车辆跟驰模型参数标定中的性能指标选择

Selection of Measures of Performance on Calibrating Parameters in Car Following Models

为揭示性能指标选择对跟驰模型参数标定结果产生的影响,基于美国I-80公路实测数据,针对Gipps,IDM和Newell三个主流跟驰模型,通过选取不同的性能指标参数,以绝对误差RMSE和相对误差U系数作为目标函数分别对上述3个跟驰模型的参数进行标定,并对标定过程中误差产生的原因进行定性及定量的分析。结果表明:在选择性能指标标定跟驰模型参数效果方面,车头间距优于跟驰车速度;此外,在优化求解过程中发现,单个观测间隔内跟驰车行驶过的位移是最基本的物理量;相对于跟驰车速度,选择车头间距作为性能指标参数虽然在一定程度上降低了单个间隔内跟驰车行驶位移的拟合度,但却保证了实测数据从整体上更准确地表征车辆的跟驰过程;因此,车头间距时间序列能够更好地代表跟驰车的跟驰运动行为,以车头间距作为性能指标对跟驰模型参数进行标定能够提高模型的可靠性和准确性。

To reveal the influence of selection of measures of performance（MOP）on the calibration of car following models,different performance indices were chosen and absolute error RMSE and relative error U coefficient were used as the objective functions to calibrate the parameters of 3 main car following models including Gipps,IDM and Newell models based on measured US I-80 highway traffic data.Then the reason of error formation was analyzed during calibration in both qualitative and quantitative methods.The results show that space headway performs better than velocity on the effect of calibrating car following models.Moreover,in the process of optimization,the displacement that follower travels in one interval is the basic variable;compared with velocity,space headway as MOP decreases the goodness-of-fit of the spatial displacement of follower that travels in one interval,but it is good at representing exact process of car following by real traffic data;therefore,the space headway has a better description capability to car following motion in general,and taking space headway as a performance index to calibrate the parameters of model is capable of improving the reliability and accuracy of car following models.