Master thesis : Cell-Free Massive MIMO with Dynamic Metasurface Antennas: Uplink Resource Allocation under Max-Min Fairness

Abstract

Cell-Free massive MIMO is a promising architecture for future wireless networks. Instead of relying on a single base station, it distributes many Access Points (APs) across the coverage area and serves all users jointly, eliminating the coverage problems that arise at cell edges. Deploying cell-free, however, requires a massive number of APs, each equipped with multiple antennas, which leads to considerable hardware complexity and energy consumption. In particular, conventional fully-digital architectures, which assign one Radio-Frequency (RF) chain per antenna element, quickly become prohibitively costly and power-hungry.

Dynamic Metasurface Antennas (DMAs) directly address this problem. A DMA performs analog beamforming through tunable metamaterial elements embedded in waveguides. A single RF chain can drive a large number of radiating elements simultaneously. This significantly reduces both hardware complexity and power consumption compared to fully-digital or hybrid architectures, making DMAs well suited for large-scale deployments.

This thesis investigates the use of DMAs as AP hardware in a Cell-Free massive MIMO network. The first part builds the necessary background: the electromagnetic behavior of DMA elements is characterized through the Lorentzian resonance model, which inherently couples the phase and amplitude of each element and constitutes the central hardware constraint of this work.

The second part addresses the uplink problem under a max-min Signal-to-Interference-plus- Noise Ratio (SINR) fairness criterion for DMA-enhanced Cell-Free MIMO, ensuring uniform Quality of Service (QoS) across all users. The resulting optimization problem involves three coupled sets of variables: digital combining vectors, per-user transmit powers, and DMA weight matrices. To solve this non-convex problem, an alternating optimization framework is developed. Within this framework, two algorithms are proposed for updating the DMA weights: one based on Semidefinite relaxation (SDR) and another based on gradient-ascent (GA) over the Lorentzian phase parameters.

Simulations in an Indoor Factory environment show that both algorithms achieve a 6–8 dB gain over an unoptimized baseline and stay within 2.5–4 dB of the fully-digital upper bound. Distributing antennas across multiple access points consistently improves the worst-user SINR, confirming the macro-diversity benefit of the Cell-Free architecture. Overall, these results demonstrate that DMA-based Cell-Free Massive MIMO constitutes a practical, cost-effective, and energy-efficient solution for next-generation industrial wireless systems.