Radar interference decision-making method, device and system, and storage medium

Abstract

The invention discloses a radar interference decision-making method, device and system and a storage medium, and the method comprises the steps: S1, obtaining a potential relation between radar working mode conversion and interference signal release of a jammer according to a radar-jammer confrontation environment; and S2, inputting the threat level corresponding to the current radar working mode as an environment state parameter into a competitive dual deep Q network decision structure based on adaptive greedy adjustment and priority dual experience playback for value evaluation, experience input, experience extraction and parameter updating operation to obtain a radar interference decision result. According to the technical scheme, the problems that a traditional reinforcement learning algorithm is poor in strategy convergence effect and insufficient in experience storage mode are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of electronic countermeasures, and in particular relates to a radar interference decision-making method and device, a system, and a storage medium. Background Art

[0002] In the modern radar electronic warfare environment, the enemy's cognitive radar can dynamically and adaptively adjust its working mode according to the battlefield situation perception results, so as to achieve the scanning, tracking and locking effects on our penetration units, posing a huge threat to our strategic goals. According to prior knowledge, the combat phase of cognitive radar can be roughly divided into five parts: coarse search phase, fine search phase, monitoring phase, tracking phase and locking phase, and the threat level to our penetration units increases in turn. The purpose of our penetration units to implement interference operations on it is to guide the cognitive radar to the mode with the lowest threat level to us. To achieve this goal, it is necessary to construct an intelligent interference decision structure for cognitive radar.

[0003] In the traditional field, there are many types of active jamming signals for radars, but they can be roughly summarized into three types of signals: suppressive jamming, deceptive jamming, and comprehensive jamming. Introducing artificial intelligence technology into the radar jamming decision-making process, using deep reinforcement learning methods, and setting appropriate environmental feature parameters and reward functions can greatly improve decision-making efficiency.

[0004] Through searching the existing technical literature, it is found that Liang Meng et al. published "Radar Autonomous Interference Decision-making Method Based on Dominant Action-Comment" in "Aerospace Electronic Countermeasures" (2023, 39 (05): 17-21.) and proposed a radar autonomous interference decision-making method based on dominant action-comment. The algorithm has the ability to quickly learn and make decisions on interference strategies in unknown environments, but the action selection strategy ε-greedy used is fixed throughout the process, so that strategy fluctuations will occur in the later stage of strategy generation, affecting the interference decision effect; in view of this, the present invention proposes an adaptive greedy adjustment algorithm, which converts the fixed greedy value ε into an ε greedy curve that changes with the number of steps, so as to fit the general thinking of human learning. Zhao Jiachen et al. published "Radar Intelligent Decision Generation Algorithm Based on Deep Reinforcement Learning" in "Modern Radar" (2022, 44 (12): 25-33.) and proposed a decision network training algorithm based on dual deep Q network (DDQN), which effectively improves the intelligence of radar and is used to overcome the over-estimation of Q value caused by the coupling of target network and evaluation network in deep Q network (DQN) algorithm, but the experience replay mechanism used is still insufficient. In view of this, the present invention proposes a dual experience replay mechanism based on an expert experience replay pool and a priority experience extraction mechanism. The mechanism divides the experience replay pool into an ordinary experience replay pool and an expert experience replay pool, and uses a priority experience sampling method in the ordinary experience replay pool to maximize the acquisition of high-value experience. On the basis of the priority experience sampling mechanism, the mechanism realizes dual depth-priority experience replay, realizes further utilization of high-value interactive experience, and thus improves the strategy accuracy of the model.

[0005] In summary, traditional reinforcement learning algorithms still have problems such as poor strategy convergence and insufficient experience storage methods. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide a radar interference decision-making method and device, system, and storage medium to solve the problems of poor strategy convergence effect and insufficient experience storage method in traditional reinforcement learning algorithms.

[0007] To achieve the above object, the present invention adopts the following technical solution:

[0008] A radar jamming decision-making method, comprising:

[0009] Step S1, according to the radar-jammer confrontation environment, obtain the potential connection between the radar working mode conversion and the jammer releasing the jamming signal;

[0010] Step S2: Input the threat level corresponding to the current radar working mode as the environmental state parameter into the competitive dual deep Q network decision structure based on adaptive greedy adjustment and priority dual experience playback to perform value evaluation, experience entry, experience extraction and parameter update operations to obtain the radar interference decision result.

