Transmit precoding and Bayesian detection for cognitive radio networks with limited channel state information
Abstract
Cognitive radio (CR) represents a recent direction for enabling coexistence among heterogeneous networks. It can be a potential solution for the problem of scarce spectrum available for wireless communication systems. The study here investigates the underlay and interweave paradigms for the coexistence of CR network of secondary users (SUs) with a primary network of primary users (PUs). Under underlay mode, both networks communicates concurrently using the same resources. With interweave, SU is able to communicate as long as (some) PUs are not active. Usually, underlay or interweave employs multiple antennas at SU to use the spectral resources better and manage the interference towards the primary network. Performance of the CR network under either paradigm depends largely on the amount and quality of channel state information (CSI) available about the different communication links. In practical systems, often CSI at SU has uncertainty since it is deviated from the true one or is not known at all. This uncertainty should be accounted when designing the precoding schemes for SU or otherwise the interference impact on primary networks would violate the quality of service (QoS) requirements for PUs. This dissertation considers two cases regarding to the availability of CSI, the first one is when CSI is imperfect and the second is when CSI is completely not known. For the underlay mode, we investigate two manifolds. The first one addresses the problem of maximizing the throughput of a multiple-input multiple-output (MIMO) SU when CSI of the interference link to PU is completely unknown or partially known. We study the achievable rates for SU under two different QoS requirements for the PU: the conventional interference temperature and leakage rate metrics. When CSI is unavailable, we develop an iterative adaptation algorithm that satisfies the QoS constraint through exploiting the side-information in the primary communication network. When CSI is inaccurate, we model the uncertainty deterministically such that the uncertainty error belongs to a convex compact set defined by the Schatten norm. We design the precoder by following the worst case formulation. We further investigate the relation between the unknown and the inaccurate CSI cases when using the interference temperature metric, and reveal that the performance of the latter is not necessarily better than the former. The second manifold assumes there is uncertainty in the SU intended link for communication as well as in the interference link from SU to PU. Similar to the first manifold, we follow the deterministic modelling using Schatten norm for the uncertainty and apply the worst case philosophy. For a given precoder matrix, we find the worst uncertainty error in the set that describes the uncertainty in each link. We further develop an iterative numerical algorithm for the precoder. Simpler solutions for the precoder and the uncertainty errors are derived under some special instances of the Schatten norm and certain requirement of transmission power. For the interweave mode, we assume there is no CSI available at SU and derive a Bayesian detector for the proposed binary hypothesis problem. For the null or noise model, we propose a conjugate prior for the unknown spatial covariance matrix. For the alternative or data model, we propose a new class of improper priors for the covariance matrix. We introduce the fractional Bayes factor (FBF) approach to enhance the detection capability of the Bayes factor. The developed FBF is compared with those using the conjugate priors for both hypotheses and generalized likelihood ratio test (GLRT), and it yields significant improvement.
Degree
Ph. D.
Thesis Department
Rights
OpenAccess.
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 License.