Toward Green and Resilient Cell-Free Massive MIMO Networks: Radio Deployment Optimization and Resource Orchestration

QC 20260813

Abstract

Mobile networks are evolving into critical societal infrastructure, while the energy footprint, deployment cost, and resilience of the radio access network (RAN) are emerging as the most pressing design constraints of the path toward future generations of mobile networks. Modular network design combined with intelligent resource orchestration holds the solution for these emerging challenges. Cell-free massive multiple-input multiple-output (MIMO) relies on coherent joint transmission by densely distributed low-cost radio units (RUs) and offers fairness among user equipment (UEs). Its distributed nature provides the required modularity for the radio access part, but efficient deployment and network orchestration frameworks are necessary to harness this modularity. To address this gap, this thesis develops deployment and resource-orchestration frameworks for cell-free massive MIMO networks. The work is structured along three research directions.

The first direction addresses efficient radio deployment for cell-free massive MIMO networks. Ray-tracing-based propagation modeling at 28, 39, and 60 GHz characterizes the millimeter-wave channel under different geometries, frame materials, and passenger configurations. Building on this channel model, a joint RU placement and resource allocation framework is formulated as a mixed-integer optimization problem that minimizes the number of deployed RUs while guaranteeing UE rate requirements under access point cooperation and precoding schemes. Numerical results on an airplane-cabin scenario show an 80% reduction in the number of RUs compared to a line-of-sight-based baseline.

The second direction develops an end-to-end power-consumption model and joint orchestration of radio, fronthaul, and cloud-processing resources for cell-free massive MIMO deployed over an open-RAN architecture with both optical and wireless fronthaul. Scenario-sampling-based group-Lasso optimization for centralized precoding and a block-coordinate-descent method for distributed precoding jointly minimize the active number of antennas, radio units, processors, and fronthaul resources. The proposed orchestration achieves up to 70% end-to-end power savings over cloud-only, and 15% over radio-only orchestration. The results reveal that distributing antennas across the coverage area is structurally more energy-efficient than concentrating them at a few sites.

The third direction addresses survivability in cell-free massive MIMO and discusses resilient network design challenges. The analysis demonstrates that dense small-radio deployment is shown to be critical for simple connectivity during disasters, whereas cell-free massive MIMO also enables high capacity in the impacted region.

Link to DiVA