Control methods and applicative scenarios for next-generation cellular telecommunication networks

The increasing demand for connectivity services to the mobile (cellular) network, together with the co-existence of former terrestrial mobile networks and the convergence of satellite telecommunication systems, lead the telecommunication operators to find solutions for enabling such convergence, by also making use of former cellular networks present on the field. Moreover, the introduction of next-generation cellular telecommunication systems leads to the need for telco operators to find new industrial use cases, which may enable new kinds of services both at the industrial and at the network level. This thesis discusses some control methodologies to be applied to facilitate the convergence of terrestrial and non-terrestrial networks in a multi-RAT environment, where different Radio Access Technologies are available at the same time, providing multi-connectivity services at increasing bandwidth and reduced latency, by also considering the users' perceived QoE. Such control techniques are mainly model-free and are based on Game Theory arguments and on Reinforcement Learning, and address two different problems: the network selection (i.e., deciding the best AP to serve a UE request) and the dynamic traffic splitting and steering (i.e., in a multi-connectivity context, the problem of deciding the quantity of traffic of each UE to be sent to each of the connected AP). Moreover, an applicative scenario of 5G network for smart grid control, and in particular for the provisioning of Frequency Regulation services using charging \acp{PEV} has been proposed in this thesis, by analyzing the regulatory framework, the technical feasibility, and the economic feasibility of the proposed approach. All the proposed approaches are provided with extensive simulations to validate the concepts and the proposed control algorithms. In addition, a 5G multi-\ac{RAT} radio access network simulator has been developed in the context of the work carried out by the Candidate for this thesis, in order to validate the proposed approaches in a realistic environment. Some of the proposed algorithms have been / will be also tested in real environments in the context of the activities of the H2020 project 5G-ALLSTAR and 5G-Solutions, which partially supported the ideas, algorithms, and results proposed in this thesis.