Updating Air Traffic Models Using Discrete Fréchet Distance

This paper extends prior research on air traffic trajectory clustering by presenting a novel two-stage approach to overcome the constraints of conventional clustering algorithms, particularly when dealing with smaller datasets. In the first stage, clustering is performed using Procrustes analysis. Then, the second stage incorporates new trajectories into the resulting clusters. This integration is achieved by assessing the incoming data using the discrete Fréchet distance and adjusting the cluster deviation and track count parameters accordingly. Through this iterative process, a more precise and comprehen-sive representation of airspace is attained, resulting in enhanced clustering metrics. Moreover, a detailed examination of non-clustered trajectories using the Fréchet distance metric reveals sparse or shifted trajectories that exhibit minimal deviation from existing clusters. The proposed method demonstrates its efficacy by notably reducing the number of non-clustered trajectories while maintaining an acceptable level of increased deviation within clusters compared to the original clustering algorithm. This refined model offers valuable insights into air traffic patterns and can be applied to various domains, including air traffic flow prediction, airspace optimisation, and facilitating the integration of UAVs into the airspace.