基于卡尔曼滤波的交叉口排队长度实时估计模型

为解决现有排队长度估计方法不能对排队长度进行实时秒级估计的问题，本文采用车联网实时数据，构建基于卡尔曼滤波的实时排队长度估计模型。首先，以当前时刻加入和离开排队队列的车辆数为输入变量构建状态转移方程，以当前排队网联车的数量和渗透率构建观测方程；其次，采用回归模型估计状态转移方程和观测方程的噪声协方差矩阵；然后，提出基于卡尔曼滤波方法估计排队长度的流程算法和模型性能评价指标；最后，基于实际数据构建仿真环境验证模型的有效性。结果表明：当网联车渗透率为30％时，平均绝对误差(MAE)，平均绝对百分比误差 (MAPE)和均方根误差(RMSE)的平均值分别为1.6辆，20.9％和2.5辆；当渗透率大于20%时，与基准方法相比，本文模型估计效果更优。

Real-time Intersection Queue Length Estimation Based on Kalman Filtering

Considering the existing queue length estimation methods cannot dynamically reflect the queue length at intersections, this paper proposes a Kalman filtering- based queue length estimation model using real-time connected vehicles data. The stepwise state transition equation is developed based on the number of vehicles joining and leaving the queue at the current moment. The observation equation is formulated through the current number of queuing connected vehicles and the penetration rate. Then, a regression model is used to estimate noise covariance matrixes of the state transition equation and the observation equation. The process of the queue estimation is established and the evaluation index is proposed to measure the effectiveness of the proposed model. A simulation evaluation is then performed based on actual data. The results show that when the penetration rate of connected vehicle is 30% , the average values of mean absolute errors (MAE) is 1.6 vehicles, the mean absolute percentage errors (MAPE) is 20.9%, and root mean square errors (RMSE) is 2.5 vehicles. When the penetration rate of connected vehicle is over 20%, the proposed method shows better performances than the benchmark method.