考虑交通信号和交通密度的城市路段个体行程时间建模

Individual Travel Time Modelling of Urban Road Section Considering Traffic Signal and Density

为探究城市路段中交通密度与交通信号对个体车辆行程时间的影响规律,提高城市道路行程时间的预测精度,基于车牌识别数据与信号机配时数据等多种融合数据,提出了由自由流行程时间、密度延误与随机误差项组成的行程时间关系函数,并在该函数基础上构建了考虑个体驾驶习惯的行程时间预测模型。首先,基于道路限速条件、车辆进入路段时的下游交叉口信号状态构建自由流行程时间函数;其次,提出了一种通过密度阈值判断车辆能否在当前信号周期通过下游交叉口的方法,并基于个体车辆进入路段时的路段密度计算出车辆所需等待信号周期个数及等待前方排队疏散所需时间,从而构建密度延误函数;然后,将个体实际行程时间与自由流行程时间、密度延误的差值作为随机误差项,通过高斯混合模型拟合随机误差项的概率分布;最后,选用安徽省宣城市多个路段作为案例,分析各路段行程时间函数各部分的具体表现,对该行程时间预测模型进行验证。结果表明:行程时间预测模型的MAPE,MAE,RMSE分别为6%~12%,5~15,13~35,在准确率方面优于其他算法,是一种有效的城市路段个体行程时间预测方法。

To explore the influence of traffic density and signal on individual vehicle travel time in urban road sections,and to improve the prediction accuracy of urban road travel time,the travel time relation function composed of free-flow travel time,density delay,and random error was proposed based on a variety of fused data,including license plate recognition data and traffic signal timing data.Furthermore,the travel time prediction model considering individual driving habits was developed based on this function.Initially,the free-flow travel time function was constructed based on road speed limits and the signal status of downstream intersections when vehicles entering the road section.Subsequently,the method was proposed to determine whether vehicles could pass the downstream intersection within the current signal cycle based on a density threshold.The needed number of signal cycles and the required time for queuing evacuation ahead were calculated based on the road section density when individual vehicles entering the road section,thus constructing the density delay function.Next,the difference between the actual travel time of an individual and the sum of free-flow travel time and density delay was used as the random error term.The probability distribution of random error term was fitted by using the Gaussian mixture model.Finally,several road sections in Xuancheng City,Anhui Province were selected as cases to analyze the specific performance of each part of the travel time function,and to validate the travel time prediction model.The result indicates that the proposed travel time prediction model has MAPE,MAE,and RMSE of 6%-12%,5-15,and 13-35 respectively,outperforming other algorithms in terms of accuracy,and proving to be an effective method for predicting individual travel time on urban roads.