Bionic swarm cooperative combat environment noise intelligent countermeasure method and system

Abstract

The application provides a bionic group cooperative combat environment noise intelligent countermeasure method and system, relates to the technical field of electronic countermeasure, and comprises the following steps: collecting multi-dimensional noise signals, performing time-frequency domain decomposition to obtain a feature matrix, constructing a bionic group cooperative model containing multiple cooperative units, establishing a local countermeasure strategy and forming a global countermeasure knowledge base, mapping the countermeasure strategy and a target area space to generate a scheme set, generating adaptive countermeasure instructions through a distributed decision mechanism, implementing noise suppression, and performing effect evaluation and optimization. The application realizes accurate identification and adaptive suppression of combat environment noise, and improves the electronic countermeasure capability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electronic countermeasures, in particular to a bionic group cooperative combat environment noise intelligent countermeasure method and system. BACKGROUND

[0002] Traditional noise countermeasure technology mainly adopts methods such as filtering suppression, shielding isolation and frequency avoidance, but with the increasing complexity of the environment and the diversification of noise sources, more intelligent and adaptive noise countermeasure technology is needed.

[0003] The existing environmental noise countermeasure technology has the following shortcomings: lack of comprehensive analysis ability of multi-dimensional noise signals, most systems only process single type noise, it is difficult to cope with the comprehensive interference of multi-source and multi-type noise in complex environment, which leads to limited countermeasure effect. Secondly, the existing technology mostly uses fixed algorithm and preset strategy, and the adaptive ability is insufficient. When facing the dynamically changing noise environment, it is difficult to adjust the countermeasure strategy in real time, which may cause the countermeasure effect to decline or even fail. SUMMARY

[0004] The bionic group cooperative combat environment noise intelligent countermeasure method and system provided by the embodiments of the present application can solve the problems in the prior art.

[0005] In a first aspect, the bionic group cooperative combat environment noise intelligent countermeasure method is provided, which comprises:

[0006] Collecting multi-dimensional noise signals in the combat environment, and performing time-frequency domain decomposition on the multi-dimensional noise signals to obtain a noise feature matrix;

[0007] Building a group cooperative model based on a bionic mechanism, the group cooperative model comprising a plurality of cooperative units, each cooperative unit establishing a local countermeasure strategy according to a feature component in the noise feature matrix, and forming a global countermeasure knowledge base through information interaction between the cooperative units;

[0008] According to the type of the feature component in the noise feature matrix, the countermeasure strategies in the global countermeasure knowledge base are spatially mapped to target regions in the combat environment to obtain a set of regional countermeasure schemes;

[0009] According to the set of regional countermeasure schemes and the combat task constraint conditions, self-adaptive countermeasure instructions are generated through the distributed decision mechanism of each cooperative unit in the group cooperative model;

[0010] Based on the self-adaptive countermeasure instructions, the noise suppression operation on the target regions is driven by the countermeasure device, and countermeasure effect evaluation information is collected. The local countermeasure strategy and the global countermeasure knowledge base are dynamically updated using the countermeasure effect evaluation information to form a closed loop optimization.

[0011] Each collaborative unit establishes a local adversarial strategy based on the feature components in the noise feature matrix, and forms a global adversarial knowledge base through information exchange between collaborative units, including:

[0012] Extract feature components of different frequency bands from the noise feature matrix, and divide the multiple cooperative units into multiple cooperative levels according to the energy distribution characteristics of the feature components;

[0013] Within each collaborative level, each collaborative unit constructs a local feature space based on the assigned feature components, and iteratively generates an initial local adversarial strategy in the local feature space through a biomimetic evolution mechanism.

[0014] Establish an information interaction topology structure between collaborative units, wherein the information interaction topology structure defines the information transmission path and collaborative weight between each collaborative unit;

[0015] Each collaborative unit transmits the initial local adversarial strategy to its neighboring collaborative units through the information interaction topology, and receives local adversarial strategies from its neighboring collaborative units. Based on the received local adversarial strategies and the collaborative weights, the initial local adversarial strategy is fused and corrected to obtain the corrected local adversarial strategy.

[0016] The modified local adversarial strategies are aggregated according to the cooperation level, and the global adversarial knowledge base is formed through the strategy consistency constraint mechanism between levels.

