Air Traffic Flow Analysis Based On ADS-B Datagram Platform

Along with rapid development of domestic civil aviation industry,increasing demands for air transportation will inevitably lead to congestion of limited airspace,which poses new challenges to the air traffic management system.The emergence of automatic dependent surveillance-broadcast(ADS-B)technique can enable us to obtain massive civil aviation flights data by using ADS-B receivers at low cost.Research and mining of these data will also provide a new angle of view for building a new generation of air traffic management system.Furthermore,the data of air traffic flow is the key information of the future denser air traffic.This thesis focuses on the research and mining of air aircraft flow in the field of air traffic management.By establishing a ground receiving platform based on the ADS-B technology,a large amount of real civil aviation flight data is collected,and the information of air aircraft flow is studied.Firstly,this thesis utilizes an ADS-B omnidirectional antenna which is set up in Nanjing University of Posts and Telecommunications,and the ADS-B data of civil aircrafts in Nanjing and its surrounding airspace is received and collected.In addition,by accessing the ADS-B data cloud platform provided by Civil Aviation University of China,the author can finally obtain detailed practical aviation data covering east China and north China.On the basis of these data,a series of data pre-processing procedures are carried out in this thesis to abtain a structured ADS-B dataset,and various other flight-related data are fused by using web crawler technique.Compared with traditional simulation data and original ADS-B data used in other researches,this thesis can obtain a multidimensional civil aviation dataset including flight planning,landing and take-off airports information,and meteorological information,which shows more practical significance and research value.Secondly,based on the obtained civil aviation dataset,this thesis mines relevant information of aircraft flow for the evaluated airspace.Combined with different application scenarios in air traffic statistics,two methods of air traffic statistics based on the ADS-B dataset are proposed,which can be applied in airports and air routes,respectively.Theoretically,the aircraft flow statistics methods can be carried out according to different time granularity and airspace range.In this thesis,the aircraft flow statistics can be conducted according to time granularity of an hpur,and some visualization tasks are carried out to analyze the flow situation and time sequence characteristics for different air routes.Finally,to capture the non-linear relationship between the air route flow and time,this thesis establishes a civil aviation flow dataset on the basis of the ststistical air route flow data,and proposes air traffic flow prediction models based on two machine learning schemes,namely Long Short-Term Memory(LSTM)and Support Vector Regression(SVR).The experimental results demonstrate that the proposed prediction model adopting LSTM can achieve better performance than the model using SVR.Specifically,the LSTM-based model shows lower RMSE value and better prediction residual distribution.At present,existing researches mostly use minority information such as historical flow and time correlation which consist of the input of the prediction model for forecasting the aircraft flow.So meteorological-relevant data are further introduced in this thesis,and the author evaluates influences of the meteorological-relevant information on the air route flow by setting control experiments.Results of the control experiments show that the RMSE values of the prediction model can be effectively reduced if the meteorological-relevant information is involved.Besides,the control ability of residual errors of the prediction model can be improved.