Titolo della tesi: Dynamic optimization of communication, computation and caching for UAV-aided edge computing
The work in this thesis focuses on the optimization of communication, computation and caching resource in the context of edge computing, with a particular emphasis on the use of aerial edge servers to enable the control over flexible and heterogeneous networks.
The emerging 5G represents a revolution in communication networks providing a common infrastructure to enable very heterogeneous services. Compared with previous generations, 5G will support not only communication, but also computation, control, and content delivery functions and will face an unprecedented increase in traffic volume and computation demands. To handle and process such a huge amount of data traffic, there is the need to bring cloud functionalities close to mobile users and then requiring to have radio access points as close as possible to the end user. In this thesis, we investigate possible synergies between communication, computation and storage, in the context of \textit{edge computing}. In the first part, we propose a novel dynamic resource allocation strategy for computation offloading in the context of UAV-aided wireless networks, to jointly orchestrate radio and computation resources. Starting from a long-term optimisation problem, we hinge on stochastic Lyapunov optimisation to solve it in a per-slot basis, with low complexity solutions for the instantaneous optimisation problem. The opportunity to offload sophisticated applications and intensive task algorithms is really crucial for resource-hungry devices, as well as the opportunity to do it in a more flexible way, for these reasons we exploit the use of aerial platforms, such as Unmanned Aerial Vehicles, acting as mobile edge servers. In this thesis, we deploy these aerial base stations for computation offloading, enabling the control of their position with the aim at minimising the global energy consumption of the system, intended as the energy spent to communicate in both directions (uplink and downlink) and the energy spent only by the server to process data and to fly. We use a Stochastic Approximation algorithm, based on the channel characteristics deeply studied in this thesis, to track the evolution of the environment and select the optimal altitude with respect to the connection link and the energy consumption. Our work aims at leverage the key possibilities in terms of high mobility, flexibility and low costs of aerial vehicles, together with the powerful tool of the Lyapunov framework to enable the edge computing. The second part of the thesis focuses on the edge caching, enabled following two different approach. We first propose an optimization algorithm aiming at balancing the cost of the energy spent to store contents at the edge and the cost to transport them through the network. We propose a novel popularity profile to rank the contents to be proactively cached at edge nodes and to measure the centrality of the best nodes to store the contents. Then, we handle the service placement and request routing problem with an optimization algorithm aiming at minimizing the total delay experienced to deliver the requested network services, under storage and computation capacity constraints. The thesis investigate the storage resource optimization to run sophisticated applications at edge clouds, managing at the same time the data offloaded by querying devices and the computation resource available at the servers.