[0017] Within each collaborative level, each collaborative unit constructs a local feature space based on its assigned feature components, and iteratively generates an initial local adversarial strategy within this local feature space through a biomimetic evolutionary mechanism, including:

[0018] Each collaborative unit constructs a local feature space based on the assigned feature components. The local feature space includes feature dimensions and feature boundaries. The feature dimensions are determined by the vector dimensions of the feature components, and the feature boundaries are determined by the value range of each component in the feature components.

[0019] Multiple candidate adversarial strategies are initialized in the local feature space, and each candidate adversarial strategy includes a set of adversarial parameters and a sequence of adversarial actions.

[0020] A biomimetic fitness evaluation function is constructed, which calculates the fitness value of each candidate adversarial strategy based on the feature components and the combat mission objective.

[0021] The multiple candidate adversarial strategies are iteratively optimized using a biomimetic evolution mechanism. The iterative optimization process includes: selecting candidate adversarial strategies based on the fitness value, mutating the set of adversarial parameters of the selected candidate adversarial strategies, and recombining the adversarial action sequence of the selected candidate adversarial strategies to generate a new generation of candidate adversarial strategies.

[0022] Repeat the iterative optimization process until the fitness value of the new generation of candidate adversarial strategies meets the preset convergence condition, and determine the candidate adversarial strategy that meets the preset convergence condition as the initial local adversarial strategy.

[0023] Based on the feature component types in the noise feature matrix, the adversarial strategies in the global adversarial knowledge base are spatially mapped to target areas in the operational environment to obtain a set of regional adversarial schemes, including:

[0024] The feature component type identifier is extracted from the noise feature matrix. The feature component type identifier represents the frequency domain attribute and energy distribution attribute of each feature component.

[0025] Acquire spatial information of a target area in the combat environment, wherein the spatial information of the target area includes the spatial coordinates and spatial extent of the target area;

[0026] Establish a mapping association rule between feature component types and target regions, wherein the mapping association rule defines the correspondence between feature component type identifiers and target region spatial information;

[0027] According to the mapping association rules, the feature component type identifiers associated with each adversarial strategy in the global adversarial knowledge base are matched with the target region spatial information to determine the applicable target region for each adversarial strategy.

[0028] For each target region, an adversarial strategy that is applicable to the target region and matches the target region is selected from the global adversarial knowledge base to form a set of candidate adversarial strategies for the target region;

[0029] Based on the feature component type identifier of each adversarial strategy in the candidate adversarial strategy set and the spatial range of the target area, calculate the spatial coverage of each adversarial strategy in the target area.

[0030] Based on the spatial coverage, the adversarial strategies in the candidate adversarial strategy set are combined and optimized to generate a regional adversarial scheme for the target area. The regional adversarial schemes for all target areas are then summarized to form the regional adversarial scheme set.

[0031] Based on the aforementioned set of regional confrontation schemes and operational mission constraints, adaptive confrontation instructions are generated through the distributed decision-making mechanism of each collaborative unit in the group collaboration model, including:

[0032] The regional confrontation scheme for each target region in the set of regional confrontation schemes is decomposed into multiple confrontation sub-tasks, and each confrontation sub-task includes confrontation actions and execution sequence.

[0033] Based on the operational mission constraints, a feasibility assessment is performed on the multiple adversarial sub-missions, and a set of candidate adversarial sub-missions that meet the constraints is selected.

[0034] The candidate adversarial subtask set is assigned to the corresponding collaborative units in the group collaborative model according to the target region, and each collaborative unit receives the assigned adversarial subtask;

[0035] Each coordinating unit independently generates local adversarial instructions based on the received adversarial sub-tasks and the constraints of the combat mission through a distributed decision-making mechanism. The local adversarial instructions include the control parameters of the adversarial device and the execution time.

[0036] A command negotiation mechanism is established between collaborative units. Each collaborative unit sends the local confrontation command to the adjacent collaborative unit through the command negotiation mechanism and receives the local confrontation command from the adjacent collaborative unit.

[0037] Each coordinating unit performs conflict detection and timing adjustment on its own local adversarial commands based on the local adversarial commands received from neighboring coordinating units, and obtains coordinated local adversarial commands.

[0038] The coordinated local adversarial instructions of all collaborative units are integrated, and the integrated instruction sequence is prioritized to generate the adaptive adversarial instructions.

