Maximizing the Average Secrecy Rate for UAV-assisted MEC: A DRL Method

In the unmanned aerial vehicle (UAV)-assisted mobile edge computing (MEC) system where eavesdropper existed, UAV was faced with the problem of secure communication when served users, so this paper proposed a security algorithm based on deep reinforcement learning which enabled UAV to find the optimal flight strategy to maximize the average secrecy rate of serving users. Firstly, the process of maximizing the average secrecy rate was modeled as a Markov decision process (MDP) without transition probability, and states, actions and reward function of UAV were defined. Secondly, UAV used the proposed algorithm to change its position through RL online learning and deep neural network offline training to find the optimal flight strategy in a battery cycle so as to maximize the average secrecy rate. Finally, the proposed algorithm was compared with the traditional algorithms, and the simulation results show that the proposed algorithm can effectively improve the average secrecy rate when UAV is serving users and has faster convergence rate than Q-Learning algorithm.