基于多用户驾驶模拟平台的雾天高速公路跟驰模型参数标定及验证

高速公路连环追尾事故多发生在雾天环境下,且容易造成严重的事故人员伤亡。当前的跟驰及追尾风险研究多集中于两车跟驰,缺乏对雾天情况下车队跟驰的研究。利用雾天环境下车队跟驰轨迹数据对传统主流跟驰模型进行标定验证,基于多用户驾驶模拟平台设计了8个不同雾天等级和限速组合的高速公路虚拟场景,开展驾驶模拟试验并采集数据。试验招募了8名男性驾驶人并通过随机调整他们在车队中的位置顺序来获得足够的车队跟驰轨迹数据,根据判定标准筛选合适的车队跟驰轨迹数据,按照2:1的原则分配标定和验证阶段的数据组。选取Newell、Gipps和IDM三个主流跟车模型进行参数标定和验证,以时间序列的车头间距和相对均方根误差(RMSPE)分别作为性能指标参数和拟合优度函数,使用遗传算法搜寻目标函数最小值以标定跟驰模型参数,并用车辆轨迹完整性(CVT)和RMSPE评价验证阶段的仿真结果。结果表明:在标定阶段,Newell、Gipps和IDM三个模型的RMSPE整体平均值分别为30.1%、18.6%和27.7%,各个试验条件下Gipps模型的RMSPE值均小于另外2个模型,说明Gipps模型能更好地拟合试验数据;在验证阶段,Gipps模型的RMSPE整体平均值为21.2%,远小于另外2个模型,可见Gipps模型在局部精确度上的鲁棒性要优于Newell模型和IDM模型;Gipps模型的CVT整体平均值和波动幅度分别为98.1%和2.0%,均是3个模型中的最小值,说明Gipps模型在整体轨迹上的鲁棒性也优于另外2个模型。雾天环境下,Gipps模型具备更好的拟合能力和鲁棒性,因此推荐仿真软件使用Gipps模型模拟雾天环境下车队跟驰行为,不同雾天等级及限速下的Gipps模型参数可参考该研究标定的参数。

Parameter Calibration and Validation for Car-following Models on Freeway Under Foggy Conditions Based on Multi-user Driving Simulator System

Multi-vehicle rear-end collisions occur frequently on highways,especially in foggy weather,and can lead to severe casualties.Recent car-following and rear-end risk research has mostly focused on two-vehicle car-following and largely ignored vehicle fleet car-following in foggy conditions.This study calibrated and validated several traditional car-following models using vehicle fleet dynamics in foggy weather.A multi-user driving simulator system was used to conduct the experiment,and data were collected from eight highway scenarios with different fog densities and speed limits.Eight male drivers were recruited in the experiment,and their positions in the fleet were randomly assigned to obtain sufficient car-following trajectory data.Several criteria were applied to screen appropriate car-following trajectory data,and the screened data were used for calibration and validation,respectively,in a ratio of 2:1.Three traditional car-following models,the Newell,Gipps,and IDM models,were selected for parameter calibration and validation.The time series of distance headway and the root mean square percentage error (RMSPE) were used as the measures of performance (MOP) and the goodness-of-fit function (GOF),respectively.A genetic algorithm (GA) was applied to obtain the minimum value of the objective function to calibrate the parameters of the car-following models.The simulation results in the validation stage were evaluated based on the completeness of vehicle trajectory (CVT) and RMSPE.The results show that in the calibration stage,the overall mean RMSPE values are 30.1%,18.6%,and 27.7% for the Newell,Gipps,and IDM models,respectively,and the RMSPE value of the Gipps model is smaller than those of the other two models under all experimental conditions.This means that the Gipps model can better fit the experimental data.In the validation stage,the overall mean value of RMSPE is 21.2% for the Gipps model,which is much smaller than those of the other two models.Therefore,the Gipps model is more robust than the Newell and IDM models in terms of local accuracy.The overall mean value and fluctuation range of the CVT of the Gipps model are 98.1% and 2.0%,respectively,which are the minimum values among the three models.This shows that the robustness of the Gipps model on the overall trajectory is also better than that of the other two models.The Gipps model has a better fitting capability and robustness under foggy conditions.As a result,it is recommended that simulation tools use the Gipps model to simulate vehicle fleet car-following behavior under foggy conditions.The parameters of the Gipps model under different fog levels and speed limits can refer to the parameters calibrated in this study.