[0039] The adaptive adversarial command drives the adversarial device to perform noise suppression operations on the target area, and collects adversarial effect evaluation information. The adversarial effect evaluation information is then used to dynamically update the local adversarial strategy and the global adversarial knowledge base, including:

[0040] The adaptive countermeasure instruction is parsed, and the countermeasure device control parameters and execution time are extracted. A timing trigger signal for the countermeasure device is generated based on the execution time. Based on the timing trigger signal and the countermeasure device control parameters, the countermeasure device is driven to perform noise suppression operation on the target area according to the predetermined execution time.

[0041] During the noise suppression operation, noise signal change data of the target area is collected in real time, and the countermeasure effect evaluation index is calculated based on the noise signal change data. The countermeasure effect evaluation index includes noise suppression rate and countermeasure response delay.

[0042] The adversarial effect evaluation index is compared with the expected adversarial effect in the adaptive adversarial instruction to generate adversarial effect deviation information; adversarial effect evaluation information is constructed based on the adversarial effect deviation information, which includes adversarial effect deviation information, noise signal change data and execution time correlation records;

[0043] The adversarial effect deviation information and noise signal change data in the adversarial effect evaluation information are fed back to each cooperative unit in the group cooperative model; each cooperative unit corrects the adversarial parameter set and adversarial action sequence according to the received adversarial effect deviation information to obtain the updated local adversarial strategy.

[0044] Each collaborative unit transmits and merges the updated local adversarial strategy through an information interaction topology, replaces and updates the adversarial strategy in the global adversarial knowledge base, and completes the dynamic update of the local adversarial strategy and the global adversarial knowledge base.

[0045] A second aspect of the present invention provides a biomimetic swarm cooperative intelligent anti-noise system for combat environments, comprising:

[0046] The first unit is used to collect multidimensional noise signals in the combat environment and perform time-frequency domain decomposition on the multidimensional noise signals to obtain a noise feature matrix;

[0047] The second unit is used to construct a group collaboration model based on a biomimetic mechanism. The group collaboration model contains multiple collaboration units. Each collaboration unit establishes a local adversarial strategy based on the feature components in the noise feature matrix and forms a global adversarial knowledge base through information interaction between collaboration units.

[0048] The third unit is used to spatially map the adversarial strategies in the global adversarial knowledge base to the target areas in the combat environment based on the type of feature components in the noise feature matrix, so as to obtain a set of regional adversarial schemes.

[0049] The fourth unit is used to generate adaptive countermeasure instructions based on the set of regional confrontation schemes and operational mission constraints, through the distributed decision-making mechanism of each cooperative unit in the group cooperation model.

[0050] The fifth unit is used to drive the adversarial device to perform noise suppression operations on the target area based on the adaptive adversarial command, and to collect adversarial effect evaluation information. The adversarial effect evaluation information is used to dynamically update the local adversarial strategy and the global adversarial knowledge base to form a closed-loop optimization.

[0051] A third aspect of the present invention,

[0052] An electronic device is provided, comprising:

[0053] processor;

[0054] Memory used to store processor-executable instructions;

[0055] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0056] Fourth aspect of the embodiments of the present invention,

[0057] A computer-readable storage medium is provided, having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0058] The beneficial effects of this application are as follows:

[0059] This invention achieves intelligent countermeasure processing of multidimensional noise signals through a biomimetic swarm collaboration mechanism, exhibiting stronger environmental adaptability and countermeasure effectiveness compared to existing technologies. Each collaborative unit can independently construct local countermeasure strategies and form a global countermeasure knowledge base through information exchange, achieving a collective enhancement of noise countermeasure capabilities and significantly improving the system's ability to cope with complex noise environments.

[0060] By employing time-frequency domain decomposition and spatial mapping techniques, countermeasures can accurately correspond to different types of noise characteristic components and their distribution in the target area, improving the targeting and effectiveness of countermeasures and significantly reducing resource waste. Especially in complex operational environments where noise characteristics change rapidly, strategies can be quickly adjusted to maintain continuous and effective countermeasure capabilities. Attached Figure Description

[0061] Figure 1 This is a flowchart illustrating the intelligent countermeasures method for combat environment noise based on biomimetic swarm collaboration, as described in an embodiment of the present invention. Detailed Implementation

[0062] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0064] Figure 1This is a flowchart illustrating the biomimetic swarm cooperative intelligent countermeasure method against combat environment noise according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0065] Collect multidimensional noise signals from the combat environment and perform time-frequency domain decomposition on the multidimensional noise signals to obtain a noise feature matrix;

[0066] A group collaboration model is constructed based on a biomimetic mechanism. The group collaboration model contains multiple collaboration units. Each collaboration unit establishes a local adversarial strategy based on the feature components in the noise feature matrix, and forms a global adversarial knowledge base through information interaction between collaboration units.

