Secure Transmission for Multi-UAV-Assisted Mobile Edge Computing Based on Reinforcement Learning

UAV communication has received widespread attention in MEC systems due to its high flexibility and line-of-sight transmission. Users can reduce their local computing pressures and computation delay by offloading tasks to the UAV as an edge server. However, the coverage capability of a single UAV is very limited. Moreover, the data offloaded to the UAV will be easily eavesdropped. Thus, in this paper, we propose two secure transmission methods for multi-UAV-assisted mobile edge computing based on the single-agent and multi-agent reinforcement learning, respectively. In the proposed methods, we first utilize the spiral placement algorithm to optimize the deployment of UAVs, which covers all users with the minimum number of UAVs. Then, to reduce the information eavesdropping by a flying eavesdropper, we utilize the reinforcement learning to optimize the secure offloading to maximize the system utility by considering different types of users' tasks with diverse preferences for residual energy of computing equipment and processing delay. Simulation results indicate that compared with the single-agent method and the benchmark, the multi-agent method can optimize the offloading in a better manner and achieve larger system utility.

Automated summary

This paper proposes two secure transmission methods for multi-UAV-assisted mobile edge computing based on single-agent and multi-agent reinforcement learning, respectively. The deployment of UAVs is optimized using the spiral placement algorithm, and secure offloading is optimized using reinforcement learning to maximize system utility. Simulation results show that compared to the single-agent method and the benchmark, the multi-agent method can optimize offloading better and achieve larger system utility.