基于改进K-Means算法的交叉口影响路段行程速度估计

Estimation of Travel Speed on Intersection Influenced Link Based on Improved K-Means Algorithm

基于低频、低覆盖率、数据来源多样的GPS浮动车数据,在现有数据预处理方法的基础上,以交叉口影响路段数据点为研究对象,研究出更合理且准确获得交通参数的技术方案。GPS浮动车数据由于其具有全天候、多覆盖等特性,能够实时监测交通参数,估计交通状态。为克服数据本身缺陷,使数据能有效利用,精确得到交通参数,本研究获取短时内路段所有数据点代表整体状态。首先基于数据的特性和在路段分布的节律,利用曲线拟合及拉格朗日中值定理确定交叉口的影响范围;其次在该范围内利用改进K-Means聚类方法,确定初始聚类中心,并以有效性指数作为优化目标确定聚类数;在此基础上分配权重,结合交叉口影响范围外的数据点,对整个交叉口影响路段的行程速度进行估计。用杭州市局部路网中GPS数据进行案例分析,验证技术方案。通过实地调查获取实验真实值,分别讨论了在主、次干路路段本方案估计差异,并与传统模型进行了对比分析。分析表明,该方法得到的路段行程速度估计值与真实值较为接近,误差较小,在城市主干路和次干路中的误差分别为4.1%和9.5%,比传统模型误差更小更稳定,能较好地满足城市智能交通控制系统对于交通参数的精度要求。

Based on GPS floating car data （FCD） which have low frequency, low coverage, and come from different vehicles, and based on existing data pretreatment, regarding the data of intersection influenced link as the research object, a more reasonable and accurate traffic parameters achieving technical scheme is studied. FCD data can monitor the traffic parameters and estimate the real-time traffic status because of its all-weather, multi coverage characteristics. In order to overcome the shortcomings of the data themselves, make the data effectively useful, and get the traffic parameters accurately, all the link data points in short term are obtained to represent the overall state. First, based on the characteristics and distribution of the data in the link, the influencing range of the intersection is determined using curve fitting and Lagrange＇s mean value theorem. Then, the initial cluster center in this influencing range is determined using an improved K-Means clustering algorithm, in which the optimal number of clusters is selected by a clustering validity index. Afterward, the weights of cluster centers are distributed, combining with the data out of the influencing range, the travel speed of the whole intersection influenced link is estimated. To verify the technical scheme, the case study is conducted choosing the GPS FCD of a local road network in Hangzhou. The true values are obtained through on-the-spot investigation, the estimation errors of arterial and minor arterial by the proposed scheme are discussed, and compared with the estimation of traditional model. The analytical result shows that the estimated link travel speed by the proposed method is close to the true value, the errors are 4. 1% on urban trunk road and 9. 5% on secondary trunk road, which are smaller and more stable than the traditional model. This algorithm can meet the precision requirement of traffic parameters for urban intelligent transport control system.