Capacity Finder: A Machine Learning-Based Decision Support Tool for Integrated Metroplex Departure Traffic Management

Multi-airport, metroplex airspace environments present a challenging air traffic sequencing and scheduling problem, especially during convective weather when the capacity of the metroplex airspace and airports is degraded. In such situations, departure flights often experience long airport surface (taxi) delays waiting in queues for open slots at convective weather-impacted overhead departure fixes and departure routes. During these situations, the available metroplex capacity can be better utilized by rerouting departure flights from overloaded or closed departure fixes/routes to less busy and open fixes/routes. However, when multiple fixes or routes are impacted by convective weather, there could be several candidate departure rerouting strategies that the FAA and airlines can collaboratively adopt to maximally utilize available capacity, and it may not be immediately clear which strategy will work the best. This paper outlines an ATAC-NASA collaborative research effort that developed a machine learning-based Air Traffic Management (ATM) decision support service called the Capacity Finder. The Capacity Finder provides data-driven decision support for determining the departure rerouting strategy that has the best chance of success given the current and predicted demand and capacity situation. The Capacity Finder makes this determination by first identifying historical operational time-periods (i.e., departure banks) that are similar to the time-period under consideration. Similarity is determined by applying machine learning (clustering) to pre-conditioned data sets that include several traffic demand and capacity-related features. Once the similar departure banks are identified, the Capacity Finder identifies the rerouting strategies that worked well during similar historical situations and suggests them to the airline and FAA users. This paper presents preliminary results on machine learning-based similarity identification for departure banks at the Dallas Fort Worth International Airport (DFW).