基于数据融合的路段行程时间估计

Link Travel Time Estimation Based on Data Fusion

基于检测器数据的路段行程时间估计通常具有精度不高和可靠性差的特点。论文引入了自适应式卡尔曼滤波,采用K近邻法寻找相似的交通流状态来标定状态转移系数,建立了基于固定型检测器数据和移动型检测器数据的路段行程时间估计融合模型。实际数据的验证结果是,平均相对误差为9.52%,相对误差的标准差为8.92%。研究表明,与基于移动检测器数据的估计方法相比较,该方法极大地改善了估计精度和可靠性,还具有收敛速度快、对初值不敏感、参数少等特点。

The estimated link travel time based on detector data have two features： low accuracy and poor reliability. The adaptive Kalman filter and K-nearest Neighbor （K-NN） are introduced to estimate link travel time based on fixed detector data and mobile detector data in this paper. The results show that the average relative error is 9.52% and the standard deviation of relative error is 8.92%. The accuracy and reliability of the estimation are largely improved when compared with those based on mobile detector data. In addition, the method converges rapidly, involves fewer parameters, and is also insensitive to initial values.