Towards Resource-Efficient Next-Generation Mobile Networks: Design and Optimization Strategies

Overview & Synopsis

This dissertation provides a synthetic overview of my research activities conducted in the last eight years at the Laboratory of Signal and Systems (L2S, UMR 8506), jointly operated by the CNRS, CentraleSupélec, and the University of Paris-Saclay. The manuscript highlights the principles motivating these research activities and the scientific connection across the various topics. The central common thread of my research activities is resource allocation and management in next-generation mobile networks. On the one hand, the work presents the evolution of Radio Access Network (RAN) architecture, focusing on Cloud-RAN and Open-RAN. These emerging architectures introduce new challenges in how network resources are managed. To address these challenges, we develop efficient and practical techniques for managing radio and computing resources, considering their interdependence. In particular, we study how to schedule these resources in a coordinated way to improve overall network performance. Moreover, we present the benefits of adopting flexible placement of virtual network functions in future RAN architectures, which allows the network to better adapt to varying Quality of Service (QoS) requirements from different users. To implement these techniques, we employ various methodological tools, including mathematical optimization models and machine learning-based solutions. On the other hand, we explore the non-orthogonal Multiple Access (NOMA) technique, which is considered a promising candidate for future generations of mobile networks. We analyze its potential benefits and limitations, particularly in large-scale network scenarios. To address the shortcomings of the NOMA technique in large-scale networks, we propose a tailored variant of NOMA designed to enhance the Quality of Service (QoS) for users located at the cell edge, while also reducing the complexity of successive interference cancellation (SIC) operations that the users located at the cell center must perform. We use different methods to develop and assess our approach, including optimization models and techniques inspired by matching theory.