Aircraft Localization Using ATC Data with Nanosecond Precision from Distributed Crowdsourced Receivers

In this paper, we present the first place solution for the Aircraft Localization Competition, which was held on the AIcrowd platform between 15 June 2020 and 31 January 2021 and was organized by the OpenSky Network and the Cyber-Defence Campus of armasuisse Science and Technology. The data for the competition was collected by the OpenSky Network from hundreds of crowdsourced low-cost receivers with nanosecond precision timestamps. Many receivers experienced clock drift and random walk and even provided fully broken timestamps. The solution combines well-known multilateration positioning with a variety of filtering methods and two tailored models for radio wave propagation and receiver clock drift to predict unknown aircraft locations. In this solution, we managed to synchronize 241 receivers, including 36 GPS-equipped, and achieved 81.9 m RMSE 2D distance prediction accuracy on 70% of samples on the private leaderboard.