The U.S. National Airspace System (NAS), its interconnected operations, and resultant dynamics (e.g., flight delays, route configurations) can be understood at different scales. At the resolution of the entire NAS, a component (e.g., Denver Center) could be considered as one homogeneous node, with dependencies on other components (e.g., adjacent centers) and exogenous stakeholders (e.g., Air Traffic Control System Command Center, airline network operations centers). Understanding the connected behavior of the NAS in a data-driven and adaptive way is critical for rigorously determining whether interventions, strategic or tactical, were successful. However, within NAS components flight operations induce a multitude of relationships between parts of the airspace at many levels. To capture, analyze, and build upon such a connected and multi-scaled system, we require graph-based network models at varying resolutions, which can be adapted to fit a particular analysis use case. As an example, graphical representation at a higher resolution within a component may be required to capture nuanced behavior in analyses focused on local perturbations. In this work, we present a pipeline that constructs flexible graph-structured data from flight trajectories, and leverage this for different case studies within the NAS, all focused on evaluating different aspects of traffic flow management, and at different scales.
Adaptable Graph Networks for Air Traffic Analysis Applications
2023-04-18
1237318 byte
Conference paper
Electronic Resource
English
Adaptable Anomaly Detection in Traffic Flow Time Series
IEEE | 2019
|Graph transformer based dynamic multiple graph convolution networks for traffic flow forecasting
DOAJ | 2023
|Graph transformer based dynamic multiple graph convolution networks for traffic flow forecasting
Wiley | 2023
|