Cooperative task offloading and resource allocation for UAV-enabled mobile edge computing systems

Abstract

Unmanned aerial vehicle (UAV)-enabled mobile edge computing (MEC) systems can provide flexible computation services to user equipments (UEs). On the one hand, the UAV can be flexibly deployed to provide computation services to UEs in remote areas or where intensive computing is required. On the other hand, the UAV can approach UEs conveniently to enhance the offloading performance. However, when UEs are widely distributed, the UEs will consume high energy to offload tasks to the UAV, or the UAV will need to consume significant energy to fly close to different UEs to ensure reliable computation offloading, which is unfriendly to the energy-constrained UAV. To tackle these issues, in this paper, we propose a cooperative task offloading scheme for the UAV-enabled MEC systems. Specifically, we consider a UAV-enabled MEC system consisting of a UAV serving as a MEC server for multiple near and far UEs, where each UE has a dividable computation task to be computed and can offload a part of the task to the UAV for computing. The task offloading of each far UE is proposed to be assisted by an associated near UE, where each far UE first sends its task to the associated near UE, and then the near UE offloads its own task and the task from the far UE to the UAV. An iterative algorithm based on the block coordinate descent method is proposed to optimize the UAV’s trajectory, the computation and communication resources for minimizing the weighted sum energy consumption of the UEs and the UAV. Specifically, the UAV’s trajectory is optimized based on the successive convex approximation method, and the computation and communication resources are optimized via the Lagrangian dual method. Simulation results are presented to verify the effectiveness of the proposed algorithm. It is shown that compared to the state-of-the-art algorithms in existing literature, the proposed algorithm achieves much lower energy consumption, especially when the UEs carry more task data.