A machine learning approach to air traffic interdependency modelling and its application to trajectory prediction

Abstract

Air Traffic Management is evolving towards a Trajectory-Based Operations paradigm. Trajectory prediction will hold a key role supporting its deployment, but it is limited by a lack of understanding of air traffic associated uncertainties, specifically contextual factors. Trajectory predictors are usually based on modelling aircraft dynamics based on intrinsic aircraft features. These aircraft operate within a known air route structure and under given meteorological conditions. However, actual aircraft trajectories are modified by the air traffic control depending on potential conflicts with other traffics. This paper introduces surrounding air traffic as a feature for ground-based trajectory prediction. The introduction of air traffic as a contextual factor is addressed by identifying aircraft which are likely to lose the horizontal separation. For doing so, this paper develops a probabilistic horizontal interdependency measure between aircraft supported by machine learning algorithms, addressing time separations at crossing points. Then, vertical profiles of flight trajectories are characterised depending on this factor and other intrinsic features. The paper has focused on the descent phase of the trajectories, using datasets corresponding to an en-route Spanish airspace volume. The proposed interdependency measure demonstrates to identify in advance conflicting situations between pairs of aircraft for this use case. This is validated by identifying associated air traffic control actions upon them and their impact on the vertical profile of the trajectories. Finally, a trajectory predictor for the vertical profile of the trajectory is developed, considering the interdependency measure and other operational factors. The paper concludes that the air traffic can be included as a factor for the trajectory prediction, impacting on the location of the top of descent for the specific case which has been studied.

Evaluation of Extended Projected Profiles impact on vertical trajectory prediction

“…The analysis is arc-length-based (see Figure 1) [5] [6]. The distance from the entry point to the sector along the central trajectory of the flow until the position of TOD is the 𝛾 𝑇𝑂𝐷 .…”

Aircraft 4D Trajectory Prediction in Civil Aviation: A Review

“…In addition to the two commonly used methods of regression model and neural network, other machine learning methods have also appeared [73][74][75][76][77][78][79], such as genetic algorithm (GA), ant colony algorithm, and support vector machine (SVM), etc. Here, this is regarded as a separate category.…”

“…Linear regression: [41,59,60] Stepwise regression: [57] Nonlinear regression: [41,58] Neural network model Feedforward neural networks: [61,70,75,84] Elman neural network: [78] LSTM: [62][63][64][65]67,71,72] DNN + LSTM: [40] CNN + LSTM: [66] GRU: [79] Bayesian neural network: [40,69] Generative adversarial network: [68] Other methods A gaussian mixture model with clustering: [82] Random forest with clustering: [83] Neural Networks with clustering: [8] Nonparametric interval prediction: [73] Genetic programming: [76]…”

Data-Driven Methods for Aviation Safety: From Data to Knowledge

“…The exploitation of Data-Driven Models (DDMs) in the ATM domain is quite extensive. Indeed, research has focused on a number of different fields, such as taxi-out time prediction [12,17], trajectory prediction [1,25], air traffic flow extraction [6,26], and flight delay prediction [5,24]. In the safety domain, some relevant applications of DDMs are proposed in literature to predict safety events or performance [7,19], or to provide safety metrics [2] or accident precursors [13].…”