Multi-UAV-Assisted Offloading for Joint Optimization of Energy Consumption and Latency in Mobile Edge Computing

To address the performance limitations caused by the insufficient computing capacity and energy of edge internet of things devices (IoTDs), we proposed a multi-unmanned aerial vehicles (UAV)-assisted mobile edge computing (MEC) application framework in this article. In this framework, UAVs equipped with high-performance computing devices act as aerial servers deployed in the target area to support data offloading and task computing for IoTDs. We formulated an optimization problem to jointly optimize the connection scheduling, computing resource allocation, and UAVs' flying trajectories, considering the device offloading priority, to achieve a joint optimization of energy consumption and latency for all IoTDs during a given time period. Subsequently, to address this problem, we employed deep reinforcement learning for dynamic trajectory planning, supplemented by optimization theory and heuristic algorithm based on matching theory to assist in solving connection scheduling and computing resource allocation. To evaluate the performance of proposed algorithm, we compared it with deep deterministic policy gradient, particle swarm optimization, random moving, and local execution schemes. Simulation results demonstrated that the multi-UAV-assisted MEC significantly reduces the computing cost of IoTDs. Moreover, our proposed solution exhibited effectiveness in terms of convergence and optimization of computing costs compared to other benchmark schemes.