Research On Antenna Selection Based On Reinforcement Learning In Wireless Communication

With the rapidly development of the fifth-generation(5G)wireless communication technology,the information security has become an issue of great concern.On the one hand,due to the broadcast nature of multimedia in wireless communication networks,some unauthorized devices can also receive codewords information transmitted by the source nodes,which greatly limits the secrecy performance of communication systems.Meanwhile,with the increase of terminal equipment and the complexity of the network structure,people have also put forward the other higher requirements for the transmission rate and service quality of mobile communications.Therefore,ensuring the secure information transmission and improving the reliability of information transmission become very important in the physical layer secure communication network.On the other hand,the battery-powered wireless devices have high maintenance costs and cannot meet the energy requirements of communications.For this reason,scholars proposed a green communication technology that harvests energy from the radio frequency signals,which greatly improves the working efficiency.However,as the number of network terminals increases and the network structure becomes more complicated,the performance analysis of the traditional wireless communication systems based on mathematical analysis becomes very complicated.In order to overcome this research difficulty,scholars have tried to investigate the performance analysis by combining machine learning algorithms and wireless communication systems,and have achieved insightful results.In recent years,the unsupervised reinforcement learning has become another research hotspot for scholars in the field of wireless communication to explore the potential advantages of machine learning algorithms.Reinforcement learning algorithms are exploited to train the wireless communication environment and allow interactive learning among the terminal nodes in the system,so that the wireless communication system can adaptively obtain an optimal transmission strategy.In addition,the use of reinforcement learning algorithms to train thecommunication environment has low complexity,and there is no need to know the wireless communication system model(such as the satellite communication,and unmanned aerial vehicle wireless communication also have good adaptability).This paper mainly considers three wireless communication scenarios: an energy-constrained untrusted relay network based on the multiple destinations diversity technology,an opportunistic antenna selection cooperative relay network based on Q-learning algorithm,and a full-duplex active eavesdropper in the dynamic transmission networks based on antenna selection.In the first scenario,a combination of wireless energy harvesting technology and physical layer security technology is used to explore the secrecy performance analysis of wireless networks,and an alternating optimization scheme is used to solve the non-convex problem of wireless communication systems in a nonlinear energy harvesting mode.In the second scenario,the system parameters are optimized to maximize the transmission rate,and the off-line updateing Q-learning algorithm combining the performance optimization is used to analyze the network performance with an optimal dynamic transmission strategy.In the last scheme,the multiple antennas source uses artificial noise-aided precoding and the maximum ratio transmission scheme to counteract the interference of full-duplex active eavesdropping nodes.Subsequently,under the condition that full-duplex active eavesdropping nodes can adjust the noise power to balance between the reduceing legal channel capacity and prevent eavesdropping from being detected,the on-line updateing of SARSA algorithm scheme is proposed based on the optimal power allocation factor.Research results show the selection scheme designed in this paper can not only avoid the waste of wireless channel resources at a degree,but also effectively improve the transmission performance.Subsequently,applying reinforcement learning algorithms to wireless networks can enable communication systems to adaptively obtain the optimal transmission performance.A reasonable performance optimization scheme and transmission parameter design can greatly enhance the performance of reinforcement learning algorithms in the wireless communication.