A Robust Decentralized Multi-Sources Joint Power Spectra Separation and DOA Estimation Algorithm

In cognitive wireless communication networks, multi-sources power spectra separation enables judicious routing, scheduling and beamforming as the interference can be effectively avoided. In this work, we propose a robust decentralized algorithm to solve the multi-source power spectra separation problem for the wireless networks consisting of multi-antenna nodes. We adopt a two-step iterative re-weighting strategy to detect the corrupt receiver reports. In each iteration, the blind spectra separation is formulated as a structured matrix factorization (SMF) problem, which is solved by the alternating direction method of multipliers (ADMM) algorithm. Moreover, we proposed a power spectral average strategy to mitigate the negative effect of noise. The direction of arrival (DOA) is then estimated in the second step during each iterative procedure. A decentralized scheme is applied to solve the resulting least squares (LS) problem. Finally, simulations are performed to validate the efficiency of the proposed robust decentralized multi-source power spectra separation and DOA estimation algorithms.