Multi-aspect Robust Adaptive Streaming Tensor Completion for Space-based Spectrum Situation Map Construction

By leveraging low Earth orbit satellites, space-based spectrum monitoring enables the construction of a wide-area spectrum situation map (SSM) to characterize spatio-temporal-frequency electromagnetic information for spectrum surveillance and management. However, due to the constraints of predetermined orbits and harsh space-ground transmission environments, the collected spectrum measurements are inherently incomplete and corrupted by anomalies. Moreover, due to the time-varying satellite coverage, the dimensions of the spectrum measurements also fluctuate over time. To recover the SSM from such suboptimal spectrum measurements, this work first formulates the problem as a low-rank tensor completion task by exploiting inherent correlations in the spectrum data. Subsequently, a dynamic tensor decomposition framework is introduced to model the newly emerging measurement patterns, which adaptively handles uncertain dimension changes by leveraging tensor factorization properties and time-series forecasting techniques. Furthermore, a multi-aspect robust adaptive streaming tensor completion scheme is proposed to incrementally update the SSM through time-aware subspace tracking. Numerical results validate the effectiveness and efficiency of the proposed scheme against state-of-the-art streaming tensor completion algorithms.