Research On VANET Based On The Traffic Flow Forecasting

With the continuous development of the automotive industry,Intelligent Traffic System(ITS)plays an important role in automotive-related industries and technologies.As the basis of the Intelligent Traffic System,the prediction of traffic flow is a hot topic in the research area among them.Formed on the basis of mobile Ad-Hoc Networks(MANET)for vehicle-to-vehicle communication,VANET has a promising development.The routing protocol,which forms the basis of VANET,is an indispensable technology in VANET and plays a decisive role in the performance of the network.Therefore,designing a routing protocol can transmit information efficiently and timely,which is vital for ITS.At first,this paper studies short-term traffic flow forecasting.In order to further improve forecasting accuracy and reduce forecasting time,the GRU model in the deep learning RNN method is used for short-term predictions of traffic flow.Besides,a method with autocorrelation analysis is proposed to select the input vector to reduce the error of multi-step prediction.What's more,it proposes a combination of blockchain technology for self-propelled nodes and other issues to inspire the vehicle nodes in the network to perform forwarding effectively and timely.Finally,this paper analyzes the GyTAR routing protocol in VANET and EP-GyTAR is proposed in conjunction with traffic flow prediction.The specific research content is as follows:Firstly,this paper studies and analyzes the availability of short-term traffic flow forecasting models,and points out some existing shortcomings in it.Then the deep learning RNN method is further studied and analyzed.Several typical models are studied.According to the characteristics of traffic flow changing,the GRU method in the LSTM model is selected to forecasting the traffic flow.According to the conventional multi-step prediction accuracy of the iteration is not high,this paper proposed autocorrelation analysis method to select the input vector so that it can reduce multi-step prediction error and improve multi-step prediction accuracy.The real traffic flow data is used as a data set extracted from the PEMS database system,and the improved method is verified by MATLAB software simulation.Compared with theconventional RBF and other machine learning prediction methods,both the single-step prediction and multi-step prediction accuracy are higher than them.Secondly,a combination of VANET and blockchain technologies is proposed to build a community and a six-tier architecture is proposed.It can theoretically encourageS the vehicle nodes in the network to perform effectively and timely forwarding,and encourages vehicles with network communication equipment to join in this community.In the community ecology,the communication performance of the entire VANET network is further improved.Thirdly,combined with the above prediction method and blockchain technology,the GyTAR routing protocol is improved from the aspect of route forwarding,and an EP-GyTAR(Enhanced Prediction-GyTAR)protocol based on traffic flow prediction is proposed and combined with six different methods.Communication methods to deal with the existing complex urban environment,the new and improved routing protocol can solve the problem of excessive end-to-end delay caused by changes in traffic flow caused by delays between roads.Finally,this paper uses SUMO combined with NS3 co-simulation platform to simulate the improved EP-GyTAR routing protocol with GyTAR,E-GyTAR,S-GyTAR and GPSR in two different conditions.The three performance indicators of quantity(end-to-end average delay,the number of gap hop count)reflect the advantages of EP-GyTAR.