimation methods,a path flow estimation method aimed at the signalized network was proposed in this study using sampled trajectory data as the sole data input,while taking the actual application condition of connected vehicle trajectory data into account.The proposed method is an extension of the classic generalized least square(GLS)model for origin-destination(OD)estimation,targeting at the minimization of the weighted error of both,the path flow and the movement flow.This way,the estimation of the path flow is realized under the generalized least square frameworks.First,based on a singular value thresholding(SVT)method for cycle-based volume estimation,the priori movement flow and the priori penetration rates of different movements were obtained using the arrival sampled trajectories in each movement.Then,using the path-movement incidence matrix and the sample path flow,the prior path flow were calculated using the prior estimated movement flow.Finally,the prior path flow and movement flow were input into the GLS model and the path flow was solved using a gradient search algorithm.The evaluation of the proposed method was done through a simulation case which was established through VISSIM,setting the roadway network in Shin an District,Qingdao City as the background.Four factors,the penetration rate,sampling method,uploading frequency of trajectory data,and weight of the objective terms were selected to design different scenarios for evaluation and sensitivity analysis of the proposed model.The results showed that,when the penetration rate is 0.1,the overall path flow estimation accuracy is 92.8%,and the accuracy reaches over 85%even when the penetration rate is 0.05.Moreover,the estimation accuracy is sensitive to the penetration rate and sampling method,while good robustness is shown in the cases where the uploading frequency is not larger than 15 s;or when the weight of path flow estimation error predominated.This shows great potential in wide applications of the proposed method in the future.
作者
姚佳蓉
曹喻旻
唐克双
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
2022年第3期226-239,共14页
China Journal of Highway and Transport
基金
上海市级科技重大专项项目(2021SHZDZX0100).
关键词
交通工程
路径流量估计
广义最小二乘法
抽样车辆轨迹数据
信号控制路网
路径-流向关联
奇异值阈值算法
traffic engineering
path flow estimation
generalized least square
sampled vehicle trajectory data
signalized roadway network
path-movement incidence
singular value thresholding algorithm
