Anti-active Jamming Method Of Intelligent Radar Network

In modern warfare,in order to gain battlefield advantage,the electronic warfare that competes for the dominant electromagnetic power in the electronic field is regarded as equally important as the traditional land,sea and air warfare.Radar network uses multiple radars to monitor the same area,so when some radars are jammed,some radars can still work,which makes them have a certain anti-jamming ability and better battlefield survivability.However,the continuous progress of active jamming technology makes it possible for multiple jammers to cooperate to jam multiple radar networks.Moreover,the dispersion of jammers in the airspace makes jamming possible in any position of the radar detection area,which also makes the anti-jamming of radar network more difficult.In order to win the challenge brought by these problems,an intelligent radar network anti-active jamming method based on reinforcement learning is proposed.The neural network is trained by the signal characteristics and measurement behavior characteristics of the received echo,thus improving the anti-jamming performance of the radar network.The work done in this paper can be summarized as follows:Firstly,the basic structure and principle of reinforcement learning are explained,the feasibility and iteration of reinforcement learning method are analyzed,and the appropriate reinforcement learning method for this topic is discussed,and illustrated by simulation experiment.Secondly,the main structure of the intelligent radar network is designed,the extraction methods of data signal and behavior characteristics are studied,and the fusion method of data processed by the fusion center is studied.The status and reward of the agent are designed and verified by simulation.Finally,the design of intelligent radar network agent based on DQN is proposed,and simulation experiments are carried out to prove the effectiveness of the method,that is,the superiority of the traditional method.In view of the slow convergence of real-time reward and the deviation from the real situation,a new reward method for training current agents using future reward is proposed and simulated.Aiming at the low sampling efficiency of training data in DQN network,the method of normalized priority sampling is adopted to improve the performance of agents.In order to cope with the worse situation,a dual DQN method is proposed to avoid the difficulty of DQN training under complex noise caused by overestimation.At the same time,the new method reduces the performance degradation of the system under worse conditions and improves the robustness of the radar network.The simulation results show that this method can improve the robustness of intelligent radar network.