A Data Analytics Approach for 4D Trajectories in ATFM

4D trajectories is a key concept of the future Air Traffic System as highlighted in relevant R&D programme both in Europe and in the US., i.e., SESAR and NEXTGen. The scope is to guarantee flexibility to Airspace Users (AUs) and, at the same time, to improve system predictability. The ambition is to allow AUs to fly their preferred route whenever possible. From the system perspective, this calls for a new class of Air Traffic Flow Management (ATFM) models that explicitly consider AUs’ preferences. This is a quite challenging task because AUs’ preferences, which depend on many factors, e.g., costs, duration, geometry etc., are not always fully known. We present a data analytics approach for 4D trajectory optimization. The approach is composed of a predictive component that learns AUs’ preferences and reduces the set of possible 4D trajectories accordingly; and of a prescriptive component that assigns 4D trajectories to flights by solving a path-based integer programming formulation. Using trajectories queried from Eurocontrol DDR2 data repositories, the learning phase is mainly based on tree classifiers, support vector machines and multiple regression. As a result, a set of trajectories and information on related AUs’ preferences are computed for each flight, and feed the optimization model.