Channel capacity, signaling design, and statistical detection for multiple-antenna channels

This dissertation addresses several key challenges in multiple-antenna communications, including information-theoretical analysis of channel capacity, capacity-achieving signaling design, and practical statistical detection algorithms. The first part of the thesis studies the capacity limits of multiple-input multiple-output (MIMO) multiple access channel (MAC) via virtual representation (VR) model. The VR model captures the physical scattering environment via channel gains in the angular domain, and hence is a realistic MIMO channel model that includes many existing channel models as special cases. This study provides analytical characterization of the optimal input distribution that achieves the sum-capacity of MAC-VR. It also investigates the optimality of beamforming, which is a simple scalar coding strategy desirable in practice. For temporally correlated channels, beamforming codebook designs are proposed that can efficiently exploit channel correlation. The second part of the thesis focuses on statistical detection for time-varying frequency-selective channels. The proposed statistical detectors are developed based on Markov Chain Monte Carlo (MCMC) techniques. The complexity of such detectors grows linearly in system dimensions, which renders them applicable to inter-symbol-interference (ISI) channels with long delay spread, for which the traditional trellis-based detectors fail due to prohibitive complexity. The proposed MCMC detectors provide substantial gain over the de facto turbo minimum-mean square-error (MMSE) detector for both synthetic channel and underwater acoustic (UWA) channels. The effectiveness of the proposed MCMC detectors is successfully validated through experimental data collected from naval at-sea experiments.