MagNet:Magnetic Networks for Navigation

Accurate and resilient localization is critical for autonomous UAVs operating in GPS-denied or contested environments. Magnetic anomaly navigation (MagNav) offers a globally available, passive, and infrastructure-free alternative by leveraging spatial variations in the Earth's crustal magnetic field. However, its effectiveness is constrained by low map resolution, sensor noise, and poor observability in magnetically flat regions. To address these limitations, this paper presents MagNet, a cooperative magnetic localization framework that enhances MagNav performance through inter-vehicle communication and sensing. We evaluate three estimation architectures: (1) an Individual EKF that fuses onboard inertial and magnetic data; (2) a Centralized EKF (C-MagNet) that jointly estimates all vehicle states and magnetic biases; and (3) a Split Distributed EKF (D-MagNet) that refines each agent’s estimate using neighbor states and inter-vehicle ranging. Simulations using real-world magnetic anomaly maps demonstrate that collaborative filtering significantly improves position accuracy and bias estimation—achieving up to 17× reduction in DRMS in ambiguous regions. The results reveal clear trade-offs between estimation accuracy, communication bandwidth, and computational scalability, offering actionable guidance for deploying multi-agent magnetic navigation systems in operational GPS-denied scenarios.