Electronic countermeasure off-line decision-making method, system and equipment and storage medium

Abstract

The invention discloses an electronic countermeasure off-line decision-making method, system and equipment and a storage medium, and aims to solve the problems that in a complex electromagnetic environment, radar state information obtained by an interference party is partially considerable and has deviation, and when an interference strategy solved by a general decision-making method is deployed in a real environment, the performance is often difficult to guarantee due to limitation, so that the interference strategy is difficult to ensure. According to the scheme provided by the invention, the electromagnetic environment is modeled into the opponent agent, and the interference agent and the opponent agent are subjected to interactive training to obtain the final interference agent, so that a radar interference decision can be better made, and the decision of the interference agent in the complex electromagnetic environment becomes efficient and safe.

Description

Technical Field

[0001] The present invention relates to the technical field of offline decision-making in electronic confrontation scenarios, and in particular to an electronic confrontation offline decision-making method, system, device and storage medium. Background Art

[0002] Radar jamming decision-making is a crucial part of radar confrontation, and the jamming strategy output by the jamming decision directly affects the quality of the jamming effect. In practical applications, although the jammer obtains the working mode of the radar through reconnaissance and other means, there are multiple jamming styles that can be selected for the same radar working mode. Traditional jamming decision-making methods rely on experience or templates to select jamming styles, which cannot guarantee that the selected jamming style is the best. Deep reinforcement learning is a method that has both perception and decision-making capabilities. Through deep reinforcement learning modeling, the jammer can continuously learn from the experience of confrontation during the confrontation with the radar, thereby dynamically selecting and optimizing the optimal jamming style. This method not only improves the accuracy and effectiveness of jamming decisions, but also enhances the adaptive ability of the jamming system in complex environments.

[0003] Online reinforcement learning models collect samples and learn strategies through online trial-and-error interactions with the environment, which is an effective means to solve sequential perception decision-making problems. However, this online interactive active learning paradigm brings high costs and safety issues when collecting samples in some complex real environments. Offline reinforcement learning, as a data-driven reinforcement learning paradigm, emphasizes learning strategies from static sample data sets and has no exploratory interactions with the environment, providing a feasible solution for the deployment of reinforcement learning algorithms in real environments. However, no relevant solutions for applying offline reinforcement learning to radar jamming decision-making have been found so far. Summary of the invention

[0004] The purpose of the present invention is to provide an electronic countermeasure offline decision-making method, system, device and storage medium, which can improve the effectiveness of interference strategies and make the interference party's decision-making in a complex electromagnetic environment efficient and safe.

[0005] The objective of the present invention is achieved through the following technical solutions:

[0006] An electronic countermeasure offline decision-making method, comprising:

[0007] Model the interferer as an interferer agent;

[0008] Take the given historical game sequence as the original data set and use it to construct the opponent agent;

[0009] The interfering agent and the opponent agent are alternately trained, and the steps include: in the current iteration, using the opponent agent to perform data extrapolation, adding the generated data to the extended data set, using the original data set and the extended data set to train the interfering agent, and the training goal is to maximize the interference value function when the interfering agent and the opponent agent confront each other; according to the offline adversarial learning method combined with the training of the interfering agent trained in the current iteration, the opponent agent is trained, and the training goal is to minimize the interference value function when the interfering agent and the opponent agent confront each other; iterate continuously to obtain the final interfering agent;

[0010] The final jammer agent is deployed in a real environment for radar jamming decision-making.

[0011] An electronic countermeasure offline decision-making system, comprising:

[0012] A jammer agent modeling unit, used for modeling the jammer as a jammer agent;

[0013] The opponent agent construction unit is used to take the given historical game sequence as the original data set and use it to construct the opponent agent;

[0014] The agent alternating training unit is used to alternately train the interfering agent and the opponent agent, and the steps include: in the current iteration, using the opponent agent to perform data extrapolation, adding the generated data to the extended data set, using the original data set and the extended data set to train the interfering agent, and the training goal is to maximize the interference value function when the interfering agent confronts the opponent agent; according to the offline adversarial learning method combined with the training of the interfering agent trained in the current iteration, the opponent agent is trained, and the training goal is to minimize the interference value function when the interfering agent confronts the opponent agent; iterate continuously to obtain the final interfering agent;

[0015] The deployment application unit is used to deploy the final jammer intelligent agent in a real environment to make radar jamming decisions.

