Study on the resolution of multi-aircraft flight conflicts based on an IDQN

Sui, Dong; Xu, Wei; Zhang, Kai

doi:10.1016/j.cja.2021.03.015

Cited by 13 publications

(1 citation statement)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Once the time has been recorded, time shifts can then be made to generate lateral conflicts. In actual flight operation, ATCs are needed to solve for 3D air traffic conflicts [6]. It is therefore important to include some semblance of the aforementioned cases in the scenarios when developing a model, especially so when the model acts as a decision support tool.…”

Section: A Simulations and Data Preparationmentioning

confidence: 99%

A Supervised Learning Approach for 4D Air Traffic Conflict Prediction under Trajectory Uncertainty

Mohamed,

Dang,

Alam

2023

2023 IEEE 26th International Conference on Intelligent Transportation Systems (ITSC)

View full text Add to dashboard Cite

This paper presents a Supervised Learning approach for the problem of air traffic conflict prediction in 4dimensional space (3-dimensional space and time) under trajectory uncertainties, resulting in non-nominal conflict points. Decision support systems for conflict prediction offer shortterm conflict alerts, triggering alarms within a two-four-minute window before loss of separation (LOS), while medium-term conflicts are flagged eight to twelve minutes prior to LOS. However, the underlying models rely on flight plans and extrapolated short-term trajectory prediction. Such models lack the capabilities of predicting emergent conflicts and new conflict birth points resulting from track deviation due to nonnominal events such as weather. These deficiencies manifest themselves in the form of misdetection in the event of nonnominal conflicts. With the goal to build better tools for conflict prediction, the present study models trajectory uncertainty in the form of weather avoidance and aircraft intent during the generation of conflict scenarios. The scenarios were then simulated in BlueSky Open Air Traffic Simulator and the resulting conflict trajectories were used as inputs for supervised machine learning. The present study also includes new features, via the introduction of the Jacobian matrix for space and time, for machine learning model training as opposed to the regular features used in the past. It is demonstrated that features with rate of change are more significant in identifying conflict as opposed to classical features. The results also demonstrated significant improvement in conflict prediction (with and without trajectory uncertainty) for a two-to-twelve-minute window, as compared to the state-of-the-art conflict detection algorithms.

show abstract

Section: A Simulations and Data Preparationmentioning

confidence: 99%