Joint Optimization of Computing Offloading in Multi-UAVs-Assisted MEC System

Abstract

Mobile edge computing (MEC) has been developed as a promising technology to extend various services to the edge of Internet of things. However, it is difficult to deploy MEC devices in special scenarios. Inspired by the high flexibility and controllability of unmanned aerial vehicle (UAV), a multi-UAVs-assisted "cloud-edge integration" network architecture for offloading computing intensive tasks in terminal devices is proposed. UAVs can provide computing resources for the users at the edge of the network. Based on this architecture, the computational offloading problem was formulated as a mixed integer nonlinear programming problem, which is usually difficult to get the optimal solution. Therefore, an efficient computing offload algorithm based on deep reinforcement learning (ISDRL) is proposed to obtain the best computing offloading and resource allocation strategy. The numerical results demonstrated that the proposed offloading algorithm has more advantages. In addition, compared with the traditional cloud architecture, the proposed network architecture is more suitable for complex scenarios.