Estimation Method Of Freeway Vehicles Location In Connected Autonomous Vehicle And Highway System

The transportation demands have shown multiple levels,high quality and individuality with the rapid incensement of motor vehicles number in last decades.However,the service level of the highways,which is an important bridge for passengers and freights between cities,has gradually descended in the context of rapid traffic growth.Expanding the high-speed kilometers has been proved ineffectively to improve LOS(the level of service),hence,active traffic management is an essential means to alleviate traffic problems.In addition,with the development of new information technology,the traditional transportation system is gradually becoming intelligent and connected.Connected and automated vehicle system is considered to be the ultimate form of intelligent transportation system.To seize market opportunities,many Internet companies and car companies have invested large amount money in the development of intelligent networked cars.It is conceivable that for a long time in the future,there will be a situation in which connected and autonomous vehicles(CAVs)are mixed with human vehicles.This paper studies the position estimation method for human vehicles in expressway with mixed traffic environment.Firstly,the development and composition of the connected and automated vehicle system are expounded based on existing research.Connected and automated vehicle system is composed by three-dimensional development system,including vehicle automation,network interconnection and system integration.The research scenarios and premise assumptions of this paper are proposed based on a detailed analysis of connected and automated vehicle system,which focus on the data sensing and transmission process.Secondly,NGSIM data,the data sources used in this study,were analyzed and processed.Considering the characteristics of NGSIM data,a trajectory data preprocessing method is proposed.In order to facilitate the modeling,the car-fed pair is extracted according to the field information.Thirdly,based on the three car-following models(OVM,FVD and IDM),the preceding vehicle position estimation method and the follow-up vehicle position estimation method are proposed respectively.In order to realize the method based on the car-following model,the BP neural network is used to estimate speed of the front car and the artificial bee colony is used to calibrate parameters of car-following model.In addition,from the perspective of data driving,two vehicle location estimation methods based on deep learning,including LSTM-RNN and GRU-RNN,are constructed.Finally,the vehicle position estimation method based on three car-following models and two RNNs are verified by the car-following data extracted by NGSIM dataset.The results show that the position estimation method based on LSTM-RNN and GRU-RNN has higher accuracy than the method based on the car-following model.