[0016] A processing device, comprising: one or more processors; a memory for storing one or more programs;

[0017] When the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned method.

[0018] A readable storage medium stores a computer program, which implements the above method when the computer program is executed by a processor.

[0019] It can be seen from the technical solution provided by the present invention that in a complex electromagnetic environment, the radar state information obtained by the interferer is partially considerable and biased. When the interference strategy obtained by the general decision-making method is deployed in a real environment, due to its limitations, the performance is often difficult to guarantee. Therefore, the solution provided by the present invention, by modeling the electromagnetic environment into an opponent intelligent agent, the interferer intelligent agent and the opponent intelligent agent are interactively trained to obtain the final interferer intelligent agent, which can better make radar interference decisions and make the interferer's decision-making in a complex electromagnetic environment efficient and safe. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0021] Figure 1 A flowchart of an electronic countermeasure offline decision-making method provided by an embodiment of the present invention;

[0022] Figure 2 A flow chart of agent training provided by an embodiment of the present invention;

[0023] Figure 3 A schematic diagram of an electronic countermeasure offline decision-making system provided by an embodiment of the present invention;

[0024] Figure 4 A schematic diagram of a processing device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0025] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the drawings 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 protection scope of the present invention.

[0026] First, the terms that may be used in this article are explained as follows:

[0027] The terms "include", "comprises", "contains", "has" or other descriptions with similar semantics should be interpreted as non-exclusive inclusion. For example, including certain technical feature elements (such as raw materials, components, ingredients, carriers, dosage forms, materials, dimensions, parts, components, mechanisms, devices, steps, procedures, methods, reaction conditions, processing conditions, parameters, algorithms, signals, data, products or products, etc.) should be interpreted as including not only certain technical feature elements explicitly listed, but also other technical feature elements known in the art that are not explicitly listed.

[0028] The term "consisting of..." means excluding any technical feature elements not explicitly listed. If this term is used in a claim, it will make the claim closed, so that it does not contain technical feature elements other than the technical feature elements explicitly listed, except for the conventional impurities related to them. If this term only appears in a clause of a claim, it only limits the elements explicitly listed in the clause, and the elements recorded in other clauses are not excluded from the overall claim.

[0029] The following is a detailed description of an electronic countermeasure offline decision method, system, device and storage medium provided by the present invention. The contents not described in detail in the embodiments of the present invention belong to the prior art known to professional and technical personnel in the field. If no specific conditions are specified in the embodiments of the present invention, the conditions are carried out according to the conventional conditions in the field or the conditions recommended by the manufacturer. If the manufacturer of the reagents or instruments used in the embodiments of the present invention is not specified, they are all conventional products that can be purchased commercially.

[0030] Embodiment 1

[0031] The embodiment of the present invention provides an electronic countermeasure offline decision-making method, such as Figure 1 As shown, it mainly includes the following steps:

[0032] Step 1: Model the interferer as an interferer agent.

[0033] Step 2: Take the given historical game sequence as the original data set and use it to construct the opponent agent.

[0034] Preferably, the opponent agent can be constructed based on the original data set through maximum likelihood estimation.

[0035] Preferably, the opponent agent represents the transfer function of the electromagnetic environment in the electronic confrontation (training process), and its optimization goal is to minimize the benefit of the interfering agent (ie, the interference value function).

[0036] Step 3: Alternately train the interfering agent and the opponent agent.

[0037] In summary, the alternating training steps of the embodiment of the present invention include: (1) in the current iteration, using the opponent intelligent agent to perform data extrapolation and adding the generated data to the extended data set; (2) using the original data set and the extended data set to train the interfering intelligent agent; (3) training the opponent intelligent agent according to the offline adversarial learning method in combination with the interfering intelligent agent trained in the current iteration; (4) continuously iterating the above (1) to (3) to obtain the final interfering intelligent agent.

