Bayesian Localization of a Moving Source via Map-Informed Sensing Using a Low-Altitude UAV

Abstract:
Accurate localization of unknown moving signal sources using a low-altitude unmanned aerial vehicle (UAV) remains challenging, particularly in urban areas with complex lineof- sight (LOS) and non-line-of-sight (NLOS) conditions. Existing localization methods typically assume static sources or ignore environmental map information essential for handling LOS/NLOS scenarios. This paper proposes a Bayesian framework that integrates a dynamic mobility model with a map-informed observation model to localize a moving signal source. The observation model distinguishes the LOS and NLOS propagation based on 3D environmental maps. Meanwhile, source mobility dynamics are captured using a probabilistic Gauss-Markov transition model. The proposed framework jointly estimates both the moving source locations and LOS/NLOS propagation parameters via an alternating optimization algorithm. Simulations conducted under an elliptical UAV trajectory, designed to avoid coverage holes, demonstrate that the proposed method achieves up to a fivefold improvement in localization accuracy compared to conventional baseline approaches.