Relaying without Decoding

Abstract: Relay networks and cooperative transmission have spurred considerable interest in recent years, promising performance improvements in, e.g., system capacity, robustness and transmission rate. In this thesis, the focus is on the so-called half-duplex relay channels, where in the system, a source node wants to transmit a message to a destination node and a relay node can potentially assist this transmission. The relay is said to be half-duplex in the sense that it cannot transmit and receive at the same time within the same frequency band. It is natural to categorize the potential relaying schemes (operations of the relay node) into two kinds. The relay can either decode the message of the source node, or it can process its received signal without decoding. The thesis investigates various schemes of the second kind and they are termed as ``relaying schemes without decoding.''The first part of the thesis is devoted to the approach of instantaneous relaying. The instantaneous relaying schemes belong to the relaying schemes without decoding, where the relay node is implemented by a deterministic single-variable function. Both linear and non-linear functions are investigated and it is demonstrated that the functions with sawtooth-like shape give higher achievable rates than other functions investigated. Furthermore, the work is extended to half-duplex multiple-access relay channels, where an extra source node is present. For such channels, the relay's operation is ``instantaneous'' in the sense that it is represented by a deterministic function of two variables. Essentially, the function handles the received signals from the two source nodes, combining them together and transmitting the combined signal to the destination node. Novel functions based on the Archimedean spiral mapping and sawtooth-like functions are proposed and demonstrated to perform well, using achievable rate regions and achievable sum rates of the two source nodes as figures of merit.In the second part of the thesis, the class of relaying schemes without decoding and with memory is investigated, where for such schemes, information theoretic source and channel coding with long codewords is used at the relay node. The two predominant schemes of such kind are the so-called compress-and-forward (CF) and quantize-and-forward (QF) schemes. The achievable rate results of the two schemes and some of their variants are derived for static channels. Furthermore, under the assumption of slow fading channels, with transmitter channel state information (CSIT) not available at the source and relay nodes, outage probabilities, expected rates as well as diversity--multiplexing trade-offs (DMT) of the respective schemes are derived and compared. In addition, to compensate for the loss due to the absence of relay CSIT, a finite-resolution feedback link from the destination node to the source node is designed for the CF and QF schemes to provide the relay node with partial CSIT, and thus, the performance of the respective schemes can be improved. Lastly, the thesis considers the problem of lack of relay CSIT from another viewpoint. The concept of hybrid digital-analog coding, as is first investigated in source-channel coding, is adapted and applied at the relay node. Such relaying schemes are termed hybrid digital-analog relaying schemes and their performance in terms of expected rate is studied. It is shown that the hybrid schemes significantly outperform the conventional digital-only (e.g., the CF scheme) and the analog-only schemes (e.g., the so-called amplify-and-forward scheme).