检测数据缺失条件下的交通流估计方法研究

A Method of Traffic Flow Estimation for Urban Network under Condition of Traffic Detectors Failure

城市路网中部分交通量检测器失效会导致路段的交通流观测数据缺失,从而影响路段实时交通拥堵状况的辨识和分析。基于路网交通流数据的时空分布特性和道路交通流特征,研究了一种改进重构算法,对缺失的交通量数据进行估计。构建了交通量时空数据的三维张量模型;综合考虑道路交通流分布与环境特性,提出了基于Tucker重构模型的目标优化函数,进而求解得到检测器失效路段的交通量估计值;对改进重构算法、Tucker重构算法和时空插值算法在不同交通量数据缺失情况下的估计效果进行了比较。实验结果表明,在6条路段组成的实验路网中,早高峰交叉口进口道实时观测交通量随机缺失50%或其中3条道路完全缺失的情况下,估计结果的平均绝对误差(MAE)分别为13.961 4和14.276 3,均方根误差(RMSE)分别为18.764 8和18.707 0。相较于Tucker重构算法,MAE分别下降了32.29%和44.26%,RMSE分别下降了31.73%和48.57%。

The failure of traffic detectors in urban road networks would result in traffic flow data missing,thus affecting identification and analysis of real time congestions at road sections.Based on spatial temporal distribution characteristics of traffic flow data and features of road traffic flow,an improved algorithm for reconstruction is studied to estimate the missing data.A three dimensional tensor model of spatial temporal data of traffic volume is developed.Considering the distribution of road traffic flow and environmental characteristics,a target optimization function based on Tucker reconstruction model is proposed to solve the estimation value of detector failure sections.In a case study of different missing volume of traffic data,estimation effects in the improved algorithm,Tucker reconstruction algorithm,and spatial temporal interpolation algorithm are compared.The results show that in the case study of 6 road sections,data collected at one entrance of an intersection during morning peak,when 50%of the real time data is randomly missing,or when the data of three sections is completely missing,the mean absolute errors(MAE)of the estimated results is 13.961 4 and 14.276 3,respectively.The root mean square error(RMSE)is 18.764 8 and 18.707 0,respectively.Compared with Tucker reconstruction algorithm,MAE decreases by 32.29%and 44.26%,respectively;RMSE decreases by 31.73%and 48.57%,respectively.