信号交叉口跟驰模型及其参数标定研究

Car-following Model and Parameter Calibration for Vehicles Passing A Signalized Intersection

为了更好地运用跟驰模型描述车辆在信号交叉口的通行过程,提出一种考虑信号灯变化对车流瞬时冲击的交叉口全速度差和加速度(full velocity difference and acceleration at intersection,FVDA-I)模型,并根据车辆轨迹数据标定了跟驰模型参数。首先,构建位置-时间临界轨迹来判断车辆能否在绿灯时间内通过交叉口,并提出转弯车辆优化速度。然后,使用S-G滤波器对车辆位置进行平滑拟合,结合期望最大化(expectation maximum,EM)算法标定模型参数。最后,运用FVDA-I模型估计车辆在交叉口的通行时间以验证模型的有效性。仿真结果表明:相较于改进全速度差(full velocity difference,FVD)模型,FVDA-I模型具有头车启动舒适缓慢、跟随车辆启动快、车队总通行时间短的特点,通行效率提高了11.3%,车辆的通行时间与实际采样轨迹的估计时间对比误差小于10%。

In order to better understand the car-following behaviors of a signalized intersection.A model of full velocity difference and acceleration at intersection(FVDA-I)is presented,together with some methods to calibrate parameters of that model using available data.In the beginning,for determining whether a vehicle can pass an intersection within remaining green time,a method is proposed for constructing position-time critical trajectory,together with a way for computing an optimal speed of a turning vehicle.For model calibration,the S-G filter is used to extract parameters of a driver’s behaviors,and an expectation-maximization algorithm is used to calibrate the rest parts.Finally,to test its effectiveness,the FVDA-I model is used to estimate the average traveling time of vehicles passing an intersection.The simulations show that:(1)Comparing with the improved full velocity difference model,the first vehicle of the FVDA-I model starts slower,the following vehicles start faster,and the total travel time of a whole platoon passing an intersection is reduced by 11.3%in average.(2)The error of the vehicle travel time computed by using the FVDA-I model against the actual sampling trajectory is less than 10%.