基于NGSIM数据的汽车变道越线时间预测

Prediction of vehicle lane change crossing time based on NGSIM data

为预测汽车变道越线时间,基于NGSIM数据提取汽车的真实越线时间,选定变道车和其周围汽车的行驶状态参数作为潜在影响因素。随机选取60组训练数据,先利用Spss软件逐步回归法对越线时间与自变量进行多元线性回归拟合,再通过Matlab编程对越线时间进行多元非线性回归拟合,分别建立多元线性和非线性回归越线时间预测模型;对模型进行训练后得到各模型的系数,分别测试除训练数据外的30组数据并对比拟合效果。结果表明,所建立的多元非线性回归模型能够有效预测汽车变道越线时间,预测有效率达到83%。

In order to predict the lane change crossing time of the vehicle,the real lane crossing time of the vehicle was extracted from NGSIM data,and the driving state parameters of the lane change vehicle and its surrounding vehicles were selected as the potential influencing factors.Randomly selected 60 set of training data are used to establish multiple linear regression fitting crossing time with the independent variables in Spss software,and through Matlab programming establish multivariate nonlinear regression fitting,respectively the coefficient of each model is obtained after training,which is used to test additional 30 data sets,the established multivariate nonlinear regression model can effectively predict the car crossing time,with a lane change prediction efficiency of 83%.