Abstract

In this thesis, we investigate several relaying strategies for cooperative networks with the aim of finding techniques to improve the performance of such networks. The objective here is to increase the spectral efficiency while achieving full diversity. Therefore, we focus on two-way
relaying and relay assignment since they are both efficient ways in improving the spectral efficiency of cooperative networks. Specifically,we propose efficient relay strategies to cope with the asymmetric data rates in two-way relay channels and address practical issues in relay assignment.

In the first part of the thesis, we consider two decode-and-forward (DF) relaying schemes for two-way relaying channels where the two sources may have different rate requirements.
One scheme combines hierarchical zero padding and network coding (HZPNC) at the relay. The novelty of this scheme lies in the way the two signals (that have different lengths) are network-coded at the relay. The other scheme is referred to as opportunistic user selection (OUS) where the user with a better end-to-end channel quality is given priority for transmission. We analyze both schemes where we derive closed form expressions for the end-to-end(E2E) bit error rate (BER). Since the two schemes offer a trade-off between performance and throughput, we analyze and compare both schemes in terms of channel access probability and average throughput. We show that HZPNC offers better throughput and fairness for both users, whereas OUS offers better performance. We also compare the performance of HZPNC with existing schemes including the original zero padding, nesting constellation modulation and superposition modulation. We demonstrate through examples the superiority of the proposed HZPNC scheme in terms of performance and/or reduced complexity.

In the second part of the thesis, we consider a hybrid relaying scheme for two-way relay channels. As per the proposed scheme, if the E2E signal-to-noise ratio (SNR) of both users is above a specified threshold, both sources transmit over orthogonal channels and the relay node uses hierarchical modulation and network coding to relay the combined signals to both sources in the third time slot. Otherwise, the user with the better E2E SNR transmits, while the other user remains silent. The advantage of the proposed scheme is that it compromises between throughput and reliability. That is, when both users transmit,
the throughput improves. Whereas when the better user transmits, multiuser diversity is achieved. Assuming asymmetric channels, we derive exact closed-form expressions for the E2E BER, access probability and throughput for this scheme and compare its performance to that of existing schemes. We also investigate the asymptotic performance of the proposed scheme at high SNRs where we derive the achievable diversity order of both users. We show
through analytically and simulation results that the proposed scheme improves 1) the overall
system throughput, 2) fairness between the two users, and 3) the transmission reliability. This all comes while achieving diversity two for both users, which is the maximal diversity.

In the third part of the thesis, we study relay assignment with limited feedback. In networks with many multiple source-destination pairs, it is normally diffcult for destinations to acquire the channel state information (CSI) of the entire network without feedback. To this end, we design a practical limited feedback strategy in conjunction with two relay assignment schemes, i.e., fullset selection and subset selection, which are based on maximizing
the minimum E2E SNR among all pairs. In this strategy, each destination acquires its SNR,quantizes it, and feeds it back to the relays. The relays then construct the E2E SNR table
and select the relay assignment permutation from all possible relay assignment permutations or only a subset of these permutations. We analyze the performance of these schemes over independent Rayleigh fading channels in terms of the worst E2E SNR. We derive closed-form
expressions for the E2E BER and investigate the asymptotic performance at high SNR. We show that relay assignment with quantized CSI can achieve the same first-order diversity as
that of the full CSI case, but there is a second-order diversity loss. We also demonstrate that increasing the quantization levels yields performance that is close to that of having full knowledge of the CSI.