[0011] Preferably, the radar includes five working modes, namely: coarse search mode, fine search mode, surveillance mode, tracking mode and locking mode, and the corresponding threat levels increase in sequence; the combat scenario is set as follows: the radar is initially in coarse search mode; after a suspected target is found, the radar switches to fine search mode; after confirming the target, the radar switches to surveillance mode; after continuous monitoring for a preset period of time, the radar switches to tracking mode; after continuous tracking for a preset period of time, the radar switches to locking mode.

[0012] Preferably, the action library of the jammer includes six different actions, namely: noise amplitude modulation interference, noise frequency modulation interference, spectrum dispersion interference, slice combination interference, intermittent sampling forwarding interference and not taking any interference measures.

[0013] The present invention also provides a radar interference decision-making device, comprising:

[0014] A building module is used to obtain the potential connection between the radar working mode conversion and the jammer releasing the jamming signal according to the radar-jammer confrontation environment;

[0015] The decision module is used to input the threat level corresponding to the current radar working mode as the environmental state parameter into the competitive dual deep Q network decision structure based on adaptive greedy adjustment and priority dual experience playback to perform value evaluation, experience entry, experience extraction and parameter update operations to obtain the radar interference decision result.

[0016] Preferably, the radar includes five working modes, namely: coarse search mode, fine search mode, surveillance mode, tracking mode and locking mode, and the corresponding threat levels increase in sequence; the combat scenario is set as follows: the radar is initially in coarse search mode; after a suspected target is found, the radar switches to fine search mode; after confirming the target, the radar switches to surveillance mode; after continuous monitoring for a preset period of time, the radar switches to tracking mode; after continuous tracking for a preset period of time, the radar switches to locking mode.

[0017] Preferably, the action library of the jammer includes six different actions, namely: noise amplitude modulation interference, noise frequency modulation interference, spectrum dispersion interference, slice combination interference, intermittent sampling forwarding interference and not taking any interference measures.

[0018] The present invention also provides a radar interference decision system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes a radar interference decision method when executed by the processor.

[0019] The present invention also provides a storage medium, on which a computer program is stored, and the computer program executes the radar interference decision method when running.

[0020] The present invention combines reinforcement learning with deep learning networks, and proposes a Dueling Double Deep Q-Network based on Adaptive Greedy Adjustment and Priority Dual Experience Replay (AGA-PDER-D3QN), which combines an adaptive greedy adjustment algorithm and a dual experience replay mechanism based on an expert experience replay pool and a priority experience extraction mechanism based on the D3QN model, so that the jammer can conduct confrontation actions with the radar in a combat environment, continuously reduce the threat level of the radar, and finally complete the predetermined penetration mission. This invention avoids the shortcomings of manual judgment in decision-making effects, optimizes the experience storage method and training result calling method in traditional reinforcement learning algorithms, and improves the scalability of the decision-making system in terms of jamming actions.

[0021] Compared with the existing interference strategy formulation method, the present invention designs a radar intelligent interference decision-making method based on AGA-PDER-D3QN deep reinforcement learning to address the problems of time-consuming manual decision-making, inability to predict the trend of radar situation changes, and the need to preset interference countermeasures in advance. The coarse search mode is used as the initial radar working mode, and the changes in Q network parameters during the training process are analyzed. The radar intelligent interference decision-making method based on deep reinforcement learning adopted by the present invention not only has a faster decision-making speed and decision-making accuracy, but also reduces the data storage pressure of the radar combat system to a certain extent. The implementation plan of the interference algorithm after the decision is most in line with the trend of changes in the radar working mode of the current environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0023] Figure 1A flowchart of a radar interference decision-making method according to an embodiment of the present invention;

[0024] Figure 2 The change of the reward value of a single round in the present invention;

[0025] Figure 3 The change in the number of steps required for a single round in the present invention;

[0026] Figure 4 This is the change in the final result of the round in the present invention. DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0029] Embodiment 1:

[0030] An embodiment of the present invention provides a radar interference decision method, including:

[0031] Step S1, according to the radar-jammer confrontation environment, obtain the potential connection between the radar working mode conversion and the jammer releasing the jamming signal;

[0032] Step S2: Input the threat level corresponding to the current radar working mode as the environmental state parameter into the competitive dual deep Q network decision structure based on adaptive greedy adjustment and priority dual experience playback to perform value evaluation, experience entry, experience extraction and parameter update operations to obtain the radar interference decision result.

