Research On Radar Jamming Decision Technology Based On Machine Learning

The complexity of the battlefield environment,the diversity of information,and the variability of state bring great challenges to electronic jamming.With the rapid development of cognitive electronic warfare,jamming decision-making technology with adaptive ability has attracted much attention.How to intelligently recognize radar working mode and choose jamming mode in complex environment has become an urgent problem in jamming decisionmaking.Machine learning is an intelligent data analysis tool.In this thesis,radar jamming decision technology based on machine learning is studied.With the development of new system radar,the difficulty of radar mode recognition is increased.The jammer cannot directly obtain the prior information of the non-cooperative radar.In order to improve the accuracy of working pattern recognition,a working pattern recognition method based on hidden markov model is proposed.In order to avoid the dependence on prior knowledge,multi-arm gambling machine based on clustering and q-learning interference decision method based on hidden markov model are proposed,Simulation results show that the proposed work pattern recognition method has a high recognition accuracy,and the decision method convergence speed is fast,the probability of choosing the optimal interference pattern is high.The main work of this thesis is as follows:The basic principle and implementation process of cognitive interference decision technology are described.A radar jamming decision model based on machine learning is established by using supervised learning and reinforcement learning in machine learning.Several working modes of phased array radar are studied,and the common suppressive jamming and spoofing jamming and their influence on the working mode of radar are discussed.The jamming effect of the jamming pattern is evaluated with the threat degree change caused by the shift of radar working mode.In order to estimate the radar working mode more accurately,a radar working mode recognition method based on hidden markov model is studied.The solution methods of hidden markov model,Baum-Welch algorithm and Viterbi algorithm are discussed.The relationship between signal characteristic parameters and radar working mode is analyzed.Through simulation experiments,the working pattern recognition method based on hidden markov model is compared with other common methods,such as radar working pattern recognition method based on clustering and BP neural network.Aiming at the problems of slow convergence speed and many interaction times of the optimal interference style selection algorithm in the current working mode,the interference decision model based on multi-arm gambling machine is studied.Several multi-arm gambling machine algorithms such as ε-greedy algorithm,UCB algorithm,UCBV algorithm and EUCBV algorithm are discussed,and the realization principle of these algorithms is analyzed.Through the simulation experiment,based on the radar working mode identification based on clustering method,the jamming decision effect of the multi-arm gambling machine algorithm is compared from the average return of jamming style and the accuracy of optimal jamming style selection.Aiming at the problem that the work mode shifts due to the interference of the jammer,the interference decision model based on markov decision process is constructed.The hidden markov model can be used to estimate the probability distribution of radar working mode transition under different jamming styles,and then the result of jamming decision can be obtained by q-learning algorithm.Through simulation experiments,the multi-arm gambling machine algorithm based on clustering,q-learning algorithm based on hidden markov model and q-learning algorithm based on statistics are compared,and the performance of different algorithms is analyzed.