Research On Precise Prediction Method Of 4D Trajectory Based On Data Mining

With the advancement of economy and technology, Chinese air traffic flow has experienced rapid growth. It is imperative to use scientific air traffic flow management to solve the emerging issues of air transport system such as low efficiency, and large-area flight delays. 4D trajectory prediction technology is one of the key technologies of air traffic flow management and core technologies of the development of ATC automation system, so the focus of current study is to explore accurate trajectory prediction technology.With the coming of big data era of ATC system, air traffic operation environment has been increasingly complex and characterization data have been of massive surge, multi-source heterogeneity, highly complexity. Thus, it is also urgent to solve how to extract useful information, elaborate air traffic situation, dig into organization pattern and evolutionary principle of traffic flow, and effectively apply them to prediction and other technologies.In terms of above aspects, this paper proposes a precise 4D trajectory prediction algorithm based on data mining. The algorithm is to establish a prediction model through deep analysis on enormous historical flight data. When predicting, the model selects the historical flight track which highly matches the current input condition and uses it as the output of 4D trajectory track. As a real flight track of higher sampling frequency(20 seconds around), the historical track is more detailed and authentic compared with the output of 4D trajectory track which is only based on fix points sequence. This trajectory track is called precise trajectory prediction in this paper. This paper firstly summarizes current status of research on 4D trajectory prediction and produces a classification of prediction algorithm. Secondly, it proposes an analysis method of historical track based on spectral clustering and kernel density estimation. On the basis of the analysis method, it uses frequent path tree and R tree method in temporal and spatial data management to establish an FPR-Tree index aimed at historical flight track. Then, it discusses the way to find k-nearest neighboring trajectories in the FPR-Tree. Finally, this paper puts forward a 4D trajectory prediction method based on FPR-Tree, expounds the algorithm process and parameters, and uses real air transport network, historical flight data of Central South area to test the algorithm. It shows that pre-tactical accuracy rating can be up to over 60%, tactical accurate rate over 80% when using the prediction method. In addition, this paper also describes the implementation of the prototype system which base on above algorithm.