Research On Road Network Traffic Situation Prediction Method In Connected Vehicle Environment

The increasingly serious traffic congestion has had a variety of adverse effects on the social and economic developments of Chinese cities.How to relieve traffic congestion in big cities has become a problem that traffic managers must face.With the continuous development of connected vehicle technologies,traffic managers try to use connected vehicle technologies for the collection and sharing of road network traffic information to provide travelers with timely and accurate traffic situation information so that travelers can select more reasonable travel routes and ease the traffic congestion of the entire road network.Therefore,the research on the forecasting method of traffic situation of road network under the environment of connected vehicle is of great significance to improving the traffic efficiency of urban road network and improving traffic congestion.Traffic situation of road network is taken as the research object in this thesis,prediction method for traffic situation of road network is studied with considering the characteristics of connected vehicle environment and validity of the prediction method based on the built connected vehicle environment is verified.First,deficiencies of existing researches are summarized and the research contents and technical route are proposed based on the review of characterization and prediction methods for traffic situation in traditional environment and connected vehicle environment.Then,the evaluation indicators of the characterization model of traffic situation in this thesis are determined after analyzing characterization models of traffic situation in China and foreign countries.On this basis,a characterization model for traffic situation based on fuzzy comprehensive evaluation is proposed.Then,three kinds of typical prediction methods of traffic situation are introduced,and the impact factors of traffic situation under the connected vehicle environment are analyzed.A prediction method for traffic situation in connected vehicle environment is proposed based on the Long Short-Term Memory(LSTM)network.Model structure,activation function,optimizer,loss function and the format of input data of the network are determined combining with the working principle of LSTM network.The designed LSTM network model is used to predict the average speed,average travel time ratio and density of the road sections respectively.Traffic situation of road network is predicted using characterization model of traffic situation based on fuzzy comprehensive evaluation.According to the characteristics of the connected vehicle environment,simulation scenario of the corresponding road network is established based on the actual road network.Secondary development of the communication modules of vehicle unit and the roadside unit and behavior modules are carried out to make intersections in road network has the capability of cooperative speed guidance.Finally,case verification is conducted for the proposed prediction method of traffic situation of road network using the established simulation scenario of connected vehicle environment.Effectiveness of the characterization model of traffic situation based on fuzzy comprehensive evaluation was analyzed based on the thermodynamic map.At the same time,prediction accuracy of the prediction method,impacts of model parameters and factors for connected vehicle environment on the prediction accuracy of the prediction method are analyzedThe case verification results show that the characterization model of traffic situation based on fuzzy comprehensive evaluation can be used to describe traffic situation of road network.For the prediction method of traffic situation of road network,the average prediction accuracy of the average speed,travel time ratio and density of the road sections are 90.3%,82.8%and 90.5%,respectively,and the prediction accuracy of traffic situation of road network is 87.0%.For the factors affecting the prediction accuracy,the prediction accuracy of the prediction method improves slowly after the number of iterations of the LSTM network reaches a certain value.The more sample sizes of the training data are,the better the prediction accuracy of the prediction method is improved.Characteristics of connected vehicle environment also impact the prediction accuracy of the prediction method.