| Unmanned Aerial Vehicles(UAVs)have the advantages of such as small size,low cost and flexible deployment,which have gradually become an important tool in various fields in many countries.In recent years,UAVs have played an increasingly important role in wireless communication.For example,they can be used as mobile space base stations or mobile relays to assist node communication for establishing line-of-sight(Lo S)communication with users.However,due to the broadcast nature of wireless communication,the security of wireless communication systems assisted by UAVs can be seriously threatened for the existence of eavesdroppers.The current UAV information security algorithms are generally fit for the fact that the source and destination nodes are fixed in specific locations.However,these related studies do not consider the cases with mobile nodes.In practical problems,it is essential to study the information security of UAV enabled communication systems with mobile communication nodes in most situations.Therefore,ensuring the secure communication between paired mobile nodes is a challenging and significant topic in the field of wireless communication.Based on the analysis of existing UAV communication information security algorithms,this paper conducts in-depth research on UAV communication security algorithms based on reinforcement learning(RL).And in order to improve the system security performance,we take measures such as introducing cooperation jamming UAV,trajectory planning,power allocation and user scheduling in the UAV relay system with mobile nodes.The main work of this thesis is as follows:The security of UAV mobile relay system and the basic theory of relay secure communication are analyzed and studied.The common algorithms,neural network models and activation functions in RL are introduced.The proximal policy optimization(PPO)algorithm is described in detail and its computational complexity is analyzed.Because PPO algorithm combines the advantages of policy gradient(PG)algorithm and actor-critic(AC)algorithm,with the characteristics of making full use of samples,limiting the update range of the algorithm for ensuring stable convergence.This thesis uses PPO algorithm to solve the problem of maximizing the secrecy rate.When the ground eavesdropper exists and is fixed,we introduce a jamming UAV to interfere with the eavesdropper.Considering multiple pairs of mobile nodes,the relay UAV assists the communication of multiple pairs of nodes in time division duplex mode.Taking the maximization of secrecy rate as the objective function,and combined with power constraints,causal information constraints and other constraints,the UAV information security transmission model is constructed.By transforming it into a Markov process and designing the corresponding action and immediate return,the optimization results are obtained by the PPO based algorithm.The simulation results in different cases show that the proposed algorithm has higher secrecy rate and can obtain better security performance in both small-scale and large-scale communication scenarios.For the presence of mobile eavesdroppers in the air,a secrecy rate maximization algorithm of UAV enabled relay system based on mobile nodes and one mobile eavesdropper is proposed.The relay UAV assists the communication of multiple pairs of mobile nodes through user scheduling.Combined with power allocation,trajectory optimization and user scheduling,the secrecy rate maximization model of UAV relay system is constructed and transformed into a Markov process.By designing the corresponding action and immediate return,the model is solved by PPO algorithm.The simulation results in different cases show that the proposed algorithm can achieve both good fairness and higher average secrecy rate. |
