MULTI-SENSOR FUSION FOR MULTI-TARGET TRACKING IN COUNTER-UAS APPLICATIONS

Reliable detection and tracking of non-cooperative drones is a central challenge in C-UAS applications, where no single sensor offers complete situational awareness. This work proposes a multi-target tracking architecture that fuses a 3D Radar with a network of three RF-DOA bearing-only sensors. A dual-pipeline design keeps the two modalities independent: each feeds a dedicated UKF tracker with JPDA. The core contribution is a triangulation engine that intersects the RF-DOA bearings, rejects ghost points through a multi-sensor consensus, and produces synthetic detections carrying a geometry-dependent covariance. The two track sets are then merged by a GM-PHD filter with a Chernoff (Covariance Intersection) updater restricted to the horizontal sub-state, while altitude is transferred from the Radar alone. A 51-run Monte Carlo evaluation using OSPA and SIAP metrics shows that fusion matches the Radar accuracy while adding the coverage and robustness contributed by the RF-DOA subsystem, with a marked improvement when RF-DOA tracks are available. A complementary sensitivity analysis on the predictor process noise confirms this complementarity: for low value of the predictor process noise the RF-DOA network sustains continuous coverage of abruptly manoeuvring targets that the Radar alone struggles to follow, while the Radar contributes superior localisation accuracy, together yielding a fused track that combines the strengths of both sensing modalities.