A flight profile clustering method combining twed with K-means algorithm for 4D trajectory prediction

4D trajectory prediction is the core technology of modern ATC automatic system. Mining nominal flight profile containing intention of controllers is the key issue in 4D trajectory prediction. In this paper, a clustering method combining time warp edit distance (TWED) with K-means algorithm is proposed to improve the accuracy of nominal flight profile. Firstly, a series of historical trajectory data with same origin and destination are pre-processed to eliminate the effect of outlier point. Secondly, a novel adaptive clustering algorithm is proposed in which the distance between different trajectories is calculated by TWED algorithm rather than the conventional elastic similarity measure. In proposed clustering algorithm, a non-fuzzy clustering model assessment index integrating the separation degree of intra-cluster and condensation degree of inter-clusters is provided to determine K clustering centers adaptively under a density threshold. Then K nominal flight profiles in each cluster are fitted based on matching rules of TWED. Finally, the predicted trajectory containing various control intentions is used to forecast aircraft trajectory in advance to improve efficiency of airspace. The experimental results show that the accuracy and stability of the proposed adaptive clustering algorithm are significantly higher than the state-of-the-art clustering algorithm with dynamic time warping (DTW).