Deep Learning in Aeronautics: Air Traffic Trajectory Classification Based on Weather Reports

Abstract

[EN] New paradigms in aviation, as the expected shortage of qualified pilots and the increasing number of flights worldwide, present big challenges to aeronautic enterprises and regulators. In this sense, a concept known as Single Pilot Operations arises in the task of dealing with these challenges, for which, automation becomes necessary, especially in Air Traffic Management. In this regard, this paper presents a deep learning-based approach to leveraging the job of both ground controllers and pilots. Making use of Meteorological Terminal Air Reports, obtained regularly from every aerodrome worldwide, we created a model based on a multi-layer perceptron capable of determining the approach trajectory of an aircraft thirty minutes prior to the expected landing time. Experiments on aircraft trajectories from Toulouse to Seville, show an accuracy, recall and F1-score higher than 0.9 for the resultant predictive model.