Predicting aircraft taxiing estimated time of arrival by cluster analysis

In order to predict the aircraft taxiing estimated time of arrival (ETA) to reduce scene conflicts and improve the operating efficiency of airports, the Kalman filter algorithm is employed to preprocess historical trajectory data for an airport scene. For the first time, in order to measure the distance between trajectory samples, the aircraft taxiing period and the number of aircraft in a scene are used as sample features for aircraft taxiing ETA prediction. The dynamic time warping algorithm is employed to extract the different features of the trajectory. The Euclidean distance between two sample features is taken as their similarity. The initial clustering centre is determined based on the divided difference maximum principle by using the K‐means algorithm to cluster samples and selecting the cluster with the highest matching degree according to the time period of the aircraft under planning and the number of aircraft operating in the scene. The trajectory sequence for the sample at the cluster centre and the static path planned by the tower are combined to predict the aircraft taxiing ETA. The accuracy of this method at predicting the aircraft taxiing ETA was demonstrated in simulations by comparing and analysing actual trajectory data with predicted taxiing ETA.