Research On Airborne Communication Waveform Parameter Decision Method Based On Reinforcement Learning

In recent years,wireless communication technology has developed rapidly,the electromagnetic environment has become more and more complex,and the requirements for the communication capacity and communication quality of the system have become higher and higher.In the process of airborne communication transmissions,factors such as complex aviation channel environment and susceptibility of wireless devices to interference have limited the communication performance of airborne radio systems.However,the traditional waveform adaptive methods still have problems such as large dependence on a priori knowledge,low decision-making efficiency,and poor environmental adaptability.Thus,research on efficient and reliable waveform parameter decision-making methods for airborne communication is imminent.To enhance the communication performance of airborne wireless communication systems,this study takes the cognitive radio adaptive waveform technology and reinforcement learning algorithm as the theoretical basis,considers the existence of various interference signals and fading in the aviation wireless channel,an intelligent decision model of airborne communication waveform parameters is established based on reinforcement learning algorithm:Firstly,to address the problem that airborne radio systems are vulnerable to interference,the anti-interference decision method is studied,and an intelligent communication frequency decision-making model is proposed based on the Deep Q-Network(DQN)algorithm,which can decide the communication frequency of the next round based on the interference discrimination information of the current round and effectively avoid interference.Compared with the traditional decision model with input interference style information,this decision model converges more stably and has a higher decision success rate.Moreover,to address the problems of large action space and slow convergence of existing algorithms,the action judgment factor G is introduced,and the G-DQN algorithm is proposed to eliminate the actions with poor feedback in the decision-making process and reduce the action space.The simulation results show that the convergence speed of the decision algorithm is greatly improved.Secondly,research on modulation coding and spreading decision-making is carried out to further improve the spectral efficiency of communication systems.The Q-learning algorithm-based modulation coding and spreading decision model is proposed by making full use of the autonomous learning characteristics of reinforcement learning.Compared with the traditional Modulation and Coding Scheme(MCS),this decision model does not rely on a priori knowledge,has a strong adaptability to different air channels,and has a superior decision performance.Moreover,to address the problems that the existing algorithms converge to local optimum,convergence speed is slow or even hard to converge,the existing algorithms are improved based on the PHC algorithm,and the DE-Q-learning algorithm is proposed,which dynamically adjusts the probability value ε through the feedback rewards and rounds to improve the success rate and convergence speed of the decision algorithm.Finally,the decision knowledge database is constructed based on the classifier algorithm to improve the decision model,and the existing decision results are stored and utilized to further improve the convergence performance of the algorithm.The reinforcement learning-based waveform parameter decision algorithm proposed in this paper does not rely on a priori knowledge and has greatly improved in convergence stability and convergence speed,which can decide the most suitable waveform parameters in complex airborne communication environments and ensure the reliable and effective transmission of information,providing reference values for the further research of waveform parameter decision methods in the future.