[0033] As an implementation mode of an embodiment of the present invention, the radar includes five working modes, namely: coarse search mode, fine search mode, monitoring mode, tracking mode and locking mode, and the corresponding threat levels are increased in sequence; the combat scenario is set as follows: the radar is initially in coarse search mode; after a suspected target is found, the radar switches to fine search mode; after the target is confirmed, the radar switches to monitoring mode; after continuous monitoring for a preset period of time, the radar switches to tracking mode; after continuous tracking for a preset period of time, the radar switches to locking mode.

[0034] As an implementation method of an embodiment of the present invention, the action library of the jammer includes six different actions, namely: noise amplitude modulation interference, noise frequency modulation interference, spectrum dispersion interference, slice combination interference, intermittent sampling forwarding interference and not taking any interference measures.

[0035] Embodiment 2:

[0036] like Figure 1 As shown, an embodiment of the present invention provides a radar interference decision method, including:

[0037] Step 1: Build an interference countermeasure environment

[0038] (1) In the present invention, the radar is set to have five working modes, namely: coarse search mode m0, fine search mode m1, monitoring mode m2, tracking mode m3 and locking mode m4, and the corresponding threat levels are increased in sequence. The combat scenario is set as follows: the radar is initially in coarse search mode; after discovering a suspected target, the radar switches to fine search mode; after confirming the target, the radar switches to monitoring mode; after continuous monitoring for a period of time, the radar switches to tracking mode; after continuous tracking for a period of time, the radar switches to locking mode.

[0039] (2) In the present invention, the action library of the jammer includes six different actions, namely: noise amplitude modulation interference a1, noise frequency modulation interference a2, spectrum dispersion interference a3, slice combination interference a4, intermittent sampling forwarding interference a5 and no interference measures a0. Among them, noise amplitude modulation interference can use its frequency band blocking characteristics to suppress the radar in the coarse search mode, so that the radar can be maintained in the coarse search mode; noise frequency modulation interference can use its frequency band aiming characteristics to further improve the suppression capability of the radar under the premise of unchanged interference power, thereby preventing the radar from switching from the fine search mode to the subsequent monitoring, tracking and locking mode; when the radar is in the monitoring, tracking and locking mode, the jammer can use deceptive interference signals, such as spectrum dispersion interference, slice combination interference, intermittent sampling forwarding interference, etc., to construct a false target with a high similarity, and induce the radar observation target to the false target, thereby ensuring the safety of the real target.

[0040] (3) Without taking any interference measures, the radar mode transition probability is set as follows:

[0041] When the jammer enters the radar detection area, the radar will switch to the current mode m i The mode transition probability under is:

[0042]

[0043] Among them, t is the current time, ω c It is the radar detection factor. The larger its value is, the stronger the detection sensitivity of the radar is.

[0044] (4) When different jamming measures are taken, the radar mode transition probability will also change accordingly:

[0045] When the jammer takes suppressive jamming action, i.e. a j =a1 or a j = a2, the radar is in the current mode m i The mode transition probability under is:

[0046]

[0047] Among them, p(m i ,a j ) is the current radar mode m i Next, implement action a j The decrease value of the post-mode transition probability.

[0048] When the jammer takes deceptive jamming action, i.e. a j =a3, a j =a4 or a j = a5, the radar is in the current mode m i The mode transition probability under is:

[0049]

[0050] Among them, p(m i ,a j ) is the current radar mode m i Next, implement action a j The probability that the radar will track a false target.

[0051] (5) Except for the coarse search mode m0 and the lock mode m4, when the radar fails to detect a target for N = 10 consecutive times in the current mode, the radar mode falls back to the previous mode.

[0052] (6) In the present invention, the state set is set as follows:

[0053] S=[m t ,v Jx ,v Jy ,l Jx ,l Jy ]

[0054] Among them, m t is the current radar working mode, v Jx and v Jy are the lateral and longitudinal components of the jammer's velocity, respectively, l Jx and l Jy are the horizontal and vertical coordinates of the jammer's position respectively.