[0038] Step 4: deploy the final jammer agent in a real environment for radar jamming decision making.

[0039] Preferably, the final interfering agent is deployed in a real environment, and the strategy of the final interfering agent can be analyzed for performance to verify the effectiveness of the lower bound of the strategy performance. The above scheme provides a priori performance results for the deployment of the strategy in a real environment, making the deployment of the interfering party in a complex electromagnetic environment safe and efficient.

[0040] In order to more clearly demonstrate the technical solution and technical effects provided by the present invention, the method provided by the embodiment of the present invention is described in detail with reference to specific embodiments below.

[0041] 1. Introduction to the principles of the plan.

[0042] In actual electronic confrontation scenarios, the interference party's cognition of the electromagnetic environment is incomplete and biased, and the performance of the interference decision based on this is difficult to guarantee in the real environment. To address this problem, the present invention constructs an opponent intelligent agent through a historical game sequence under incomplete electromagnetic environment cognition. The interference party intelligent agent is trained in turn with the constructed adversarial intelligent agent without interacting with the real environment. Finally, a conservative environmental model and a performance-guaranteed interference strategy are obtained, which makes the interference party's decision-making in a complex electromagnetic environment efficient and safe.

[0043] Specifically: In the electronic countermeasure scenario, as the interfering party, the interfering party is usually modeled as an intelligent agent, and the electromagnetic environment in which it is located is regarded as the environment in reinforcement learning. At the same time, the present invention introduces the idea of ​​robust adversarial reinforcement learning (RL) and introduces a new intelligent agent, which represents the transfer function of the electromagnetic environment in which it is located, with the purpose of making the interfering party's benefits as small as possible (i.e., minimizing the benefits of the interfering party's intelligent agent). The optimization goal set here is:

[0044]

[0045] Among them, π is the strategy of the jammer agent (i.e., the jamming decision to be solved in the end), M is the electromagnetic environment in which the jammer agent and the opponent agent (jammer and radar) are located, is the strategy of the opponent agent; Π, are the strategy π of the interfering agent and the strategy of the opponent agent respectively. A collection of; Indicates that the strategy of the interfering agent is π, and the strategy of the opponent agent is Interference value function of the interfering agent when the electromagnetic environment is M.

[0046] Combining this idea with offline reinforcement learning, the radar jamming decision problem of the jammer agent can be modeled as: In the case of, find a strategy π that satisfies the following constraints:

[0047]

[0048] in, is the opponent agent, is the interference value function when the interfering agent with strategy π competes with the opponent agent, is a set of state transition functions, expressed as:

[0049]

[0050] In the above formula, represents expectation, s represents state, and a represents action. represents the state transition law of the opponent agent, ξ represents the degree of variational distance closeness, due to and Both represent the state and reward at the next moment after taking action a in state s. Since the possible state and reward at the next moment are uncertain, they are represented by dots; TV(P1,P2) is the variational distance between distributions P1 and P2. It represents the maximum likelihood estimate of the given MDP of the historical game sequence. The above constraint formula defines a pessimistic electromagnetic environment model, because the purpose of this pessimistic electromagnetic environment model is to minimize the gain of the interferer.

[0051] In the embodiment of the present invention, the above optimization objective formula is equivalent to the above constraint formula, mainly because the present invention uses the strategy Abstract as the corresponding agent (i.e. the opponent agent)

[0052] Those skilled in the art can understand that the historical game sequence is a set of four-tuples {state, action, next moment state, reward}.

[0053] The performance of the strategy obtained from the pessimistic electromagnetic environment model is analyzed: T is used to represent the state transition function in the real environment. It can be proved that for any strategy π of the interfering agent, there is a probability of 1-σ:

[0054]

[0055] Among them, σ is a set value (a value close to 0), indicating that the above formula is very likely to be true. represents the interference value function when the interfering agent with strategy π confronts the opponent agent, The interference value function represents the interference value of the interference agent with strategy π when deployed in the real environment to fight against the actual opponent. The above formula indicates that the performance of any interference strategy in the real environment is at least as good as the performance of the strategy in the pessimistic model, that is, the performance of the strategy π in the pessimistic model is the lower bound of the performance of the strategy in the real environment.

