What if an unstable aircraft approach could be identified early enough for a flight instructor to intervene?
That is the problem behind this open source project: AI Early Warning for Unstable Approaches in GA Flight Training.
The project uses public ADS-B data from the OpenSky Network around Daytona Beach International Airport (KDAB) to analyze general aviation training approaches.
It currently:
The initial dataset covered 3 hours of flight activity:
105 approaches analyzed
90 training aircraft approaches
72 approaches selected for instructor review
7 approaches flagged by at least one rule
One Cessna 172S approach triggered both speed and sink-rate warnings, highlighting how data-driven methods could support instructor review.
Importantly, the project does not claim that every automated flag represents a genuinely unstable approach. Ground speed is affected by wind, ADS-B measurements have limitations, and instructor review is essential.
That is what makes the project interesting: the next step is validating and calibrating the rules against expert human judgment.
This is a good example of how open source aviation data can be turned into a practical safety research problem.
🔗 GitHub: https://github.com/samsuseelan/ga-unstable-approach-warning