Position Estimation of Unmanned Aerial Vehicles in Contested Environments using Dense Matching Networks

Abstract

This paper presents an approach to predicting the trajectory and the position of unmanned aerial vehicles (UAVs) in contested global navigation satellite system (GNSS) environments. Our approach utilizes dense matching networks to analyze and predict patterns in UAV movement and then effectively recovers temporal trajectory information and accurately predicts UAV position changes using images and altitude data. This problem was part of the KDDBR 2022 Kaggle Competition, where we obtained a Root Mean Squared Error (RMSE) of 0.29603, which made it the runner-up solution. This could have significant implications for the future of UAV navigation, potentially leading to safer and more efficient operations in various applications.