Research On Intelligent Anti-interference Method For Wireless Communication Based On Machine Learnin

With the continuous development of jamming technology in wireless communications,malicious jamming is becoming increasingly efficient and intelligent,posing an increasingly serious challenge to the reliable transmission capability of wireless communication systems.According to the idea of anti-intelligence with intelligence,there is an urgent need for intelligent anti-jamming methods in communications to cope with the escalating threats of malicious jamming.This thesis addresses the core problem of how to improve the reliability of wireless communication in the jamming environment with intelligent anti-jamming methods,and focuses on the three key elements of efficient and reliable communication in an environment where users face frequency jamming between malicious users and users,dynamic unknown jamming and multi-parameter dynamic jamming,aiming to provide preliminary solutions and theoretical technical support for improving the intelligence of anti-jamming in wireless communications.The main works of the dissertation are summarized as follows.(1)A multi-users distributed anti-jamming coalition formation algorithm is proposed for external malicious jamming and intra-user mutual interference problems.By optimizing the grouping strategy of users,an efficient distributed cooperative anti-jamming decision is achieved,which reduces both the external malicious jamming threat and the mutual interference among users.In addition,a reciprocity criterion is proposed to theoretically demonstrate that the proposed game model can obtain stable Nash equilibrium solutions with the help of an exact potential game,which provides a theoretical basis for the anti-jamming performance of the distributed algorithm.The simulation results show that the proposed algorithm has better antijamming performance compared with the direct transmission algorithm.(2)A Q-learning anti-jamming routing algorithm based on improved confidence interval upper bound is proposed for the anti-jamming routing problem in wireless communication networks,which can quickly converge to the optimal anti-jamming path without the a priori knowledge of the jamming environment.The improved confidence interval upper bound algorithm is used instead of the ε-greedy or Softmax algorithm in traditional Q-learning to balance the "exploration" and "exploitation" of the reinforcement learning,which can avoid the decisive influence of preset parameters on the convergence speed of the anti-jamming routing algorithm in traditional methods and improve the speed of convergence of the communication network.The simulation results show that the statistical average packet acceptance rate of the proposed algorithm is better than Dyna＿Q algorithm and traditional Q learning algorithm.Moreover,compared with UCBQ algorithm without considering the effect of reward variance and initial value,the iteration times of the algorithm can be reduced and the speed of exploration to the destination node can be accelerated when the algorithm achieves the same anti-jamming performance.(3)To address the problem that the traditional single-dimensional domain anti-jamming method is difficult to ensure effective protection against various new jamming threats in complex electromagnetic environments,an anti-jamming intelligent algorithm based on proximal policy optimization is proposed.The algorithm first models the communication problem under smart targeting jamming as a Markov decision process,and uses a proximal policy optimization algorithm based on the solution to obtain the optimal joint anti-jamming method of the system,which rapidly approximates the optimal transmission strategy and improves the reliability and effectiveness of communication.Simulation results show that the statistical average packet reception rate performance of the proposed algorithm is significantly better than that of deep Q networks and Q-learning algorithms,and the number of iterations of the algorithm can be scaled down and the convergence speed can be accelerated when the same anti-jamming performance is achieved.