Towards sustainable cell-free massive MIMO : initial access and pilot assignment

Abstract

Cell-free massive multiple input multiple output (MIMO) is a key enabling technology for future wireless networks, promising high data rates and improved reliability. By eliminating the concept of cell boundaries, it addresses the traditional cell-edge problem, characterized by low received power and high interference, thereby enabling a more uniform quality of service across the network. In these systems, a large number of distributed access points (APs) cooperate to jointly serve the users, with each user being served by a subset of nearby APs. This distributed and cooperative architecture is fundamental to the performance gains of cell-free systems. However, the dense deployment of APs and the cooperation required to realize these benefits introduces significant challenges, primarily in terms of system complexity, coordination overhead and energy consumption. These factors present major obstacles to the development of practical and sustainable cell-free massive MIMO implementations. The primary goal of this dissertation is to address the critical challenges of high complexity and energy consumption that hinder the practical deployment of sustainable cell-free massive MIMO systems. This is achieved by developing novel, low-complexity algorithms for two fundamental network procedures: initial access and pilot assignment.To understand the practical challenges of interference and network management, an in-depth analysis of a commercial network deployment is presented. This analysis is based on a large-scale measurement campaign conducted across diverse urban and suburban environments in Vienna,capturing both cell-level long term evolution (LTE) and beam-level fifth generation (5G) data.The cell-level analysis quantifies the severe impact of inter-cell interference in traditional cellular systems, highlighting the need for cooperative approaches like cell-free MIMO. The more granular 5G beam-level analysis reveals key spatial characteristics of the radio environment, such as the stability of beam connections and, most importantly, the high concentration of received signal power within a small subset of dominant beams. These real-world observations provide thenecessary basis and motivation for the development of data-driven, spatially aware frameworks.Guided by these real-world insights, the thesis addresses the challenges of initial access. It proposes a novel, two-category AP architecture framework that divides the network into a sparseset of always-on signaling APs and a pool of on-demand data APs. This approach reducespower consumption and mitigates interference from physical cell identity (PCI) conflicts. The framework is enabled by a data-driven method that uses GSP to learn an AP neighborhood graph from user-reported measurements. The learned graph enables an advanced graph-coloring algorithm for low-conflict PCI planning and intelligent AP sampling strategies to select active signaling APs.For the pilot assignment, the thesis introduces innovative, low-complexity schemes to mitigate pilot contamination. First, an AP clustering-based scheme is developed that decentralizes the assignment process by grouping users based on their serving APs’ cluster identity. Second, a novel, beam-domain-based scheme is introduced that leverages spatial information from 5G beam management to construct a highly accurate interference graph. This improves not only pilot assignment but also enables dynamic RF chain switching for further energy savings.