Thesis Examination Committee

Prof Xiangtong QI, IEDA/HKUST (Chairperson)
Prof Wai Ho MOW, ECE/HKUST (Thesis Supervisor)
Prof Tat Ming LOK, Department of Information Engineering, The Chinese University of Hong Kong (External Examiner)
Prof James SHE, ECE/HKUST
Prof Ling SHI, ECE/HKUST
Prof Cunsheng DING, CSE/HKUST

Abstract

As a feasible way to improve the network throughput, channel coded physical-layer network coding (PNC) is an attractive technology for communications in wireless relay networks.  A desirable decoder for channel coded PNC should satisfy the following properties: low computational complexity, good error performance and high robustness against symbol misalignment. This thesis focus on designing decoding schemes for the channel coded PNC over the two-way relay networks (TWRN).

Firstly, we examine the convolutionally coded PNC (CC-PNC) over the TWRN without a direct end-to-end link.  To enhance decoding performance by considering more than one best candidate codewords, a novel efficient list-output decoder at the relay is proposed. Through combining the tree and the trellis structures of the convolutional code representation, the proposed list decoder achieves a lower computational complexity than the existing list decoders at a comparable error performance. In addition, we present a new way to estimate the error performance of the proposed decoder with different list sizes.

Secondly, we consider the asynchronous CC-PNC over the TWRN in which signals from the two end nodes arriving at the relay are subject to symbol misalignment. A joint channel-network decoding scheme is devised by jointly representing the channel codes, PNC, and symbol misalignment. Unlike the known joint channel-network decoders in the literature, the applicability of the proposed decoder is not limited to channel codes with a cyclic structure. Moreover, the associated bit error rate bound is analyzed and shown to be consistent with simulation results.

Thirdly, we investigate the end-to-end decoding problem for channel coded PNC over the full-duplex (FD) TWRN with a direct end-to-end link. To mitigate the effect of error propagation induced by the decoding error at the relay, two error models, namely, the trellis-error model and the Tanner graph error models, are proposed and convert the end-to-end decoding problem for TWRN into an equivalent point-to-point channel decoding problem. The proposed error models not only provide better error performance comparing with various benchmarking schemes, but also facilitate the asymptotic error performance analysis for channel coded PNC over the FD TWRN.