Adaptive hybrid filtering for UAV swarm navigation under abnormal measurements

In complex environments, unmanned aerial vehicle (UAV) swarms face challenges such as GNSS signal interference and abnormal or incomplete measurements, severely affecting navigation and positioning accuracy. To address traditional Distributed Consensus Kalman Filter (DCKF) limitations in anomaly identification and robustness, this paper proposes an adaptive hybrid filtering algorithm. First, a Residual-Adaptive Distributed Consensus Kalman Filter (RADCKF) is designed, utilizing chi-square residual tests to isolate anomalous nodes in real time and adaptively allocating state estimation weights based on neighborhood information. To further enhance system integrity, a sliding-window chi-square mechanism integrates the RADCKF with a federated filter. Compared to existing methods, the main novelty lies in this autonomous mode-switching capability, which dynamically recovers navigation accuracy when GNSS is denied. The experimental and simulation results demonstrate that under node failure conditions, the navigation accuracy of abnormal nodes has improved by 82.72%, 81.68%, and 80.05% in the northeast, east, and south directions, respectively, compared to previous methods. The system maintains high robustness even under multi-node anomalies, providing a highly reliable solution for UAV swarm navigation in complex environments.