[0067] Based on the feature component types in the noise feature matrix, the adversarial strategies in the global adversarial knowledge base are spatially mapped to the target areas in the combat environment to obtain a set of regional adversarial schemes;

[0068] Based on the set of regional confrontation schemes and the constraints of the combat mission, adaptive confrontation instructions are generated through the distributed decision-making mechanism of each collaborative unit in the group collaboration model.

[0069] Based on the adaptive adversarial command, the adversarial device performs noise suppression operations on the target area and collects adversarial effect evaluation information. The adversarial effect evaluation information is used to dynamically update the local adversarial strategy and the global adversarial knowledge base to form a closed-loop optimization.

[0070] In one optional implementation, each coordinating unit establishes a local adversarial strategy based on the feature components in the noise feature matrix, and forms a global adversarial knowledge base through information exchange between coordinating units, including:

[0071] Extract feature components of different frequency bands from the noise feature matrix, and divide the multiple cooperative units into multiple cooperative levels according to the energy distribution characteristics of the feature components;

[0072] Within each collaborative level, each collaborative unit constructs a local feature space based on the assigned feature components, and iteratively generates an initial local adversarial strategy in the local feature space through a biomimetic evolution mechanism.

[0073] Establish an information interaction topology structure between collaborative units, wherein the information interaction topology structure defines the information transmission path and collaborative weight between each collaborative unit;

[0074] Each collaborative unit transmits the initial local adversarial strategy to its neighboring collaborative units through the information interaction topology, and receives local adversarial strategies from its neighboring collaborative units. Based on the received local adversarial strategies and the collaborative weights, the initial local adversarial strategy is fused and corrected to obtain the corrected local adversarial strategy.

[0075] The modified local adversarial strategies are aggregated according to the cooperation level, and the global adversarial knowledge base is formed through the strategy consistency constraint mechanism between levels.

[0076] A noise feature matrix is ​​obtained, which contains characteristic component information of the noise signal in different frequency bands. To effectively process this feature information, characteristic components of different frequency bands are extracted from the noise feature matrix, and the energy distribution characteristics of these characteristic components are analyzed. Based on the energy distribution characteristics, multiple cooperative units are divided into multiple cooperative levels. For example, when the noise feature matrix contains characteristic components of three frequency bands—low frequency (0-500Hz), mid frequency (500-2000Hz), and high frequency (2000-8000Hz)—the cooperative units can be divided into three levels according to the energy distribution: a base layer, an intermediate layer, and a top layer. If the energy proportion of the low-frequency component exceeds 50%, the cooperative unit processing this frequency band is assigned to the base layer; if the energy proportion of the mid-frequency component is between 30% and 50%, the corresponding cooperative unit is assigned to the intermediate layer; if the energy proportion of the high-frequency component is less than 30%, the corresponding cooperative unit is assigned to the top layer.

[0077] Within each collaborative layer, each collaborative unit constructs a local feature space based on its assigned feature components. Taking the base layer as an example, after a collaborative unit is assigned feature components in the 0-200Hz frequency band, it extracts parameters such as amplitude, phase, and time-varying characteristics within that band, constructing a set of feature vectors containing these parameters, forming a local feature space. Within this local feature space, the collaborative unit iteratively generates initial local adversarial strategies through a biomimetic evolutionary mechanism. This mechanism simulates the biological evolution process, including four steps: strategy generation, evaluation, selection, and mutation. In the strategy generation stage, the collaborative unit randomly generates multiple sets of noise adversarial parameter combinations. For example, for low-frequency noise in the 0-200Hz range, it may generate 20 initial strategies containing parameters such as filter strength, phase compensation, and time-domain modulation. In the evaluation stage, the noise suppression effect of each strategy in the local feature space is quantitatively evaluated, such as calculating the percentage reduction in noise energy. In the selection stage, the 10 best-performing strategies are retained. In the mutation stage, these 10 strategies undergo parameter fine-tuning and cross-combination to generate a new generation of 20 strategies. Through multiple iterations (e.g., 50 times), the initial local adversarial strategy that performs best in the local feature space is finally obtained.

