Optimum group detection for spatial multiplexing systems

Abstract

Large MU-MIMO (Multi-User Multiple-Input Multiple-Output) systems are advanced wireless communication technologies that support a large number of antennas at both the transmitter and receiver ends. These systems provide spatial multiplexing, diversity, and beamforming, significantly enhancing wireless communication’s capacity and reliability. As a result, large MU-MIMO systems have become an essential component of 5G and beyond-5G wireless networks, meeting the increasing demands for higher data rates. The load factor refers to the ratio of the number of user equipment (UEs) to the number of antennas at the base station (BS). When the load factor approaches zero, the system achieves a favorable propagation condition that offers significant diversity gain and reliability. In this scenario, linear receivers such as zero-forcing (ZF) and minimum mean square error (MMSE) perform near-optimal. In contrast, the system maximizes the multiplexing gain and capacity when the load factor equals one, called the full-load factor. However, this scenario results in a non-favorable propagation condition, severely degrading the linear receivers’ performance and making them sub-optimal algorithms. Maximum-likelihood group detection (GD-ML) receiver is an algorithm that improves the performance of linear receivers without significantly increasing their complexity. The GD-ML technique involves dividing the received symbols vector into groups after applying a linear projection. The optimal Maximum-likelihood (ML) detection is applied to each group. While this technique has been researched for conventional multiple-input multiple-output (MIMO) systems with uncorrelated Rayleigh channels, its potential for large MU-MIMO systems remains largely unexplored. In this thesis, we aimed to achieve two primary objectives related to the effectiveness of the GD-ML receiver in large MU-MIMO systems with full-load factor and correlated Rayleigh channel. The first objective involves obtaining the equation for the receiver’s complexity and assessing the compromise between performance and complexity. The second objective is to derive an analytical expression for GD-ML receiver performance. The aforementioned objectives were accomplished using the following methodology: a literature review, which encompassed relevant works and background information; creation of a system model that assumes a wireless cellular infrastructure with M distributed and uncorrelated antennas at the BS, N closely-located single-antenna and correlated UEs, and a GD-ML receiver with group size Nu, (N _ Nu); evaluation metrics that included bit error rate (BER) and vector error rate (VER) as performance metrics, along with the floating-point operations (FLOPs) metric for computational complexity; mathematical analysis which involved the formulation of analytical equations to assess the receiver’s performance and complexity using multivariate random theory, stochastic ordering, and matrix operations. We provided the FLOPs equation to evaluate the GD-ML receiver’s complexity. We observed that the GD-ML algorithm has almost the same complexity as ZF and MMSE, where the ML detection and grouping stages add a negligible complexity compared to linear projection operation. We derived a closed-form expression for the average group VER to assess the GD-ML receiver’s performance. Our analytical results indicated that the GD-ML receiver provides a diversity gain proportional to M − N + Nu. We also found that the GD-ML receiver’s performance decreases as the UEs’ correlation coefficients increase. Numerical results revealed that the GD-ML receiver outperforms both ZF and MMSE receivers, and validated the derived performance expression. We observed that the analytical expression and numerical outcomes remain close for small Nu. At a moderate signal-to-noise ratio (SNR), we observed that the analytical expression and simulation results closely match and become perfect as the UEs’ correlation increases.

Date	4 Oct 2023
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Georges Kaddoum (Supervisor) & François Gagnon (Co-supervisor)