Uncertainty-Aware GP Bayesian Optimization for UAV-Based Closed-Loop RSSD Interference Localization

Abstract

Unmanned aerial vehicle (UAV)-based interference source localization is increasingly important in dense 5G-Advanced and emerging 6G networks, where unknown interferers degrade communication reliability and bias measurement-driven channel modeling. However, propagation uncertainty caused by multipath, blockage, and non-line-of-sight (NLOS) propagation makes the measurement-to-location relationship highly nonlinear and time-varying, which limits the reliability of conventional localization methods. To address this challenge, we propose a single-UAV localization framework that uses spatial received signal strength indicator (RSSI) measurements collected along an adaptive flight trajectory to localize a dominant interference source. A robust channel-aware localization algorithm estimates the source location together with its associated uncertainty. A Gaussian process (GP) Bayesian optimization method selects waypoints by maximizing the expected reduction in localization uncertainty. By integrating channel-aware localization with uncertainty-aware waypoint selection, the proposed closed-loop received signal strength difference (RSSD) interference localization framework iteratively updates the source estimate and plans subsequent waypoints. At a signal-to-noise ratio (SNR) of 20 dB under multipath propagation, the proposed framework reduces the flight distance for a given localization root mean square error (RMSE) by about 60% and the localization RMSE at a given flight distance by about 70% compared with the baseline methods.