[0055] (7) In the present invention, the action set is set as follows:

[0056] A=[a0,a1,a2,a3,a4,a5]

[0057] (8) In the present invention, the single-step reward function is set as follows:

[0058] First, the changes in radar operating modes should be considered:

[0059]

[0060] That is, when the radar working mode is improved, the intelligent agent should be punished; when the radar working mode is reduced, the intelligent agent should be rewarded; when the radar working mode remains unchanged, no reward or punishment should be given to the intelligent agent.

[0061] Secondly, the penetration distance of the jammer should be considered:

[0062]

[0063] That is, the longer the jammer's penetration distance is, the more rewards will be given to the agent.

[0064] In addition, the global rewards after the final breakthrough should be considered, namely:

[0065]

[0066] Therefore, the single-step reward function can be expressed as follows:

[0067]

[0068] Where x is the total number of penetration steps required at the end of a single training round.

[0069] The state set, action set, and single-step reward function are the core prerequisite structures of reinforcement learning, which provide the necessary prerequisites for the training of interference strategies.

[0070] Step 2: Build a radar jamming decision-making mechanism

[0071] (1) In the present invention, Dueling Double Deep Q-Network (D3QN) is used as the basic decision model. Its Q-value network adopts a fully connected network (Deep Neural Network, DNN) structure, and additionally adds a state attention layer and an action value layer to the hidden layer, so that the decision model has the ability to distinguish the state-affected subject, avoiding the fluctuation of training effect caused by environmental changes. In addition, the model transfers the right to choose the next moment estimated action from the Q target network to the Q estimation network, while the Q value calculation work at the next moment is still completed by the Q target network, thereby realizing the decoupling between action selection and Q value calculation, and effectively reducing the over-estimation probability of the model.

[0072] (2) In the present invention, the action selection mechanism of the decision model should be more in line with general learning thinking, that is, exploration and trial in the early stage, and obtaining as diverse experience feedback as possible; taking into account exploration and learning in the middle stage, guiding the model to implement actions with higher value; and consolidation in the later stage, so as to generate feasible interference strategies as soon as possible. Therefore, the present invention designs an ε-greedy action selection mechanism based on an adaptive greedy adjustment algorithm.

[0073] The adaptive greedy adjustment algorithm consists of two parts: the baseline curve and the greedy adjustment coefficient. First, a monotonically decreasing curve l(t) with a variation interval of [0,1] is selected:

[0074]

[0075] Then use past historical experience to calculate the greed adjustment coefficient

[0076]

[0077] Among them, T μ is the variation period of the greedy adjustment coefficient, and the initial value of the greedy adjustment coefficient μ0=0.

[0078] The above formula can be understood as follows: when the mean value of past historical rewards is less than or equal to 0, it means that the model has not learned effective experience. At this time, the greed value needs to be increased to make the model more inclined to continue exploring the environment; when the mean value of past historical rewards is greater than 0, it means that the model has learned effective experience. At this time, the greed value needs to be reduced to make the model more inclined to select the optimal action. Therefore, the greed value at the current moment can be defined as:

[0079] ε t =l(t)+μ i

[0080] This mechanism will dynamically adjust the greediness value based on historical experience and the benchmark greediness adjustment curve, so that the agent can adaptively adjust its own action selection tendency according to the learning effect, effectively improving the flexibility of the model's action selection.

[0081] (3) In the present invention, in order to further improve the efficiency of the decision-making model in utilizing experience, a dual experience replay mechanism based on an expert experience replay pool and a priority experience extraction mechanism is designed. This mechanism divides the experience replay pool into a common experience replay pool and an expert experience replay pool, and uses a priority experience sampling method in the common experience replay pool to obtain as many high-value experiences as possible. At the same time, a random sampling mechanism is used to sample the expert experience replay pool, which is mixed with the samples obtained from the common experience pool and then handed over to the model for learning.

[0082] (4) The D3QN model mentioned in (1) is combined with the mechanisms in (2) and (3) to obtain a competitive dual deep Q network based on adaptive greedy regulation and priority dual experience replay. This model is used to learn the interference strategy.

[0083] (5) Obtain the current radar working mode. According to the radar working mode, jammer speed and its relative position, the current state parameter s is obtained. t , and send it to the Q estimation network for evaluation. After evaluating all action Q values, the current action a is selected based on the adaptive greedy adjustment algorithm and the ε-greedy action selection mechanism. t ∈A, where A is the action set.

