Advanced Techniques for Cell-free Massive MIMO Networks

Abstract

To further address the increasing data demands, support emerging applications, enhance user experiences, and achieve sustainable communication goals [1], [2], beyond-5G communication plays a crucial role in bridging 5G and 6G communications. Among numerous innovative technologies in beyond-5G communication, cell-free massive multiple-input multiple-output (MIMO) has emerged as a promising solution for next-generation network architectures, leveraging the macro-diversity gains provided by a large number of distributed access points (APs). This thesis investigates the fundamental benefits of cell-free massive MIMO compared to the cellular architecture while providing solutions for three issues raised by cell-free massive MIMO networks, including the pilot assignment, uplink data quantisation, and the application of reconfigurable intelligent surface (RIS) in cell-free massive MIMO. Mathematical analysis is adopted to model different stages of a communication system, and the performance of proposed algorithms is evaluated through numerical simulations.

The first part of the thesis focuses on pilot assignment in a user-centric cell-free network to reduce pilot contamination caused by pilot-sharing user equipment (UEs). A linear minimum mean square error (LMMSE) estimator is adopted for channel estimation. Then, the normalised mean square error (NMSE) is derived to evaluate the accuracy of the channel estimate. Inspired by the NMSE expression, the relation of UEs' channel correlation matrices is considered while deciding the set of UEs sharing the same pilot sequence. With the proposed method, UEs with relatively different eigenspaces are assigned to the same pilot, and the level of pilot contamination is confined.

The second part addresses the problem of data quantisation when APs need to forward received uplink signals to the central processing unit (CPU) through realistic fronthaul links. It starts with introducing a lattice-based vector quantisation scheme and Bussgang decomposition for modelling the non-linear quantisation distortion. To improve the quantitation efficiency for a given rate of fronthaul links, the proposed method adjusts the shape of the codebook according to the distribution of the input signals of the quantiser. The quantisation efficiency is effectively enhanced by exploiting the local spatial correlation at each AP.

The final part focuses on applying RIS in a multi-user cell-free massive MIMO network. It starts with a RIS splitting scheme for multiple UEs served by the same surface. To increase the degrees of freedom for inter-user interference suppression, the proposed method exploits the spatial diversity of the distributed MIMO network and configures RIS elements so that different UEs' signals are re-radiated to different groups of APs. Short-term and long-term phase-shift configurations are proposed. Different channel estimation strategies are needed for different configuration schemes, leading to varying levels of overhead and distinct system performances.

Date of Award	13 May 2025
Original language	English
Awarding Institution	University of Bristol
Supervisor	Angela Doufexi (Supervisor) & Mark A Beach (Supervisor)