路段行程时间估计的浮动车数据挖掘方法  

Mining method of floating car data based on link travel time estimation

作　　者：李慧兵[1] 杨晓光[2] 罗莉华[1] 

机构地区：[1]上海海事大学交通运输学院,上海201306 [2]同济大学交通运输工程学院,上海201804

出　　处：《交通运输工程学报》2014年第6期100-109,116,共11页Journal of Traffic and Transportation Engineering

基　　金：国家自然科学基金项目(61304203);上海市科研计划项目(12ZR1444800)

摘　　要：基于浮动车数据,提出一种信号配时信息缺失下的路段行程时间估计方法,由交叉口范围动态划分、路段影响范围划分、浮动车数据提取与路段行程时间估计4个模块组成,每个模块的实现均需借助于前一模块的输出。根据交叉口信号控制下的车辆行驶状态,在交叉口范围动态划分与路段影响范围划分2个模块中,利用密度法将单元路段划分为不同区域。根据路段行程时间估计原理,利用浮动车数据提取模块过滤掉受信号控制影响较大的浮动车数据,提取路段行程时间估计的目标数据。利用路段行程时间估计模块挖掘历史浮动车数据,根据浮动车目标数据点存在区域的不同,将浮动车数据分为3类,并对不同类型数据采取相应的断面通过时刻估计方法,建立基于不同数据条件下的行程时间估计模型。利用VISSIM软件对路段行程时间估计方法进行仿真验证,并与直接法和间接法进行对比分析。分析结果表明:对于粗粒度浮动车数据,路段行程时间估计方法的平均绝对误差和平均相对误差分别为12s和8.67%,优于传统的直接法与间接法。Based on floating car data,a link travel time estimation method without signal timing data was proposed.The method consisted of four modules,which were intersection boundary dynamic partition module,link influence range partition module,floating car data extraction module,and link travel time estimation module,and the implementation of each module relied greatly on the output of previous one.According to vehicle travel state under the influence of signal control, link unit was divided into different segments by using density method in intersection boundary dynamic partition module and link influence range partition module. According to link travel time estimation mechanism,floating car data that were seriously affected by signal control were filtered off in floating car data extraction module,so the target floating car data could be obtained.Historical floating car data were excavated in link travel time estimation module,and floating car data were divided into 3 types according to different exsited regions of target data.Corresponding section travel time estimation methods were used for different types of data,and corresponding section travel time estimation models were established.Link travel time estimation method was simulated and verified by using software VISSIM,and its result was compared with the results of direct and indirect methods.Analysis result indicates that for coarse-grained floating car data,the average absolute error and average relative error of link travel time estimation method are 12 s and 8.67% respectively,so it performs better than traditional direct and indirect methods.3 tabs,22 figs,20 refs.

关 键 词：智能交通系统 路段行程时间估计 浮动车数据 信号配时 粗粒度数据 

分 类 号：U491.1[交通运输工程—交通运输规划与管理]