Description: The previous work on this SBIR was aimed toward using machine learning to analyze National Airspace System (NAS) data to identify days that have characteristics similar to the characteristics that the user specifies. Those days would then be used in the scenario generation phase of modeling and simulation of the NAS to analyze new air traffic management (ATM) concepts and technologies. The proposed SBIR Phase II-E work is aimed toward supporting near real-time analysis (rather than modeling and simulation). For example, an airline operations center or FAA traffic flow management (TFM) supervisor might look at a current day’s weather forecast and want to know, based on prior data, the best way to react. Our tool would then use machine learning techniques to identify not only historical days with similar weather characteristics as those currently being experienced/forecast, but also to identify the traffic management initiatives (TMI) that were instituted on those historical days to see which were successful and which were not. A key part of the proposed work is to continue to use machine learning to produce meta data based on historical NAS data. The NAS data, thus far, have included airport traffic data, TFM metrics related to ground delay and ground stops, wind data, geometric data such as sectors, centers, airways, and much more. These data are processed offline to create labeled metadata using a clustering procedure (machine learning) and stored in a database to enable search. In the proposed work, we will augment the data, create additional metrics, develop analytics to support decision-making, and enhance the user interface.

Benefits: This work fits very well with the prospective Digital Services for Aviation (DS4A) sub-project; one objective of DS4A is : Similar day analysis, and more generally classifications that help drive decisions and impact notification. Simple examples of this may be identifying a ‘high traffic management restriction’ day for a particular airport or region. While there could be many definitions for this classification, a community wide example could help expedite the ability to predict when a target classification is likely to occur.

Airlines and FAA users could use our tool. For example, including historical airline decision support and operational outcome data would assist a partner airline in determining flight preferences for communicating to the FAA as a part of the collaborative decision-making (CDM) process.