Cooperative Spectrum Sensing for Amplify-and-Forward Cognitive Networks

TL;DR

This paper proposes a Bayesian framework for spectrum sensing in cooperative amplify-and-forward cognitive radio networks, introducing algorithms to approximate likelihood ratios under various CSI conditions and evaluating their performance.

Contribution

It introduces novel approximation algorithms for likelihood ratio tests in spectrum sensing, addressing intractability issues with Gaussian, Laplace, and Laguerre series methods.

Findings

Algorithms achieve accurate spectrum detection under imperfect CSI.

Performance bounds validate the approximation accuracy.

Simulation results confirm the effectiveness of the proposed methods.

Abstract

We develop a framework for spectrum sensing in cooperative amplify-and-forward cognitive radio networks. We consider a stochastic model where relays are assigned in cognitive radio networks to transmit the primary user's signal to a cognitive Secondary Base Station (SBS). We develop the Bayesian optimal decision rule under various scenarios of Channel State Information (CSI) varying from perfect to imperfect CSI. In order to obtain the optimal decision rule based on a Likelihood Ratio Test (LRT), the marginal likelihood under each hypothesis relating to presence or absence of transmission needs to be evaluated pointwise. However, in some cases the evaluation of the LRT can not be performed analytically due to the intractability of the multi-dimensional integrals involved. In other cases, the distribution of the test statistic can not be obtained exactly. To circumvent these difficulties we design two algorithms to approximate the marginal likelihood, and obtain the decision rule. The first is based on Gaussian Approximation where we quantify the accuracy of the approximation via a multivariate version of the Berry-Esseen bound. The second algorithm is based on Laplace approximation for the marginal likelihood, which results in a non-convex optimisation problem which is solved efficiently via Bayesian Expectation-Maximisation method. We also utilise a Laguerre series expansion to approximate the distribution of the test statistic in cases where its distribution can not be derived exactly. Performance is evaluated via analytic bounds and compared to numerical simulations.