交通网络中动态起点-迄点矩阵估计的一种实用算法

Novel Algorithm for Estimating Dynamic Origin-Destination Matrix in Traffic Networks

基于广义最小二乘模型建立了一种带滑动窗的动态起点 迄点(OD)矩阵估计算法,可通过对路段交通量和行程时间的检测来估计时变的OD数据.对任意估计时段的OD流,通过假定各车辆间的时头距均匀分布且可按相同比例拉伸或压缩,得出模型中关键的分配矩阵的解析算式.该算法是一种递推的估计过程,仅需较少的先验信息而估计过程不会发散.滑动窗的引入可充分利用量测信息,抑制量测噪声.大量仿真实验表明,所提出的方法在估计精度上明显优于Cascetta的递推估计法,但计算量并无显著增加.

Based on generalized least square (GLS) model, a dynamic origindestination (OD) matrix estimation algorithm with sliding window is established. The dynamic OD matrix can be estimated from the surveillance of traffic counts and traveling time on links in a traffic network. On the assumption that headtimes between vehicles in arbitrary OD flow for each interval are uniformly distributed and can be 'stretched' or 'squeezed' with the same proportion as they traverse the network, an analytical formula to calculate the key assignment matrix is obtained. The algorithm is a recursive procedure with few a priori data, and there exists no divergence in the estimation. And more surveillance data can be utilized effectively, and measurement noises can be restrained efficiently by introducing a sliding window. A large number of simulation tests show that the estimation accuracy of the proposed method is much higher than that of Cascetta's recursive algorithm, and there is only a little increase in computation cost.