ATC Decision Support
A research prototype testing whether air traffic control conflict alerting can be composed from a population of small, independent, deterministic detector functions rather than a single monolithic rule engine. Every alerting rule is a plain function a domain expert can read; the same traffic picture produces the same alerts every time.
This is not certified software and is not usable for operational air traffic control. Thresholds are untuned. Live mode classifies only a fraction of nearby traffic, several detector inputs are synthesised because ADS-B does not carry them, and the incident scenarios are reconstructions with the outcome already known. Everything here is designed to sit underneath a controller's judgment, never in place of it.
What you can do here
- Scenarios — twenty-eight deterministic scenarios: nine reconstruct documented incidents (Tenerife 1977 through the 2025 Potomac midair, sources cited per scenario), eleven form a negative-control corpus that must stay quiet, the rest are synthetic stress cases. Each loads paused.
- Live mode — the same detector population running on live ADS-B and METAR for six US airports. No key required.
- Squawks — transponder code reference.
Data sources
Live traffic from adsb.lol, a volunteer ADS-B network. Weather from the NOAA Aviation Weather Center. Neither endorses this work. Incident reconstructions draw on published NTSB, ICAO, CIAIAC and ANSV reports.
Author
Filipe Brito Ferreira · [email protected] · fbritoferreira.com. The whitepaper documenting the architecture, the measurement methodology and the findings that ran against the design is published at A Multi-Detector Decision-Support Architecture for Air Traffic Control. The source repository is scheduled for publication under Apache-2.0.
← Back to scenarios