Interference detection for UAV swarm satellite communications with spatiotemporal manifold filtering

Co-channel interference poses significant challenges to Unmanned Aerial Vehicle Swarm (UAVS) satellite communications within the Space-Air-Ground Integrated Network (SAGIN). To address this issue under conditions of weak interference, limited samples, and non-stationarity, we propose a novel interference detection method based on SAGIN. The core idea involves modeling the covariance matrices of multi-UAV received signals on a Riemannian manifold, thereby transforming interference detection into a geometric classification problem on the manifold. We further design a Spatiotemporal Manifold Filtering (STMF) mechanism, which suppresses noise and enhances weak interference perception through weighted geometric averaging of spatial neighbors’ covariance matrices and temporal trajectory smoothing. Simulation results show that STMF significantly outperforms traditional methods such as energy detection and trace-based detection under weak interference (ISR < -4 dB), achieving up to 38.4% higher average detection probability. Crucially, STMF operates without requiring large labeled datasets, enabling practical online detection for UAVS satellite communications in dynamic electromagnetic environments.