| In recent years,with the introduction of artificial intelligence technology,cognitive electronic warfare has gradually replaced the traditional electronic countermeasure and become the main form of modern war.As a jammer,jamming confrontation must be carried out if we want to take the initiative in the war and protect our own targets.In the real battlefield environment,the radar side and the jammer are non cooperative.The confrontation between the two sides is a dynamic game process.The traditional single jamming method can not achieve good jamming effect.Especially for the multi-functional radar with changeable state and strong anti-jamming ability,the static jamming strategy relying on template matching or intelligent algorithm is not suitable for modern war.Using artificial intelligence technology to dynamically adjust the jamming strategy according to the battlefield environment is more in line with the development direction of electronic countermeasures.Based on the idea of reinforcement learning,this paper studies the intelligent jamming decision-making method against multifunctional radar in different countermeasure environments.The research contents of this paper are as follows:The radar working state and active jamming style involved in the follow-up research are studied.The common working states of radar are analyzed in detail,and the working tasks,working principles and signal characteristics of each state are studied.The working principle of typical active jamming style is studied,and the jamming effect is simulated and analyzed.This paper analyzes the intelligent algorithms commonly used in interference decision-making,studies their ideological source and solution process,and points out the reasons why these algorithms are not suitable for modern war.Aiming at the problem that the known jamming decision-making methods can not adaptively and dynamically select jamming strategies according to the state changes of multi-functional radar,reinforcement learning is introduced into jamming decision-making,and the jamming decision-making methods of radar and jammer in "one-to-one" and "many to many" environments are studied respectively.Aiming at the one-to-one jamming multi-functional radar scene,the jamming decision model based on reinforcement learning is established,and the one-to-one jamming decision method based on adaptive HAQL is designed.The main factors affecting the formulation of radar jamming strategy are analyzed,and the jamming style and power are combined into a joint jamming action.Considering that the evaluation result of interference effect,as the reward value in interference countermeasure based on reinforcement learning,is an important basis for selecting interference action.A jamming effect evaluation method based on the change of radar working state is proposed.The support vector machine optimized by the improved sparrow search algorithm is used to identify the radar working state,and compared with other algorithms to verify its feasibility.The interference experience table is established and combined with heuristic Q-learning to guide the selection of interference actions.Simulation results show that this method can give the optimal jamming strategy for each state of multi-functional radar,maximize the utilization of jamming resources,and effectively accelerate the convergence speed and make jamming decisions compared with Q-learning algorithm.Aiming at the scene of many to many jamming,the corresponding relationship between radar state,jammer and jamming style is more complex.A jamming decision method based on is proposed.By analyzing DDQN and priority experience playback,the jamming decision-making model based on PER-DDQN is established,and the radar state,transfer relationship,jamming action and corresponding reward are determined.The background of multi jammer flying with UAV against multi-functional radar network is designed,and the effectiveness of static jamming strategy and this algorithm for multi-functional radar network jamming decision is simulated and compared.The influence of DDQN and PER-DDQN on jamming decisionmaking is compared from two aspects of training error and penetration success rate,and the superiority of PER-DDQN algorithm in stability and convergence speed is verified. |