[0056] 2. Detailed introduction of the plan.

[0057] Figure 2 The training process is shown; it mainly includes:

[0058] 1. Use historical game sequences as original data sets And normalize it.

[0059] 2. Using the original data set, a simple method (such as maximum likelihood estimation) is used to give the initial radar state transfer function (i.e., the opponent agent ).

[0060] 3. Alternately train the agent’s strategy and the opponent’s agent.

[0061] Each iteration consists of the following steps:

[0062] (1) Through the opponent agent Perform K-step extrapolation to generate new data and add it to the extended sequence set (i.e., extended dataset).

[0063] (2) Train the interfering agent. The training goal is to maximize the interference value function when the interfering agent with strategy π competes with the opponent agent. The training data comes from Use reinforcement learning to update the policy π and Q network of the interfering agent

[0064] (3) Train the opponent agent. The training goal is to minimize the interference value function when the interference agent with strategy π competes with the opponent agent. The opponent agent is updated using an offline adversarial learning method. Specifically, according to the offline adversarial learning method, the strategy of the interfering agent after the current iterative training is combined to calculate the strategy gradient, and the strategy gradient is used to guide the training of the opponent agent.

[0065] In the embodiment of the present invention, the update of strategy π can adopt conventional reinforcement learning methods (for example, actor-critic algorithm), so it will not be described in detail. The key issue is how to update the strategy π from a given historical game sequence. Starting from, an opponent intelligent agent is constructed to represent the transfer function of the electromagnetic environment. To represent the opponent agent, is the probability of obtaining reward r and transferring to state s′ after executing (s, a). The present invention represents the opponent agent through an integration of neural networks, each of which generates a Gaussian distribution about the next state and reward: represents the Gaussian distribution of obtaining reward r and transferring to state s′ after executing (s, a), u φ (s,a) is the mean corresponding to (s,a), ∑ φ The standard deviation corresponding to (s,a) is Indicates that the mean is u φ (s,a), the standard deviation is ∑ φ (s,a) is a Gaussian distribution, (s,a) means taking action a when in state s. The action of the opponent agent is the next state and interference benefit detected by the jammer from the electromagnetic environment. Under different action strategies, the electromagnetic environment state information and interference benefits obtained by the jammer are different. The purpose of the opponent agent is to make the jammer's benefits (i.e. ) as small as possible. In order to achieve the above goals, the present invention uses the policy gradient method. For the opponent agent, the actual work to be done is to minimize the value function of the policy π (that is, the value function of the policy π mentioned above). ), so this value function can in turn be used as an indicator to guide the training of the opponent agent. The policy gradient can be calculated as:

[0066]

[0067] in, For expectations, Indicates that state s obeys The distribution of represents the access distribution of state s under strategy π, that is, the steady-state distribution or weighted distribution of the state under a given strategy π, which depends on the transition probability distribution and strategy π; a~π means that the action selection strategy is π, Indicates that the state s′ and reward r at the next moment come from the Gaussian distribution of the opponent agent γ is the setting coefficient; and The meaning of is the same. Here, the neural network model is used to represent the opponent agent, and φ is the neural network model parameter, that is, Represents the interference value function when the interfering agent with strategy π confronts the neural network model, represents the gradient with respect to φ.

[0068] In addition, for the opponent agent, the training loss function also needs to consider being close to the maximum likelihood estimate on the data set. The training loss function can be expressed as:

[0069]

[0070] in, It is related to the maximum likelihood estimate given by the historical game sequence, λ is the set coefficient, Indicates that (s, a, r, s′) comes from the original dataset (i.e. historical game sequence).

