Broadband UAV DoA Estimation Using a Multi-Channel MZ–Sagnac Fiber-Optic Acoustic Sensing Array
Abstract
Low-altitude unmanned aerial vehicle (UAV) activity in urban areas has created increasing demand for passive monitoring methods. Fiber-optic acoustic sensing offers a promising alternative to conventional electrical microphones for acoustic UAV monitoring, owing to its multiplexing capability and immunity to electromagnetic interference (EMI). This paper presents a compact, field-deployable UAV direction-finding platform in which a tetrahedral array of fiber-optic acoustic sensors (FOASs) provides non-coplanar spatial baselines for joint azimuth–elevation estimation, while a wavelength-multiplexed Mach–Zehnder–Sagnac (MZ–Sagnac) interferometric readout architecture enables simultaneous four-channel phase interrogation and enhances the end-to-end response over a broad acoustic band. Frame-level direction-of-arrival (DoA) observations are obtained using steered response power with phase transform (SRP-PHAT) on a spherical grid. Urban rooftop measurements of a quadrotor UAV (DJI Mavic Pro) at different distances are used to quantify observation quality and determine the effective localization range. The 200–6000 Hz band provides the best practical localization performance, enabling reliable localization at approximately 100–120 m under the tested conditions. In this band, a lightweight temporal stabilization module based on prediction-window peak search and error-state Kalman filtering maintains a continuous DoA trajectory under intermittent high-level broadband interference, reducing the peak-level miss rate from 12.8% to 3.5% and achieving a tracking RMSE of 1.27$^\circ$.