[0084] The ε-greedy action selection mechanism after integrating the adaptive greedy adjustment algorithm can be summarized as follows:

[0085]

[0086] where ε t It can be obtained according to the greedy value calculation method in (2).

[0087] (6) Change the current action a t Applied to the environment, get the next state s t+1 , and according to s t and t+1 Calculate the current reward r t . The experience string [s t ,a t ,r t ,s t+1 ] are stored in the common experience replay pool in order, and the priority value corresponding to the experience string is calculated. The calculation process is as follows:

[0088] p i =|δi |+β

[0089] Among them, δ i is the timing error of the i-th experience string, and β is a very small constant not equal to 0 to ensure that the extraction priority of all experiences is greater than 0. In addition, the timing error of the i-th experience string can be expressed as:

[0090] δ i =r i +γ·Q(s i ′,a i ′;θ - )-Q(s i ,a i ;θ)

[0091] Among them, γ is the attenuation factor, which is used to control the influence of past experience on the current moment value.

[0092] From this we can see that the extraction probability of each experience string is:

[0093]

[0094] Among them, α∈[0,1] is the priority extraction emphasis. The higher α is, the more the model tends to use the priority sampling method to extract empirical samples. When α=0, the sampling method changes from priority sampling to random sampling.

[0095] Finally, the priority value is stored in the sum tree structure.

[0096] (7) Based on the state value and reward value in the current experience string, measure whether it meets the conditions for entering the expert experience replay pool. Each state value corresponds to an entry threshold, and the initial threshold is 0. When the reward value of the experience is greater than the threshold, it will be entered into the expert experience replay pool, and the entry threshold will be adjusted to the reward value of the experience.

[0097] (8) Repeat (5)-(7) until the data in the common experience replay pool meets the minimum sampling requirements.

[0098] (9) The priority experience sampling mechanism is used to sample the ordinary experience replay pool, and the random sampling mechanism is used to sample the expert experience replay pool. The sampled data of the two experience pools are mixed and then handed over to the Q-value network for learning.

[0099] According to the next moment state s in the sample t+1 And the estimated action a′ calculates the Q target network output value:

[0100]

[0101] According to the current state s in the sample tand the current action a t Calculate the Q estimation network output value:

[0102] Q evaluate =Q(s t ,a t ;θ)

[0103] Using the above calculation results and the current reward r t , attenuation factor γ and other parameters, calculate the model loss function, where the loss function is in the form of a mean square error function:

[0104]

[0105] The gradient value of this function is used to perform reverse gradient update on the Q estimation network.

[0106] (10) Setting the network parameter migration interval, that is, after a certain number of Q estimation network weight updates, the internal weight parameters of the Q estimation network are completely copied to the Q target network in a hard update manner.

[0107] (11) Repeat (5)-(10) until the Q-value network converges, that is, the decision results of the Q target network and the Q estimation network are almost the same.

[0108] After the model training is completed, the Q target network gives the best action suggestion according to the current state. After the jammer executes the best action, the environment changes. By looping the above operations, a continuous jamming strategy can be obtained.

[0109] like Figure 2 As shown in the figure, the change of the reward value for a single round in the present invention is shown. It can be seen that as the model continues to learn the interference interaction experience, its average round reward value at about 300 rounds can basically remain in the positive range, which shows that after 300 rounds, the interference strategy learned by the agent has a certain effectiveness.

[0110] like Figure 3 As shown in the figure, the number of steps required for a single round in the present invention changes. It can be seen that as the model continues to learn interference interaction experience, the number of action steps required for a single round after about 300 rounds is reduced to about 4 steps, indicating that the agent has learned the fastest interference implementation path to reach the non-threatening radar working mode.

[0111] like Figure 4 As shown in the figure, the round results in the present invention change. It can be seen that the agent continuously performs trial and error and correction before reaching 300 rounds. After reaching 300 rounds, the subsequent rounds can successfully reduce the radar's working mode to the lowest coarse search mode, and the threat level is reduced to the lowest.

[0112] From this we can see that the interference algorithm combination given after the interference decision has an extremely excellent comprehensive interference effect on the radar.