[0078] Next, an information interaction topology is established between the collaborative units, defining the information transmission paths and collaborative weights between each unit. In a system containing nine collaborative units, a 3×3 grid topology can be constructed, with each collaborative unit connected to its neighboring units. The connection weights are determined based on feature similarity. For example, the connection weight between collaborative units processing adjacent frequency bands is set to 0.8, while the weight between units processing significantly different frequency bands is set to 0.4. This topology ensures efficient information transmission between highly correlated collaborative units.

[0079] Each coordinating unit transmits its initial local adversarial strategy to its neighboring units through the aforementioned information exchange topology, while simultaneously receiving local adversarial strategies from its neighbors. Taking the coordinating unit at the center of the grid as an example, it receives adversarial strategies from its four neighboring units (top, bottom, left, and right), and calculates a weighted average of these strategies according to predefined coordinating weights, then merges this average with its own initial strategy. Specifically, for each parameter in the strategy, a weighted average is calculated as the merged parameter value. For example, if the central unit's filter strength parameter is 0.65, and the corresponding parameters of the four neighboring units are 0.7, 0.6, 0.68, and 0.62, with weights of 0.8, 0.8, 0.4, and 0.4 respectively, then the merged filter strength parameter is 0.65×0.5+0.7×0.16+0.6×0.16+0.68×0.08+0.62×0.08=0.653 (assuming its own strategy weight is 0.5). In this way, the collaborative unit generates a modified local adversarial strategy that retains the specificity of its own feature space while incorporating the advantageous strategies of neighboring units.

[0080] The modified local adversarial strategies are aggregated according to cooperative levels, and a global adversarial knowledge base is formed through a strategy consistency constraint mechanism between levels. During the aggregation process, the modified local adversarial strategies of all units within each cooperative level are first merged to generate hierarchical strategies. For example, the base layer contains four cooperative units, each responsible for a specific low-frequency band. The modified strategies of these four units are merged to form a hierarchical strategy covering the entire low-frequency band. The strategy consistency constraint mechanism between levels ensures that strategies of different levels maintain consistency in overlapping areas, avoiding strategy conflicts. For example, at the frequency band boundary between the base layer and the intermediate layer near 500Hz, the strategy parameters of the two levels need a smooth transition. Gradual weights are applied in the 480-520Hz range to smoothly transition the strategy parameters from the base layer values ​​to the intermediate layer values. In this way, a global adversarial knowledge base containing noise adversarial strategies for each frequency band is ultimately formed, capable of effectively suppressing complex noise environments.

[0081] This global adversarial knowledge base can be stored and dynamically updated. In practical applications, the most suitable combination of adversarial strategies can be retrieved from the knowledge base according to changes in the noise environment, achieving intelligent noise adversarial control. Actual tests show that the adversarial system built using this method can achieve a noise suppression effect of 15-25dB in complex noise environments, which is 40% more efficient than traditional single-strategy methods.

[0082] In one optional implementation, within each collaborative level, each collaborative unit constructs a local feature space based on its assigned feature components, and iteratively generates an initial local adversarial strategy within the local feature space through a biomimetic evolutionary mechanism, including:

[0083] Each collaborative unit constructs a local feature space based on the assigned feature components. The local feature space includes feature dimensions and feature boundaries. The feature dimensions are determined by the vector dimensions of the feature components, and the feature boundaries are determined by the value range of each component in the feature components.

[0084] Multiple candidate adversarial strategies are initialized in the local feature space, and each candidate adversarial strategy includes a set of adversarial parameters and a sequence of adversarial actions.

[0085] A biomimetic fitness evaluation function is constructed, which calculates the fitness value of each candidate adversarial strategy based on the feature components and the combat mission objective.

[0086] The multiple candidate adversarial strategies are iteratively optimized using a biomimetic evolution mechanism. The iterative optimization process includes: selecting candidate adversarial strategies based on the fitness value, mutating the set of adversarial parameters of the selected candidate adversarial strategies, and recombining the adversarial action sequence of the selected candidate adversarial strategies to generate a new generation of candidate adversarial strategies.

