Examination Committee

Prof Andrew B HORNER, CSE/HKUST (Chairperson)
Prof Wai Ho MOW, ECE/HKUST (Thesis Supervisor)
Prof Chi-Sing LEUNG, Department of Electronic Engineering, City University of Hong Kong (External Examiner)
Prof Weichuan YU, ECE/HKUST
Prof Ling SHI, ECE/HKUST
Prof Cunsheng DING, CSE/HKUST
 

Abstract

Relay-assisted communication has been shown to improve the reliability and throughput of real-world wireless communication networks. In this thesis, we re-examine the decoding problem of the coded decode-and-forward (DF) relay channels. The error-free decoding at the relay is typically assumed by the conventional decoder at the destination. However, unsuccessful decoding at the relay can cause error propagation especially when the quality of the source-to-relay link is poor.

We first propose a near BER-optimal decoding (NBOD) algorithm at the destination for convolutionally coded relay channels. Its decoding complexity is linear in the information block length, while the exact BER-optimal decoding algorithm with a sub-exponential complexity is still unknown. The only major approximation involved in NBOD is the pairwise error probability approximation. The NBOD can perform close to the near maximum likelihood decoding (MLD) performance bound on BER in the three-node DF relay channel. Inspired by NBOD, we propose a trellis error model in this scenario. Moreover, we extend NBOD for more general single-source single-destination DF relay networks.

We further study the decoding problem for distributed Turbo codes (DTC). Accordingly, two trellis-error-model-based iterative decoders are derived. Moreover, we propose a near MLD performance bound on BER of the DTC. The extrinsic information transfer chart analysis is then performed to characterize the asymptotic gain achieved by our decoders for large code lengths. In addition, we extend the proposed decoders to general distributed concatenated coding systems with DF-based relays.

Besides, we study the decoding problem for the LDPC-coded relay channels. We propose a Tanner graph error model to represent the relay decoding errors and characterize their statistics. A Tanner-graph-error-model-based message passing decoder is developed. Furthermore, the error model can transfer the relay channel as a point-to-point channel in a more accurate way. The proposed decoder can be applied to the convolutionally coded relay channels to realize NBOD such that the complexity is independent of the number of states.

Finally, we examine the decoding problem of the general two-way relay channel with binary network coding. We consider the practical convolutional codes and LDPC codes, which are included in the wireless Local Area Network standard (i.e., IEEE802.11n). Simulation results show the superiority of our decoders based on error models.