Identification Of Prevalent Air Traffic Flow In Terminal Airspace Based On 3D Trajectory Spectral Clustering

With the increase of air traffic flow and the application of new navigation technology(e.g.: RNP, RNAV and PBN etc.) in air traffic management, the controllersbased on the conditions such as flight, weather, and alternate, sort planes and direct aircraft to go round, which inevitably makes actual aircraft trajectory deviate from the standard routes. So the routeshaven't been able to adapt to the change of the traffic flow. The clustering of terminal area historical trajectory and identification of the prevalent air traffic flow has great significance for the improvement of route design and airspace sectordivision.The thesis described the calculation method of trajectory similarity metrics, the application objects and its scope. One model of calculating trajectory similarity metrics based on 3D space grid is proposed. In the clustering analysis, the spectral clustering algorithm is improved to fit terminal area trajectory clustering. Trajectory clustering is transformed into graph partitioning. Trajectory is viewed as a vertex V of an undirected graph G(V,E),and the trajectory similarity value represents a set of weight of edges, In order to react the real and complete terminal air traffic, trajectory outliers are not processed in advance in the clustering.In the identification of prevalent air traffic flow and trajectory outlier, the concept of trajectory relative distance is introduced. A method of identifyingprevalent air traffic flow and trajectory outlier based on trajectory kernel density estimation is proposed, which can identify the two at the same time, eliminate the adverse effects caused by the trajectory outlier, and detect the degree of cluster.Finally, Runway 23 R of Xi'an Xianyang airport's approach trajectory data is used to cluster analysis to verify the accuracy of the method proposed in this thesis. The thesis also gives some possible applications.