[0087] Repeat the iterative optimization process until the fitness value of the new generation of candidate adversarial strategies meets the preset convergence condition, and determine the candidate adversarial strategy that meets the preset convergence condition as the initial local adversarial strategy.

[0088] Each coordinating unit first receives the assigned feature components, which may include target identification features, threat assessment features, resource allocation features, etc. For example, in a drone swarm, target distance and target speed can be assigned to the first-level coordinating unit, while threat level and defense capabilities can be assigned to the second-level coordinating unit.

[0089] Each coordinating unit constructs a local feature space based on the received feature components. The dimension of the feature space is directly determined by the vector dimension of the feature components. For example, if the feature components received by the coordinating unit include target distance, target velocity, and target direction, then the constructed local feature space is a three-dimensional space. The feature boundaries are determined by the value range of each component; for example, the target distance ranges from 0 to 100 kilometers, and the target direction ranges from 0 to 360 degrees, thus forming a three-dimensional feature space with clear boundaries.

[0090] In the constructed local feature space, multiple candidate adversarial strategies are initialized. Each candidate adversarial strategy consists of two parts: a set of adversarial parameters and a sequence of adversarial actions. The set of adversarial parameters includes specific parameters such as setting the attack distance threshold to 25 kilometers, the evasion angle to 45 degrees, and the resource allocation ratio to 0.7; the sequence of adversarial actions is a series of action instructions triggered by time or conditions, such as "reconnaissance first, then attack, and finally evade." During the initialization phase, 20-100 different candidate adversarial strategies can be generated randomly in the feature space to ensure the diversity of the initial population.

[0091] Subsequently, a biomimetic fitness evaluation function is constructed to evaluate the effectiveness of each candidate adversarial strategy. This function is designed based on feature components and operational mission objectives. For example, in an interception mission, the interception success rate can be weighted at 0.6, resource consumption rate at 0.3, and time efficiency at 0.1, and the fitness value can be calculated comprehensively. In practical implementation, each candidate adversarial strategy can be simulated in a local feature space, and the execution results can be recorded. For example, if a strategy has an interception success rate of 85%, a resource consumption rate of 65%, and a time efficiency of 70%, then its fitness value is:

[0092] 0.6×85%+0.3×(100%-65%)+0.1×70%=75.5%.

[0093] The process enters the iterative optimization phase of the biomimetic evolutionary mechanism, optimizing candidate adversarial strategies based on Darwinian evolutionary theory. Each iteration includes three operations: selection, mutation, and recombination. The selection operation employs a roulette wheel method, where candidate adversarial strategies with higher fitness values ​​have a greater probability of being selected. For example, among 50 candidate strategies, the strategy with the highest fitness (80%) has a selection probability of approximately 5%, while a strategy with a fitness of only 40% has a selection probability of approximately 2.5%.

[0094] The mutation operation targets the set of adversarial parameters of the selected strategy and randomly adjusts the parameter values ​​with a certain probability. For example, if the attack distance threshold in the original strategy is 25 kilometers, it may be adjusted to 23 kilometers or 27 kilometers after mutation; the evasion angle may be mutated from 45 degrees to 50 degrees. The mutation probability is usually set to 10%-30%, and the mutation range can be set to 5%-20% of the parameter value.

[0095] The recombination operation targets the adversarial action sequence of the selected strategy, cross-combining the action sequences of the two parent strategies to generate a new action sequence. For example, if the action sequence of parent strategy A is "reconnaissance-attack-evasion" and the action sequence of parent strategy B is "concealment-flanking-attack", recombination may generate a new sequence of "reconnaissance-flanking-attack". The recombination point is usually randomly determined to ensure that the generated new sequence still has logical rationality.

[0096] Through selection, mutation, and recombination, an equal number of new generation candidate adversarial strategies are generated to replace the previous generation's strategy. The fitness values ​​of the new generation strategies are then recalculated to determine if they meet preset convergence conditions. Convergence conditions are typically set as follows: the fitness value of the optimal strategy in the population has not changed by more than 1% over 10 consecutive generations, or the preset maximum number of iterations (200) has been reached.

