A Performance Learning Method for Aircraft Trajectory Modeling

In this paper, an aircraft trajectory modeling method is presented to extract and identify flight status on the basis of performance learning. The aim is to facilitate understand of the intrinsic mechanism of flight operations and improve the quality of trajectory analysis and waypoint identification, as it is of vital significance for the implementation of TBO in the future. In the first step, the spatial and status data of real trajectory for different aircraft types and flight phases is categorized to generate a subset of performance data. Next, Gaussian mixture clustering algorithm and a comprehensive evaluation index are utilized to extract flight status features from the subset data. The KD-Tree method is used to quickly and effectively classify the input trajectory points and identify corresponding flight status with the extracted features as evaluation criteria. Finally, the K-fold cross-validation method is applied to assess the feasibility. Experimental results based on real data show the method could achieve a satisfying accuracy in trajectory waypoint categorization.