[0113] Embodiment 3:

[0114] An embodiment of the present invention provides a radar interference decision-making device, including:

[0115] A building module is used to obtain the potential connection between the radar working mode conversion and the jammer releasing the jamming signal according to the radar-jammer confrontation environment;

[0116] The decision module is used to input the threat level corresponding to the current radar working mode as the environmental state parameter into the competitive dual deep Q network decision structure based on adaptive greedy adjustment and priority dual experience playback to perform value evaluation, experience entry, experience extraction and parameter update operations to obtain the radar interference decision result.

[0117] As an implementation mode of an embodiment of the present invention, the radar includes five working modes, namely: coarse search mode, fine search mode, monitoring mode, tracking mode and locking mode, and the corresponding threat levels are increased in sequence; the combat scenario is set as follows: the radar is initially in coarse search mode; after a suspected target is found, the radar switches to fine search mode; after the target is confirmed, the radar switches to monitoring mode; after continuous monitoring for a preset period of time, the radar switches to tracking mode; after continuous tracking for a preset period of time, the radar switches to locking mode.

[0118] As an implementation method of an embodiment of the present invention, the action library of the jammer includes six different actions, namely: noise amplitude modulation interference, noise frequency modulation interference, spectrum dispersion interference, slice combination interference, intermittent sampling forwarding interference and not taking any interference measures.

[0119] Embodiment 4:

[0120] An embodiment of the present invention further provides a radar interference decision system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes a radar interference decision method when executed by the processor.

[0121] Embodiment 5:

[0122] An embodiment of the present invention further provides a storage medium, on which a computer program is stored, and the computer program executes the radar interference decision method when running.

[0123] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A radar jamming decision method, characterized in that: include: Step S1, according to the radar-jammer confrontation environment, obtain the potential connection between the radar working mode conversion and the jammer releasing the jamming signal; Step S2: Input the threat level corresponding to the current radar working mode as the environmental state parameter into the competitive dual deep Q network decision structure based on adaptive greedy adjustment and priority dual experience playback to perform value evaluation, experience entry, experience extraction and parameter update operations to obtain the radar interference decision result.

2. The radar interference decision-making method according to claim 1, characterized in that: The radar has five working modes: coarse search mode, fine search mode, surveillance mode, tracking mode and lock mode, and the corresponding threat levels increase in sequence. The combat scenario is set as follows: the radar is initially in coarse search mode; after finding a suspected target, the radar switches to fine search mode; After confirming the target, the radar switches to monitoring mode; after continuously monitoring for a preset period of time, the radar switches to tracking mode; after continuously tracking for a preset period of time, the radar switches to locking mode.

3. The radar interference decision-making method according to claim 2, characterized in that: The jammer's action library contains six different actions, namely: noise amplitude modulation interference, noise frequency modulation interference, spectrum dispersion interference, slice combination interference, intermittent sampling forwarding interference and no interference measures.

4. A radar jamming decision-making device, characterized in that: include: A building module is used to obtain the potential connection between the radar working mode conversion and the jammer releasing the jamming signal according to the radar-jammer confrontation environment; The decision module is used to input the threat level corresponding to the current radar working mode as the environmental state parameter into the competitive dual deep Q network decision structure based on adaptive greedy adjustment and priority dual experience playback to perform value evaluation, experience entry, experience extraction and parameter update operations to obtain the radar interference decision result.

5. The radar interference decision-making device according to claim 4, characterized in that: The radar has five working modes: coarse search mode, fine search mode, surveillance mode, tracking mode and lock mode, and the corresponding threat levels increase in sequence. The combat scenario is set as follows: the radar is initially in coarse search mode; after finding a suspected target, the radar switches to fine search mode; After confirming the target, the radar switches to monitoring mode; after continuously monitoring for a preset period of time, the radar switches to tracking mode; after continuously tracking for a preset period of time, the radar switches to locking mode.

6. The radar interference decision-making device according to claim 5, characterized in that: The jammer's action library contains six different actions, namely: noise amplitude modulation interference, noise frequency modulation interference, spectrum dispersion interference, slice combination interference, intermittent sampling forwarding interference and no interference measures.

7. A radar jamming decision system, characterized in that: include: A memory and a processor, wherein the memory stores a computer program executed by the processor, and when the computer program is executed by the processor, the radar interference decision method according to any one of claims 1 to 3 is executed.

8. A storage medium, characterized in that: The storage medium stores a computer program, which executes the radar interference decision method according to any one of claims 1 to 3 when running.

Images