Spectrum sensing and throughput maximization in cognitive radio networks

Abstract

In cognitive radio (CR) systems, reliable spectrum sensing techniques are required in order to avoid interference to the primary users (PUs) of the spectrum. In this dissertation two spectrum sensing techniques are developed, and sensing time and power allocation are optimized in multi-input multi-output (MIMO) CR systems. The motivation of the first proposed spectrum sensing technique is that, in practice, CRs also have to cope with various types of non-Gaussian noise such as man-made impulsive noise, and co-channel interference. However, most of the existing literature on spectrum sensing only considers impairment by additive white Gaussian noise (AWGN). To address this issue, we propose an Lp-norm detector which has tunable parameters that can be adjusted for the underlying type of noise. We also propose an adaptive algorithm for optimization of the Lp-norm parameters which does not require any a priori knowledge of the noise statistics. The motivation for the second proposed spectrum sensing technique is that the signals transmitted by PUs often also contain known pilot symbols for synchronization and channel estimation purposes. Coherent correlation based spectrum sensing techniques can exploit these known symbols but waste the energy contained in the data symbols. Hence, while considering AWGN impairment, we propose a hybrid coherent/energy detection scheme which exploits both the pilot and the data symbols transmitted by the PU. Since the complexity of the globally optimal hybrid detection metric is very high, we develop a simple locally optimal hybrid metric, which turns out to be a linear combination of an energy detection metric and a correlation metric. While the proposed methods improve the accuracy of spectrum sensing, there exists a tradeoff between sensing time and transmission time for CR networks. In this thesis, we investigate this issue for conventional energy detection in MIMO CR networks. Specifically, we optimize the sensing threshold, sensing time, and transmit power of both single-band and multi-band MIMO CR systems for maximization of the opportunistic throughput under transmit power, probability of false alarm, and probability of detection constraints. We also develop efficient iterative algorithms for solving these non-convex optimization problems based on the concept of alternating optimization.