Short-term Traffic Flow Prediction In Heterogeneous Wireless Networks

With the development of society,the road traffic environment is getting worse and worse,and road traffic data is showing an explosive growth trend.In this regard,this paper builds a heterogeneous wireless network based on the NB-IOT technology and LTE-V-Direct technology under the edge computing environment,and studies the two-level short-term traffic prediction model DBN-HMM based on this to alleviate road traffic congestion.In order to solve the problem that GPS speed measurement cannot work normally in specific environments such as tunnels and overpasses,an RFID road system based on the AEKF algorithm to predict the speed of high-speed vehicles.The main results of this thesis are summarized as follows:1.A heterogeneous wireless network is established in the edge computing environment,and a two-level data-driven short-term traffic flow prediction model is proposed.Firstly,the traffic characteristics between road occupancy and road flow are effectively extracted by using the DBN,and the predicted road flow is used as the input data of the HMM.Secondly,the statistical relationship between road flow and speed can be effectively established through the HMM,and road speed prediction for each road section is realized.Then the weighted average method is used to solve the problem of wrong or abnormal data collected by the road loop detector.Finally,we proposed model can further reduce RMSE and MAPE and improve prediction accuracy compared with the SAE,LSTM,and GRU models.2.In situations such as overpasses and tunnels,where GPS speed measurement cannot be used normally and effectively,an RFID data-driven vehicle speed prediction based on an AEKF model is proposed.Firstly,the current vehicle traveling in the RFID system can obtain the speed information of the preceding vehicle through the communication between the RFID tag on the road and the on-board RFID reader.Secondly,AEKF model is designed by combining adaptive forgetting factor and EKF algorithm to improve the accuracy of vehicle speed prediction.The driving environment was then simulated using three different vehicle speed models.Two evaluation indicators,MSE and MAE,are introduced to better evaluate vehicle speed prediction errors.Finally,we analyzed the impact of ? on the AEKF model and further reduced the prediction error compared with the EKF model.