Large-scale multiple-input multiple-output (MIMO) technique has been recognized as an enabling technology for the sixth generation (6G) network to meet stringent requirements of innovative applications. It is thus essential to determine the fundamental limits of large-scale MIMO systems and investigate the related system design. However, the large number of parameters, complex system, and novel channel conditions cause the performance analysis of large-scale MIMO systems very challenging and very limited results are available in the literature. Fortunately, the high dimensionality makes asymptotic random matrix theory (RMT) a powerful tool for the concerned analysis. In this thesis, we utilize asymptotic RMT to provide rigorous performance analysis for large-scale MIMO systems and inves...[ Read more ]

Large-scale multiple-input multiple-output (MIMO) technique has been recognized as an enabling technology for the sixth generation (6G) network to meet stringent requirements of innovative applications. It is thus essential to determine the fundamental limits of large-scale MIMO systems and investigate the related system design. However, the large number of parameters, complex system, and novel channel conditions cause the performance analysis of large-scale MIMO systems very challenging and very limited results are available in the literature. Fortunately, the high dimensionality makes asymptotic random matrix theory (RMT) a powerful tool for the concerned analysis. In this thesis, we utilize asymptotic RMT to provide rigorous performance analysis for large-scale MIMO systems and investigate the related system design. We will tackle the challenges from several aspects, including the novel (non-Gaussian) fading channels, complex systems (intelligent reflecting surface (IRS)-aided channel), and innovative applications (ultra-reliable and low-latency communications (URLLC), and secure communications).

• Non-Gaussian Fading: We evaluate the ergodic mutual information (EMI) of large-scale MIMO systems with general non-centered non-Gaussian fading. To this end, we derive a closed-form expression for the EMI and show that the bias between the EMI over Gaussian and non-Gaussian fading channels is determined by the pseudo-variance and fourth-order cumulant of the fading. This result is then utilized to refine the outage probability evaluation.

• Two-hop Channels: We explore the fundamental limits of IRS-aided MIMO systems. In particular, we derive the closed-form EMI and outage probability, and propose a gradient-based algorithm to minimize the outage probability by optimizing the phase shifts at the IRS.

• URLLC: We consider the URLLC design for large-scale MIMO systems. For that purpose, we derive the bounds for the optimal average error probability with finite block-length over Rayleigh-product and IRS-aided MIMO channels. Based on the analysis results, we propose a gradient-based algorithm to increase the reliability of IRS-aided MIMO systems, which is shown to be effective in decreasing the error probability of practical coding schemes.

• Physical Layer Security: To unveil the secrecy performance of IRS-aided MIMO systems, we give the closed-form evaluation for the ergodic secrecy rate (ESR) and secrecy outage probability (SOP). The results are then utilized to 1) maximize the artificial noise-aided ESR by jointly designing signal, the artificial noise, and phase shifts at IRSs; and 2) minimize the SOP by optimizing the phase shifts.

This thesis provides an RMT framework for the performance analysis and system design of large-scale MIMO systems. The accuracy of the derived theoretical results and the effectiveness of proposed algorithms are validated by comprehensive simulations. The proposed framework can be utilized to tackle other high-dimensional problems beyond wireless communications, such as theoretical machine learning.