融合卡尔曼滤波的高速公路状态估计误差界限分析

Analyzing Error Bounds of Highway Traffic State Estimation Via Kalman Filter Fusion

为分析高速公路交通流检测数据质量,本文构建平方流量误差界(Squared FlowError Bound,SFEB)和扩展卡尔曼滤波(Extended KalmanFilter,EKF)的决策级融合模型SFEB-EKF,在检测器空间覆盖不足情况下,计算检测路段和无检测器路段的交通状态估计误差界限。与SFEB算法相比,融合模型利用EKF交通状态估计模型估计全路段交通状态,基于得到的估计样本计算全路段交通状态估计误差下界。同时,采用最近邻法(Nearest NeighborMethod,NNM)计算全路段交通状态估计误差上界。应用开源高速公路数据集测试模型,结果表明,与需要输入真实样本的SFEB算法相比,融合模型SFEB-EKF在缺少真实样本情况下,能取得相似的结果且误差保持在5%以内,不同检测器覆盖率实验下模型表现出良好的稳定性。本文模型通过给出无检测器路段交通状态估计界限,为高速公路交通检测器布设方案提供参考。

This paper analyzes the data quality of highway traffic flow and proposes a decision-level fusion model based on Squared Flow Error Bound(SFEB)and Extended Kalman Filter(EKF),namely SFEB-EKF.It is used to calculate the error bounds of traffic state estimation for the road sections with and without detectors when the detection spatial coverage is insufficient.Compared with the SFEB algorithm,the fusion model uses the EKF traffic state estimation model to estimate traffic states of the whole highway section.The lower bound of traffic state estimation error is calculated based on the obtained estimated samples.At the same time,the Nearest Neighbor Method(NNM)is used to calculate the upper bound of traffic state estimation error of the whole road sections.The model is tested with an open-source highway dataset,and results show that,compared with the SFEB algorithm that needs to input real samples,the fusion model SFEB-EKF can achieve similar results in the absence of real samples and the gap is kept within 5%.The model shows good stability under different detection coverage experiments.The research results can give the estimation boundary of the traffic state of the road section without detectors,and provide a reference for the layout plan of highway traffic detectors.