基于参数预测的高速公路交通状态判别  

Identification of Freeway Traffic State Based on Parameter Forecasting

作　　者：韦清波 苏奎 何兆成[2] 聂佩林[3] 杨敬锋 

机构地区：[1]广州市公共交通数据管理中心,广州510620 [2]中山大学工学院智能交通研究中心,广州510275 [3]佛山科学技术学院,广东佛山528000

出　　处：《交通信息与安全》2012年第6期81-86,共6页Journal of Transport Information and Safety

基　　金：国家科技计划支撑项目(批准号:2011BAG02B02);广州市经贸委技术改造投资项目(批准号:11010653902000800)资助

摘　　要：针对目前基于单截面检测数据的高速公路交通状态判别算法存在着判断阈值多,对拥挤样本依赖性强而拥挤样本采集困难等问题,提出了基于交通流预测的交通状态判别模型。预测过程中以车辆的平均占用时间作为预测的目标参数,利用神经网络建立预测模型,并通过相关系数法确定神经网络的输入层。在预测的基础上,以实测值与预测值之间的差值作为判别的依据,判别道路的交通状态。应用广深高速公路实测数据对判别模型的有效性进行检验,并与经典的McMaster检测算法做了对比,结果表明,所提出算法对拥挤样本依赖较少,判别精度高,鲁棒性高。Current methods for identifying freeway traffic state with the data from a single detector require many traffic samples and quite a few thresholds to judge whether congestion takes place.A novel identification model based on traffic flow forecasting is presented in this study.In this model,the average vehicle occupancy time is used as forecasting target.Neural network is used to construct the forecasting model,whose input layer is determined through a correlation analysis.Then,traffic state is identified according to the difference between measured values and forecasted values.The presented algorithm and model are tested by the field data collected from Guangzhou-Shenzhen Freeway.Compared with the classical algorithm,McMaster,it is found out that the proposed model only requires a small congestion sample and has a high accuracy and robustness.

关 键 词：交通状态 交通流预测 神经网络 平均占用时间 

分 类 号：U491[交通运输工程—交通运输规划与管理]