A Framework for Unlocking Hidden Air Traffic Management Data Sources Through Advanced Technologies

Our proposed innovation, Air Traffic Language Analysis System for Coordination, Operational Prediction, and Enhancement (ATLASCOPE), is an Artificial Intelligence-based airspace service that improves the efficiency of traditional civil aviation missions in the near-term NAS, thus directly addressing Subtopic A3.01. ATLASCOPE does this by unlocking crucial ATM information hidden in unstructured, non-traditional data sources. A key example is unpublished traffic management initiatives (TMIs) that are negotiated between FAA facilities via phone conversations and implemented without being published via SWIM or other sources. Lack of information on these TMIs prevents airlines from making preemptive adjustments to minimize impact to their networks, leading to significant, avoidable operating costs. Information on unpublished TMIs can be inferred indirectly from controller-pilot conversations. ATLASCOPE Monitors these voice comm data feeds, Converts the audio to text, Extracts crucial information from the text via AI-based NLP, and Provides insights and predictions based on this information to airlines and other NASA Digital Information Platform (DIP) users. ATLASCOPE supports NASA ATM-X’s DIP sub-project’s plans for demonstrating the pre-departure rerouting digital service (called CDDR) in the SNFP-Ops demos, by providing a hitherto unavailable source of information on the impact of unpublished TMIs on departure route capacities and departure delays. This increases the effectiveness of CDDR. The SBIR brings together an innovative approach for creating training datasets for fine-tuning audio-to-text conversion Large Language Models, cutting-edge AI model training techniques that avoid pitfalls of over- and under-fitting, and an innovative flight track processing-based approach for creating unpublished TMI activity truth dataset. The anticipated Phase I result is demonstration of the ATLASCOPE prototype on one or more historical unpublished TMIs scenarios.