[0097] When the convergence condition is met, the candidate adversarial strategy with the highest fitness value in the population is determined as the initial local adversarial strategy. For example, in a drone swarm, after 157 iterations, the optimal strategy converges to "when the target distance is less than 30 kilometers, execute the attack action with a resource ratio of 0.8, and after the attack is completed, evade at a 60-degree angle while maintaining tracking of the target". The fitness value of this strategy reaches 92.3%, which meets the convergence condition that the change does not exceed 1% over 10 consecutive generations. Therefore, it is determined as the initial local adversarial strategy of this cooperative unit.

[0098] All collaborative units within each collaborative level generate their initial local adversarial strategies using the aforementioned method, laying the foundation for subsequent strategy collaborative optimization. This local feature space strategy generation method, based on a biomimetic evolutionary mechanism, fully utilizes the natural selection mechanism of biological evolution, enabling it to find near-optimal adversarial strategies in complex environments.

[0099] In one optional implementation, the adversarial strategies in the global adversarial knowledge base are spatially mapped to target areas in the operational environment based on the feature component types in the noise feature matrix, resulting in a set of regional adversarial schemes, including:

[0100] The feature component type identifier is extracted from the noise feature matrix. The feature component type identifier represents the frequency domain attribute and energy distribution attribute of each feature component.

[0101] Acquire spatial information of a target area in the combat environment, wherein the spatial information of the target area includes the spatial coordinates and spatial extent of the target area;

[0102] Establish a mapping association rule between feature component types and target regions, wherein the mapping association rule defines the correspondence between feature component type identifiers and target region spatial information;

[0103] According to the mapping association rules, the feature component type identifiers associated with each adversarial strategy in the global adversarial knowledge base are matched with the target region spatial information to determine the applicable target region for each adversarial strategy.

[0104] For each target region, an adversarial strategy that is applicable to the target region and matches the target region is selected from the global adversarial knowledge base to form a set of candidate adversarial strategies for the target region;

[0105] Based on the feature component type identifier of each adversarial strategy in the candidate adversarial strategy set and the spatial range of the target area, calculate the spatial coverage of each adversarial strategy in the target area.

[0106] Based on the spatial coverage, the adversarial strategies in the candidate adversarial strategy set are combined and optimized to generate a regional adversarial scheme for the target area. The regional adversarial schemes for all target areas are then summarized to form the regional adversarial scheme set.

[0107] In a specific embodiment, feature component type identifiers are extracted from the noise feature matrix. These identifiers characterize the frequency domain attributes and energy distribution attributes of each feature component. For example, a noise feature matrix N may contain multiple feature components n1, n2, ..., nk, each with its specific frequency domain attributes (e.g., low-frequency, mid-frequency, high-frequency) and energy distribution attributes (e.g., concentrated, dispersed). In practical applications, feature component n1 can be identified as "LF-C" to represent a low-frequency concentrated feature, and n2 can be identified as "HF-D" to represent a high-frequency dispersed feature. This identification method facilitates subsequent matching with spatial regions.

[0108] Acquire spatial information about target areas in the operational environment, including the spatial coordinates and spatial extent of the target area. In practice, a target area can be represented as a coordinate set R = {P1(x1, y1, z1), P2(x2, y2, z2), ..., Pm(xm, ym, zm)} and a spatial extent S = {length L, width W, height H}. For example, the center point of a target area R1 can be P(120.35, 36.28, 15.5), and the spatial extent can be S1 = {500 meters, 300 meters, 100 meters}. This precise spatial information provides the foundation for subsequent strategy mapping.

[0109] Establish mapping and association rules between feature component types and target regions, defining the correspondence between feature component type identifiers and target region spatial information. In practical applications, the mapping and association rules can be defined as an association matrix M, where M(i,j) represents the fit between feature component type i and target region j. For example, for feature component type "LF-C" and target region R1, if this type of feature is valid within region R1, then set M("LF-C", R1) = 0.85, representing an 85% fit. This mapping relationship can be determined comprehensively based on factors such as historical adversarial data, electromagnetic propagation characteristics, and terrain features.

[0110] Based on the established mapping and association rules, the feature component type identifiers associated with each adversarial strategy in the global adversarial knowledge base are matched with the target region spatial information to determine the applicable target region for each adversarial strategy. Assume the global adversarial knowledge base contains a strategy set S = {s1, s2, ..., sn}, and each strategy si is associated with a specific feature component type set T(si). The suitability of strategy si for region Rj is determined by calculating the average fit A(si, Rj) between each feature component type in T(si) and the target region Rj. When A(si, Rj) exceeds a preset threshold (e.g., 0.7), strategy si is determined to be an applicable strategy for region Rj.

