09 - Crewed and Uncrewed Aircraft Operations and Traffic Management

WEATHER-DRIVEN DEEP LEARNING FOR AIRPORT RUNWAY DIRECTION FORECASTING

H.W. Kim¹, Y.J. Eun¹, D.K. Jeon¹; ¹Korea Aerospace Research Institute, South Korea

We propose a weather-only deep learning model for hourly runway-configuration forecasting up to 24 hoursrnahead at Incheon International Airport. An encoder-only inverted Transformer with a switch-aware trainingrnobjective improves transition-timing localization while preserving per-hour configuration accuracy, achievingrnover 90% overall accuracy on five years of data.

View Paper

DOI: 10.71945/icas2026_0838