A data-driven approach for dynamic and adaptive aircraft trajectory prediction

Abstract

Accurate and reliable trajectory prediction (TP) is a fundamental requirement to support trajectory-based operations (TBOs). Particularly, the mismatch between planned and flown trajectories (caused by operational uncertainties from airports, Air Traffic Control interventions, Airspace Users behavior and changes in flight plan data) act as a driver for shortcomings in flow and capacity management (e.g., congestion and suboptimal decision making) and as a precursor for potential safety conflicts. Therefore, enhanced traffic forecasts (which integrate uncertainty assessment and include different sources of relevant flight information) may enable improved demand-capacity balancing and conflict detection and resolution (CD&R) models. Moreover, new methodological approaches, as the exploitation of historical data by means of machine-learning techniques is expected to boost TP performance. This paper presents the data-driven, dynamic and adaptive TP framework achieved within DIAPasON project, considering adaptation to different Airspace Users’ characteristics and strategies. The main target is the development of a methodology for TP and traffic forecasting in a pre-tactical phase (one day to six days before the day of operations), when few or no flight plans are available. This can be adjusted to different time scales (planning horizons), taking into account the level of predictability of each of them.