Research On Physical Layer Security Of UAV Communications Based On Reinforcement Learning

Physical layer security is a key technology to achieve wireless communication security,and can also be used to achieve information transmission security of UAV communication.Using UAVs as mobile base stations can achieve more stable and secure information transmission for long-distance communications,at the same time it can improve the communication quality and coverage of the wireless communication system.However,when the UAV base station communicates with the legitimate receivers on the ground in the presence of a potential eavesdropper,due to the open characteristics of wireless communication,the communication quality is improved,and the intensity of information received by the eavesdropper is also enhanced,which increases the security risks in the communication process.Compared with the traditional security technology to encrypt information,the physical-layer security technology makes use of the physical characteristics of wireless communication channels and provides a technical scheme to realize secure communication.In the existing research,the flight trajectory optimization of the UAV generally adopts an offline method so that the optimized trajectory obtained is fixed,which leads to the lack of the ability of secure transmission to cope with the change of communication environment and limits the communication performance of the system.To solve this problem,this thesis proposes an online flight path optimization strategy for the UAV base station communication system based on reinforcement learning,so that the UAV base station can re-plan the flight path according to the environmental changes,and realize the safety and reliability of the transmission of confidential information by the UAV base station.According to the different number of UAV base stations required by the system model,this thesis studies the two situations of single UAV base station communication and multi-UAV base station communication respectively.1.Research on the optimization of flight path for single UAV base station communication system.In this system,the UAV base station transmits confidential information to a legitimate receiver in the presence of an eavesdropper on the ground,and the confidential information can be safely transmitted through online optimization of the flight path of UAVs,and the communication average secrecy rate during the flight of UAVs is maximized.Specifically,the UAV uses the instantaneous secrecy rate of the location in the flight area to guide the selection of different flight directions and finally converges to the optimal trajectory to achieve the optimization goal of maximizing the average secrecy rate.Before the UAV takes off,because the communication environment is unknown and unpredictable,it is impossible to directly solve the optimal flight trajectory like an offline algorithm.To solve this problem,this thesis adopts the Temporal-Difference learning in reinforcement learning,utilizing the autonomous interaction of the UAV and the communication environment to strive for a balance between exploration and exploitation,and finally,finds an optimal flight trajectory.Simulation results show that,compared to the benchmark algorithm,the proposed algorithm can effectively improve the average security rate,and achieve the security of the wireless communication system.2.Research on the optimization of flight path and power of a multi-UAV base station communication system.In the case of large communication range and multiple ground receivers,the scheme proposed above cannot meet the communication quality requirements.Therefore,this thesis also considers maximizing the average security rate of UAVs by jointly optimizing the flight trajectory and power of multiple UAVs.The communication environment is unknown to the UAVs,so the reinforcement learning algorithm is adopted to enable the UAVs to interact autonomously with the environment,and finally the optimized flight trajectory can be obtained to achieve the communication security and reliability of the wireless system.According to the simulation results,compared with the benchmark algorithms,the joint optimization algorithm proposed in the thesis is feasible and more effective.