Abstract:
Next-generation wireless networks are expected to support a massive amount of Internet of Things (IoT) devices in delivering a wide range of novel resource intensive applications. IoT devices, despite their enhanced capabilities, fall short of meeting the computational requirements of these applications within a strict deadline. This thesis leverages a hierarchical computational model where IoT devices offload their computational tasks to a number of unmanned aerial vehicles (UAVs) that can be deployed in a way to establish strong Line-of-Sight (LoS) links. Being equipped with computational resources, UAVs process part of the tasks within the
set time threshold while the other tasks are transferred to edge and cloud servers for potential processing. This work aims at optimizing the number and position of deployed UAVs, IoT-to-UAV association, resource allocation, and task offloading to UAVs, edge servers, and the cloud, while ensuring various system constraints.
The problem is formulated as a mixed integer programming problem and solved using Successive Convex Approximation. An efficient solution is then proposed to decompose the main problem into two subproblems that are solved iteratively. The first subproblem minimizes the number of IoT clusters and positions a UAV to serve each cluster while the second subproblem maximizes the percentage of admitted tasks using the resulting number of UAVs determined by the first subproblem. As long as not all devices are fully served, an additional UAV is introduced and the two
subproblems are solved repeatedly. The proposed decomposition solution has been evaluated as a function of various system parameters and application use cases to show optimized performance and high scalability.
