A Reliability-aware Computation Offloading Solution via UAV-mounted Cloudlets

Abstract

Multi-access Edge Computing (MEC) has enabled low-latency computation offloading for provisioning latency-sensitive 5G services that may also require stringent reliability. Given the growing user demands incurring communication bottleneck in the access network, Unmanned Aerial Vehicles (UAVs) have been proposed to provide edge computation capability, through mounting them by cloudlets, hence, harnessing their various advantages such as flexibility, low-cost, and line of sight communication. However, the introduction of UAV-mounted cloudlets necessitates a novel study of the provisioned reliability while accounting for the high failure rate of UAV-mounted cloudlets, that can be caused by various factors. In this paper, we study the problem of reliability-aware computation offloading in a UAV-enabled MEC system. We aim at maximizing the number of served offloading requests, by optimizing the UAVs' positions, users' task partitioning and assignment, as well as the allocation of radio and computational resources. We formulate the problem as a non-convex mixed-integer program, and due to its complexity, we transform it into an approximate convex program and provide a low-complexity iterative algorithm based on the Successive Convex Approximation (SCA) method. Through numerical analysis, we demonstrate the efficiency of our solution, and study the achieved performance gains for various latency and reliability requirements corresponding to different use cases in 5G networks.