Date

2017

Document Type

Dissertation

Degree

Doctor of Philosophy

Department

Electrical Engineering

First Adviser

Li, Jing

Other advisers/committee members

Yan, Zhiyuan; Venkitasubramaniam, Parvathinathan; Wu, Ping-Shi

Abstract

This dissertation focuses on transmission and estimation schemes in wireless relay network, which involves a set of source nodes, a set of destination nodes, and a set of nodes helps communication between source nodes and destination nodes, called relay nodes. It is noted that the overall performance of the wireless relay systems would be impacted by the relay methods adopted by relay nodes. In this dissertation, efficient forwarding strategies and channel coding involved relaying schemes in various relay network topology are studied.First we study a simple structure of relay systems, with one source, one destination and one relay node. By exploiting “analog codes” -- a special class of error correction codes that can directly encode and protect real-valued data, a soft forwarding strategy –“analog-encode-forward (AEF)”scheme is proposed. The relay node first soft-decodes the packet from the source, then re-encodes this soft decoder output (Log Likelihood Ratio) using an appropriate analog code, and forwards it to the destination. At the receiver, both a maximum-likelihood (ML) decoder and a maximum a posterior (MAP) decoder are specially designed for the AEF scheme.The work is then extended to parallel relay networks, which is consisted of one source, one destination and multiple relay nodes. The first question confronted with us is which kind of soft information to be relayed at the relay nodes. We analyze a set of prevailing soft information for relaying considered by researchers in this field. A truncated LLR is proved to be the best choice, we thus derive another soft forwarding strategy – “Z” forwarding strategy. The main parameter effecting the overall performance in this scheme is the threshold selected to cut the LLR information. We analyze the threshold selection at the relay nodes, and derive the exact ML estimation at the destination node. To circumvent the catastrophic error propagation in digital distributed coding scheme, a distributed soft coding scheme is proposed for the parallel relay networks. The key idea is the exploitation of a rate-1 soft convolutional encoder at each of the parallel relays, to collaboratively form a simple but powerful distributed analog coding scheme. Because of the linearity of the truncated LLR information, a nearly optimal ML decoder is derived for the distributed coding scheme. In the last part, a cooperative transmission scheme for a multi-source single-destination system through superposition modulation is investigated. The source nodes take turns to transmit, and each time, a source “overlays” its new data together with (some or all of) what it overhears from its partner(s), in a way similar to French-braiding the hair. We introduce two subclasses of braid coding, the nonregenerative and the regenerative cases, and, using the pairwise error probability (PEP) as a figure of merit, derive the optimal weight parameters for each one. By exploiting the structure relevance of braid codes with trellis codes, we propose a Viterbi maximum-likelihood (ML) decoding method of linear-complexity for the regenerative case. We also present a soft-iterative joint channel-network decoding. The overall decoding process is divided into the forward message passing and the backward message passing, which makes effective use of the available reliability information from all the received signals. We show that the proposed “braid coding” cooperative scheme benefits not only from the cooperative diversity but also from the bit error rate (BER) performance gain.

Share

COinS