Abstract

Empowered by recent technological advances and driven by the ever-growing population density and needs, the conception of 5G has opened up the expectations of what mobile networks are capable of to heights never seen before, promising to unleash a myriad of new business practices and paving the way for a surging number of user equipments to carry out novel service operations. The advent of 5G and networks beyond will hence enable the vision of Internet of Things (IoT) and smart city with its ubiquitous and heterogeneous use cases belonging to various verticals operating on a common underlying infrastructure, such as smart healthcare, autonomous driving, and smart manufacturing, while imposing extreme unprecedented Quality of Service (QoS) requirements in terms of latency and reliability among others. Due to the necessity of those modern services such as traffic coordination, industrial processes, and mission critical applications to perform heavy workload computations on the collected input, IoT devices such as cameras, sensors, and Cyber-Physical Systems (CPSs), which have limited energy and processing capabilities are put under an unusual strain to seamlessly carry out the required service computations. While offloading the devices' workload to cloud data centers with Mobile Cloud Computing (MCC) remains a possible alternative which also brings about a high computation reliability, the latency incurred from this approach would prevent from satisfying the services' QoS requirements, in addition to elevating the load in the network core and backhaul, rendering MCC an inadequate solution for handling the 5G services' required computations. In light of this development, Multi-access Edge Computing (MEC) has been proposed as a cutting edge technology for realizing a low-latency computation offloading by bringing the cloud to the vicinity of end-user devices as processing units co-located within base stations leveraging the virtualization technique. Although it promises to satisfy the stringent latency service requirements, realizing the edge-cloud solution is coupled with various challenges, such as the edge servers' restricted capacity, their reduced processing reliability, the IoT devices' limited offloading energy, the wireless offloading channels' often weak quality, the difficulty to adapt to dynamic environment changes and to under-served networks, and the Network Operators (NOs)' cost-efficiency concerns. In light of those conditions, the NOs are consequently looking to devise efficient innovative computation offloading schemes through leveraging novel technologies and architectures for guaranteeing the seamless provisioning of modern services with their stringent latency and reliability QoS requirements, while ensuring the effective utilization of the various network and devices' available resources. Leveraging a hierarchical arrangement of MEC with second-tier edge servers co-located within aggregation nodes and macro-cells can expand the edge network's capability, while utilizing Unmanned Aerial Vehicles (UAVs) to provision the MEC service via UAV-mounted cloudlets can increase the availability, flexibility, and scalability of the computation offloading solution. Moreover, aiding the MEC system with UAVs and Intelligent Reflecting Surfaces (IRSs) can improve the computation offloading performance by enhancing the wireless communication channels' conditions. By effectively leveraging those novel technologies while tackling their challenges, the edge-cloud paradigm will bring about a tremendous advancement to 5G networks and beyond, opening the door to enabling all sorts of modern and futuristic services.

In this dissertation, we attempt to address key challenges linked to realizing the vision of a low-latency and high-reliability edge computation offloading in modern networks while exploring the aid of multiple 5G network technologies. Towards that end, we provide novel contributions related to the allocation of network and devices' resources as well as the optimization of other offloading parameters, and thereby efficiently utilizing the underlying infrastructure such as to enable energy and cost-efficient computation offloading schemes, by leveraging several customized solutions and optimization techniques. In particular, we first tackle the computation offloading problem considering a multi-tier MEC with a deployed second-tier edge-cloud, where we optimize its use through proposed low-complexity algorithms, such as to achieve an energy and cost-efficient solution that guarantees the services' latency requirements. Due to the significant advantage of operating MEC in heterogeneous networks, we extend the scenario to a network of small-cells with the second-tier edge server being co-located within the macro-cell which can be reached through a wireless backhaul, where we optimize the macro-cell server use along with the other offloading parameters through a proposed customized algorithm based on the Successive Convex Approximation (SCA) technique. Then, given the UAVs' considerable ability in expanding the capabilities of cellular networks and MEC systems, we study the latency and reliability aware optimized positioning and use of UAV-mounted cloudlets for computation offloading through two planning and operational problems while considering tasks redundancy, and propose customized solutions for solving those problems. Finally, given the IRSs' ability to also enhance the channel conditions through the tuning of their passive reflecting elements, we extend the latency and reliability aware study to a scenario of an IRS-aided MEC system considering both a single-user and multi-user OFDMA cases, where we explore the optimized IRSs' use in order to reveal their role in reducing the UEs' offloading consumption energy and saving the network resources, through proposed customized solutions based on the SCA approach and the SDR technique.