[0071] Through iterative alternating training steps, we can finally get a trained interference agent, which includes a strategy π and Q network The strategy π is a performance-guaranteed strategy, and the Q network The corresponding value function is a pessimistic value function, which is the lower bound of the value function in the real environment, and can reduce the number of interactions during the training process. Specifically: during online training, the jammer (corresponding to the jammer agent) has to interact with the radar to gain experience, and the present invention uses historical data to build an opponent agent, and then the radar interacts with this opponent agent, reducing the number of interactions between the jammer and the radar. The interaction between the two is only in the data collection stage and provides valuable reference for the deployment of the strategy in the real environment. The strategy π is deployed to the environment for operation, and its performance is analyzed. The analysis results are compared with the performance in the pessimistic model to verify the effectiveness of the lower bound of the strategy performance. This method provides a priori performance results for the deployment of the strategy in the real environment, making the deployment of the reinforcement learning algorithm in a complex electromagnetic environment safe and efficient.

[0072] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above embodiments can be implemented by software, or by means of software plus necessary general hardware platforms. Based on such understanding, the technical solutions of the above embodiments can be embodied in the form of software products, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0073] Embodiment 2

[0074] The present invention also provides an electronic countermeasure offline decision-making system, which is mainly used to implement the method provided in the above embodiment, such as Figure 3 As shown, the system mainly includes:

[0075] A jammer agent modeling unit, used for modeling the jammer as a jammer agent;

[0076] The opponent agent construction unit is used to take the given historical game sequence as the original data set and use it to construct the opponent agent;

[0077] The agent alternating training unit is used to alternately train the interfering agent and the opponent agent, and the steps include: in the current iteration, using the opponent agent to perform data extrapolation, adding the generated data to the extended data set, using the original data set and the extended data set to train the interfering agent, and the training goal is to maximize the interference value function when the interfering agent confronts the opponent agent; according to the offline adversarial learning method combined with the training of the interfering agent trained in the current iteration, the opponent agent is trained, and the training goal is to minimize the interference value function when the interfering agent confronts the opponent agent; iterate continuously to obtain the final interfering agent;

[0078] The deployment application unit is used to deploy the final jammer intelligent agent in a real environment to make radar jamming decisions.

[0079] Considering that the relevant technical details involved in the system have been introduced in detail in the previous embodiments, they will not be repeated here.

[0080] Technical personnel in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the system can be divided into different functional modules to complete all or part of the functions described above.

[0081] Embodiment 3

[0082] The present invention also provides a processing device, such as Figure 4 As shown, it mainly includes: one or more processors; a memory for storing one or more programs; wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the methods provided in the aforementioned embodiments.

[0083] Furthermore, the processing device also includes at least one input device and at least one output device; in the processing device, the processor, memory, input device, and output device are connected via a bus.

[0084] In the embodiment of the present invention, the specific types of the memory, input device and output device are not limited; for example:

[0085] The input device may be a touch screen, an image acquisition device, a physical button or a mouse, etc.;

[0086] The output device may be a display terminal;

[0087] The memory may be a random access memory (RAM) or a non-volatile memory, such as a disk memory.

[0088] Embodiment 4

[0089] The present invention also provides a readable storage medium storing a computer program, which implements the method provided in the above embodiment when the computer program is executed by a processor.

[0090] In the embodiment of the present invention, the readable storage medium is a computer-readable storage medium and can be set in the aforementioned processing device, for example, as a memory in the processing device. In addition, the readable storage medium can also be a U disk, a mobile hard disk, a read-only memory (ROM), a disk or an optical disk, etc., which can store program codes.

[0091] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by any technician familiar with the technical field within the technical scope disclosed in the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims. The information disclosed in the background technology section of this article is only intended to deepen the understanding of the overall background technology of the present invention, and should not be regarded as an admission or in any form that the information constitutes prior art known to those skilled in the art.

Claims

1. An electronic countermeasure offline decision-making method, characterized in that: include: Model the interferer as an interferer agent; Take the given historical game sequence as the original data set and use it to construct the opponent agent; The interfering agent and the opponent agent are alternately trained, the steps comprising: in the current iteration, using the opponent agent to perform data extrapolation, adding the generated data to the extended data set, using the original data set and the extended data set to train the interfering agent, the training goal of which is to maximize the interference value function when the interfering agent confronts the opponent agent; training the opponent agent according to the offline adversarial learning method combined with the training of the interfering agent trained in the current iteration, the training goal of which is to minimize the interference value function when the interfering agent confronts the opponent agent; Continue to iterate and obtain the final interfering agent; The final jammer agent is deployed in a real environment for radar jamming decision-making.

