Design and Analysis of Multilevel Coded Modulation for Multi-Node Networks

Abstract

During the last several decades, information theory has made significant advances in the analysis of the limits of communication in multi-node networks and the methods that can approach those limits. This dissertation studies new multi-level architectures for coded modulation in multi-node networks that aim to approach, in the practical realm, the capacity limits unveiled by information theory.

For the two-user additive white Gaussian noise (AWGN) broadcast channel, a multi-level coding architecture is proposed whose performance can approach the entire capacity region, and whose attractive features include a convenient partition of the two users’ data so that one and only one of the modulation bit levels (and the corresponding encoder) must contend with both users’ data. Practical aspects of the problem, including allocation of levels to users and finding level-wise code rates, have been addressed.

For the full-duplex decode-forward relay channel, a pragmatic yet capacity-approaching construction is proposed that synthesizes the components of full-duplex transmission via distinct signal levels of a multi-level code at the source and at the relay. The rate penalty due to linearity of component codes is analyzed and to avoid it, a solution is proposed involving the labeling of signal constellations. Simulations show that the proposed architecture together with good point-to-point codes can achieve excellent performance.

For the full-duplex decode-compress-forward relay channel, a multi-level coding architecture is proposed and analyzed that achieves rates very close to the best known (information theoretic) achievable rates. The performance of the proposed architecture is evaluated using a combination of low-density parity-check (LDPC) codes and polar codes.

For the design of coded modulation for the discrete-input, Gaussian noise wiretap channels, a rate splitting method is proposed to allow a convenient construction of wiretap channel codes via a combination of two separate encoders operating on the data and the dither components. This technique leads naturally to the construction of multilevel codes for the AWGN wiretap channel where the message and the dither are encoded through separate levels without compromising secrecy. The effect of maximum likelihood decoding and multistage decoding at the legitimate receiver, as well as the effect of modulation labeling, are studied.