基于城市主干路交通流数据的跟驰模型标定

根据某种机理建立的交通流模型需要通过模型标定和验证后才能具体应用到实际中.通过采用视频处理技术,对陕西省西安市二环主干路和浙江省舟山市昌洲大道上上下高峰时期内的车辆微观运动录像进行技术处理,提取得到了包括位移、速度和加速度的车辆微观运动轨迹数据.根据这些交通流数据,采用Levnberg-Marquardt算法,分别对跟驰理论中2个典型的跟驰模型,即惯性模型中的敏感系数、安全时间间隔、最小安全车间距、允许速度和智能驾驶人模型中的理想速度、安全时间间隔、静止安全距离、启步加速度和舒适加速度进行了标定和验证.针对惯性模型,当允许速度大于实际速度时,位移均方差和速度均方差的平均值分别为2.8m和0.58 m/s,当允许速度小于实际速度时,位移均方差和速度均方差的平均值分别为2.22m和0.49 m/s;针对智能驾驶人模型,利用早、中、晚3组数据进行标定,得到的位移均方差和速度均方差的平均值分别为0.12m和0.10 m/s,0.07m和0.10 m/s,0.75m和0.27 m/s.因此,惯性模型与智能驾驶人模型都可用于描述城市主干路近饱和状态(即跟随车辆的最大速度远小于允许速度的行驶状态)下的车辆跟驰行为,而且当智能驾驶人模型中的加速度指数取较大的值时,它较前者更为适合.

Calibrating Car-Following Models Based on Urban Arterial Video Data

A traffic model, which is established based on certain mechanism, can be used only after calibration and validation. In this paper, the peak-hour data from vehicle micro-motion camera on the Second Ring Road in Xi＇an, Shaanxi Province and the Changzhou Road in Zhoushan, Zhejiang Province are processed by using video processing tech- nique. The traffic parameters, including displacement, speed and acceleration, are extracted from the videos. Based on the data, by using Levnberg-Marquardt algorithm, two typical car-following models, the Inertia Model and the Intelligent Driver Model, are calibrated and validated. Sensitivity, safety time gap, minimum safety distance, allowable speed in the Inertia Model and ideal speed, safety time gap, safety distance, starting acceleration and comfortable acceleration in the Intelligent Driver Model are all determined. In the Inertia Model, the average mean square error of displacement is 2.8 m and that of speed is 0.58 m/s when allowable speed is higher than actual speed, and these two values are equal to 2.22m and 0.49m/s respectively when allowable speed is lower than actual speed. For the Intelligent Model, the average mean square errors of displacement and speed are 0.12m and 0.10m/s, 0.07m and 0.10m/s, and 0.75m and 0.27m/s respec- tively in the morning, afternoon and evening. Hence, the Inertia Model and the Intelligent Driver Model are available to describe the car-following behavior on urban main roads; however, in comparison with the Inertia Model, the Intelligent Driver Model is better when the index of acceleration is high.