| Intersection is an important part of urban traffic network,and areas of high incidence of traffic accidents too.The mixture of pedestrians,vehicles brings drivers a great deal of driving difficulty in intersection.In order to improve safety of vehicles and reduce difficulty of driving for driver in intersection,In this paper,time series models and neural network model are used to predict speed values and steering intention respectively of vehicles into intersection.Experiments data verity that the model prediction effect meet the requirements.This paper describes the contents are as follows:(1)ARMA model is used to forecast the vehicle speed values in the intersection.The vehicles’ speed timing series at straight road of the intersection is regarded as pilot timing series and treated with method of stabilization.And then check its autocorrelation and partial autocorrelation function,set of model order,estimation of model parameter.Finally the ARMA model is built to forecast the vehicle speed values in intersection.(2)EMD is used to extract the intersections fluctuation characteristic of vehicle speed at straight road of intersection area.After analyzing the vehicle driving characteristics at intersection,the paper concludes: the preparation that the driver make for steering when entering into intersection will be reacted to fluctuations of vehicle speed at straight road of intersection area.Based on this conclusion,this paper obtained the intrinsic mode signal curve of vehicle speed at straight road with the EMD.And then get its number of extreme point and variance as the characteristics of fluctuations of vehicle speed as the inputs for establishing a BP neural network model below.(3)Forecast the driver steering intention when entering into of intersection with BP neural network model.At first,determine the structure of the BP neural network,activation function and training algorithm based on vehicle speed fluctuation characteristic data above.Then use the training data to train BP neural network model.Finally,use test data to test the prediction effect of BP neural network model. |