2. The electronic countermeasure offline decision-making method according to claim 1, characterized in that: The constructing of the opponent intelligent agent includes: constructing the opponent intelligent agent according to the original data set through maximum likelihood estimation.

3. The electronic countermeasure offline decision-making method according to claim 1, characterized in that: The radar jamming decision-making problem of the jammer agent is modeled as: given a historical game sequence, find a strategy π that satisfies the following constraints: Among them, π is the strategy of the interfering agent, Π is the set of strategies π of the interfering agent, is the transfer function of the opponent agent, representing the electromagnetic environment during training; is the set of state transition functions, is the interference value function when the interfering agent with strategy π competes with the opponent agent.

4. The electronic countermeasure offline decision-making method according to claim 1, characterized in that: The training of the opponent agent according to the offline adversarial learning method in combination with the jammer agent after the current iteration training comprises: The offline adversarial learning method is combined with the strategy of the interfering agent after the current iteration training to calculate the strategy gradient, and the strategy gradient is used to guide the training of the opponent agent; Among them, the calculation method of the policy gradient is expressed as: in, For expectations, Indicates that state s obeys The distribution of represents the access distribution of state s under strategy π, a~π means the action selection strategy is π, represents the interference value function when the interference agent with strategy π confronts the neural network model, the neural network model represents the opponent agent, φ is the neural network model parameter, Indicates that the state s′ and reward r at the next moment come from the Gaussian distribution of the opponent agent It indicates that after executing (s, a), the reward r is obtained and the state s′ is transferred to the Gaussian distribution. (s, a) indicates that action a is taken in state s. γ is the setting coefficient. represents the gradient with respect to φ.

5. An electronic countermeasure offline decision-making method according to claim 1 or 5, characterized in that: The training loss function of the opponent agent is expressed as: in, is the training loss function of the opponent agent, λ is the setting coefficient, represents the Gaussian distribution of obtaining reward r and transferring to state s′ after executing (s, a), Represents (s, a, r, s′) from the original dataset 6. The electronic countermeasure offline decision-making method according to claim 1, characterized in that: Also includes: The final interfering agent is deployed in a real environment, and the performance of the strategy of the final interfering agent is analyzed to verify the effectiveness of the lower bound of the strategy performance.

7. The electronic countermeasure offline decision-making method according to claim 1, characterized in that: When analyzing the performance of the final interfering agent's strategy, T is used to represent the state transition function in the real environment. For any strategy, there is a 1-σ probability: Among them, σ is a set value to make the above equation valid. Represents the interfering agent and the opponent agent with strategy π Interference value function during confrontation, opponent agent The transfer function representing the electromagnetic environment during training, is the set of state transition functions, Represents the interference value function when the interfering agent with strategy π is deployed in a real environment to compete with an actual opponent.

8. An electronic countermeasure offline decision-making system, characterized in that: include: A jammer agent modeling unit, used for modeling the jammer as a jammer agent; The opponent agent construction unit is used to take the given historical game sequence as the original data set and use it to construct the opponent agent; The agent alternating training unit is used to alternately train the interfering agent and the opponent agent, and the steps include: in the current iteration, using the opponent agent to perform data extrapolation, adding the generated data to the extended data set, using the original data set and the extended data set to train the interfering agent, and the training goal is to maximize the interference value function when the interfering agent confronts the opponent agent; according to the offline adversarial learning method combined with the training of the interfering agent trained in the current iteration, the opponent agent is trained, and the training goal is to minimize the interference value function when the interfering agent confronts the opponent agent; iterate continuously to obtain the final interfering agent; The deployment application unit is used to deploy the final jammer intelligent agent in a real environment to make radar jamming decisions.

9. A processing device, characterized in that: include: one or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

10. A readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images