Adaptive Federated Kalman Filter With Fault Detection for INS/GNSS/VC/SFC Integration in Fixed-Wing UAVs

This paper proposes an adaptive federated Kalman filter (AFKF)–based INS/GNSS integration framework enhanced with velocity and specific-force constraints (VC and SFC) to ensure reliable UAV navigation during GNSS outages. The proposed method operates in three phases: (1) an adaptive Kalman filter is first used during GNSS-available intervals to tune the process noise covariance Q; (2) classical centralized INS/GNSS fusion when GNSS is available; and (3) a fault-tolerant AFKF with fault detection and isolation (FDI) is employed during GNSS outages when pseudo-measurements are used. Simulation results based on a UAV trajectory demonstrate that incorporating VC and SFC pseudo-measurements significantly improves INS accuracy during GNSS outages (reducing RMS position error from 138.0 m to 38.2 m). When VC becomes faulty, the classical fusion deteriorates severely (3479.4 m RMS error), while the AFKF with FDI mitigates this degradation (672.8 m RMS error). Finally, using finely tuned Q from adaptive estimation further enhances the FDI performance, achieving complete fault isolation and reducing RMS error to 30.1 m. The proposed AFKF with adaptive tuned Q provides a robust and fault-tolerant solution for UAV navigation in GNSS-denied conditions.