Offloading strategy for UAV-assisted mobile edge computing based on reinforcement learning

Abstract

With the rapid development of mobile communication technology, mobile edge computing (MEC) is increasingly becoming an effective technology solution to the excessive pressure of traffic data. That is, by offloading the computing tasks to the edge server for computing, the delay-sensitive tasks can effectively meet the requirements of ultra-low latency. However, in the actual scenario, most edge servers were deployed on fixed base stations, which can not provide timely and effective services in the face of natural disasters such as earthquakes and fires that cause damage to fixed base stations and prevent communication. With the advantages of low cost and high mobility, Unmanned Aerial Vehicle (UAV) can be considered to offload computational tasks to UAV for better meet some of the unexpected communication needs. To make better use of the UAV, this paper constructs an independently divisible task model, where the user equipment (UE) can choose to offload part of the task to the UAV for computation and the other part through local computation. By jointly optimizing the resource scheduling of UAVs, the offloading ratio of tasks, and the Proximal Policy Optimization(PPO) based on reinforcement learning, the unloading strategy was optimized. Numerous experiments have proven that the algorithm has great advantages in terms of convergence speed and training speed. At the same time, compared with the baseline method, the algorithm is more generalized, which means that the learning ability is stronger, and the delay can still be effectively reduced in the face of environmental changes.