Compressive Spectrum Sensing in Cognitive IoT

PhDWith the rising of new paradigms in wireless communications such as Internet of things (IoT), current static frequency allocation policy faces a primary challenge of spectrum scarcity, and thus encourages the IoT devices to have cognitive capabilities to access the underutilised spectrum in the temporal and spatial dimensions. Wideband spectrum sensing is one of the key functions to enable dynamic spectrum access, but entails a major implementation challenge in terms of sampling rate and computation cost since the sampling rate of analog-to-digital converters (ADCs) should be higher than twice of the spectrum bandwidth based on the Nyquist-Shannon sampling theorem. By exploiting the sparse nature of wideband spectrum, sub-Nyquist sampling and sparse signal recovery have shown potential capabilities in handling these problems, which are directly related to compressive sensing (CS) from the viewpoint of its origin. To invoke sub-Nyquist wideband spectrum sensing in IoT, blind signal acquisition with low-complexity sparse recovery is desirable on compact IoT devices. Moreover, with cooperation among distributed IoT devices, the complexity of sampling and reconstruc- tion can be further reduced with performance guarantee. Specifically, an adaptively- regularized iterative reweighted least squares (AR-IRLS) reconstruction algorithm is proposed to speed up the convergence of reconstruction with less number of iterations. Furthermore, a low-complexity compressive spectrum sensing algorithm is proposed to reduce computation complexity in each iteration of IRLS-based reconstruction algorithm, from cubic time to linear time. Besides, to transfer computation burden from the IoT devices to the core network, a joint iterative reweighted sparse recovery scheme with geo-location database is proposed to adopt the occupied channel information from geo- location database to reduce the complexity in the signal reconstruction. Since numerous IoT devices access or release the spectrum randomly, the sparsity levels of wideband spec-trum signals are varying and unknown. A blind CS-based sensing algorithm is proposed to enable the local secondary users (SUs) to adaptively adjust the sensing time or sam- pling rate without knowledge of spectral sparsity. Apart from the signal reconstruction at the back-end, a distributed sub-Nyquist sensing scheme is proposed by utilizing the surrounding IoT devices to jointly sample the spectrum based on the multi-coset sam- pling theory, in which only the minimum number of low-rate ADCs on the IoT devices are required to form coset samplers without the prior knowledge of the number of occu- pied channels and signal-to-noise ratios. The models of the proposed algorithms are derived and verified by numerical analyses and tested on both real-world and simulated TV white space signals