Dynamic UAV Routing for Multi-Access Edge Computing

Multi-access Edge Computing (MEC) is a paradigm in which the computation of traffic and services is moved from a centralized cloud to the edge of the network and closer to the customer. This paradigm is expected to effectively reduce latency, making it suitable for real-time applications. A key technology for this paradigm is the use of unmanned aerial vehicles (UAVs) to mount edge servers that meet customers' computing requirements. The use of UAVs in MEC is attractive because they are close to customers and can be deployed dynamically. In real-time applications, users update the terrestrial network with their locations and computing needs by sending real-time requests. Therefore, from the terrestrial network's perspective, both the locations and arrival times of tasks are randomly distributed. It is crucial to manage the UAV efficiently to maximize the number of computational tasks completed during its flight time, as UAVs have limitations in terms of energy, power and speed. Due to the random nature of the arrival of tasks, existing studies address this problem with the traditional approach of using machine learning (ML). This approach suffers from two drawbacks: 1) ML algorithms are computationally intensive and 2) they are heuristic methods that offer no guarantees of optimality. In this paper, we present an alternative approach for UAV management. We propose an algorithm for UAV routing and task scheduling based on Lyapunov optimization that has a low computational cost while considering energy consumption and path stability. We also developed Theorem 1, which provides upper bounds on the long-term performance of the UAV. Numerical results based on realistic UAV parameters show the feasibility of our proposed algorithm and its advantage over ML algorithms. We have also investigated the conditions that can further improve the performance of task offloading in terms of uplink power, UAV speed, and CPU resources.