Receiver Design and Performance Study for Amplify-and-Forward Cooperative Diversity Networks with Reduced CSI Requirement

Abstract

This thesis aims to tackle the theoretical challenges of characterizing the fundamental performance limits of amplify-and-forward (AF) cooperative networks and to resolve the practical challenges in the receiver design for AF systems.

First of all, we study the Shannon-theoretic channel capacity which serves as a benchmark for practical wireless communications systems. Specifically, we derive exact expressions of the ergodic capacity in a single-integral form for general multi-branch AF relay networks with/without the direct link (DL). Moreover, we derive closed-form and tight upper bounds on the ergodic capacity, which facilitate the evaluation of the ergodic capacity. These expressions provide useful theoretical tools for the design of practical wireless AF relaying systems.

We then tackle the practical challenges involved in the design of AF receivers, aiming to substantially reduce the channel state information (CSI) signaling overhead yet achieving satisfactory error performance.

We take the maximum-likelihood (ML) and generalized likelihood ratio test (GLRT) approaches to develop detectors under four typical wireless communications scenarios with little/no knowledge of the CSI. Firstly, for a semi-coherent scenario where only the product of channel coefficients of each relay branch is known, we develop the ML symbol-by-symbol (SBS) detector, which reduces the instantaneous CSI signaling overhead by 50% while achieving comparable performance to the ideal coherent receiver. Secondly, for the noncoherent scenario with only the (second-order) channel statistics and noise variances, we develop a noncoherent ML SBS detector for AF networks employing differential modulations. Thirdly, for AF networks with only the knowledge of the noise variance, we develop a sequence detector using GLRT. Lastly, for a completely blind scenario where the instantaneous CSI, channel statistics, and noise variances are all unknown, we develop a GLRT-based sequence detector. The proposed detectors achieve significant performance improvements over the state-of-the-art counterparts.

The conducted theoretical analysis and practical design will facilitate the design of reliable communications over wireless AF networks with reduced CSI requirement.