[0111] For each target region, adversarial strategies that match the target region are selected from the global adversarial knowledge base, forming a set of candidate adversarial strategies for that target region. For example, for target region R1, the above matching process may yield a set of candidate adversarial strategies C(R1) = {s3, s7, s12, s25}. Each candidate strategy has a high degree of adaptability to region R1 and is suitable for deployment in that region.

[0112] Based on the characteristic component type identifiers of each adversarial strategy in the candidate adversarial strategy set and the spatial range of the target area, the spatial coverage of each adversarial strategy within the target area is calculated. Spatial coverage represents the effective coverage ratio of the adversarial strategy within the target area. In specific implementation, the effective coverage volume V(si, Rj) of strategy si in region Rj can be calculated according to its characteristic component type and propagation characteristics, and then divided by the total volume V(Rj) of region Rj to obtain the spatial coverage Cov(si, Rj) = V(si, Rj) / V(Rj). For example, if the effective coverage volume of strategy s7 in region R1 is 12,000,000 cubic meters, and the total volume of R1 is 15,000,000 cubic meters, then the spatial coverage of s7 in R1 is 0.8, or 80%.

[0113] Based on the calculated spatial coverage, adversarial strategies in the candidate adversarial strategy set are combined and optimized to generate a regional adversarial scheme for the target area. The optimization process considers multiple factors: strategy coverage, strategy complementarity, and resource consumption. A greedy algorithm is used, starting with the strategy with the highest spatial coverage and gradually adding complementary strategies that provide additional coverage until the required regional coverage is achieved (e.g., 95%) or resources are exhausted. For the target area R1, the optimized final adversarial scheme may be P(R1) = {s7, s12}, where s7 provides 80% basic coverage and s12 provides complementary coverage, together achieving 96% regional coverage.

[0114] All regional confrontation plans for all target areas are aggregated to form a set of regional confrontation plans. In practical applications, assuming the operational environment includes five target areas R1, R2, R3, R4, and R5, confrontation plans P(R1), P(R2), P(R3), P(R4), and P(R5) for each area are generated respectively, and these are aggregated into a total set of regional confrontation plans P = {P(R1), P(R2), P(R3), P(R4), P(R5)}. This set provides a basis for subsequent confrontation action planning and resource allocation.

[0115] In a practical application case, the above method was implemented in a combat zone with a complex electromagnetic environment. This zone contained three target sub-regions. By analyzing 15 feature component types in the noise feature matrix, applicable strategies were selected from a knowledge base containing 250 countermeasure strategies. Ultimately, a set of regional countermeasure schemes with a coverage rate of 94.3% was formed, which significantly improved the countermeasure effect and saved about 35% of resource consumption compared with traditional methods.

[0116] In one optional implementation, the adaptive countermeasure instructions are generated through a distributed decision-making mechanism of each cooperative unit in the group cooperative model, based on the set of regional confrontation schemes and operational mission constraints, including:

[0117] The regional confrontation scheme for each target region in the set of regional confrontation schemes is decomposed into multiple confrontation sub-tasks, and each confrontation sub-task includes confrontation actions and execution sequence.

[0118] Based on the operational mission constraints, a feasibility assessment is performed on the multiple adversarial sub-missions, and a set of candidate adversarial sub-missions that meet the constraints is selected.

[0119] The candidate adversarial subtask set is assigned to the corresponding collaborative units in the group collaborative model according to the target region, and each collaborative unit receives the assigned adversarial subtask;

[0120] Each coordinating unit independently generates local adversarial instructions based on the received adversarial sub-tasks and the constraints of the combat mission through a distributed decision-making mechanism. The local adversarial instructions include the control parameters of the adversarial device and the execution time.

[0121] A command negotiation mechanism is established between collaborative units. Each collaborative unit sends the local confrontation command to the adjacent collaborative unit through the command negotiation mechanism and receives the local confrontation command from the adjacent collaborative unit.

[0122] Each coordinating unit performs conflict detection and timing adjustment on its own local adversarial commands based on the local adversarial commands received from neighboring coordinating units, and obtains coordinated local adversarial commands.

[0123] The coordinated local adversarial instructions of all collaborative units are integrated, and the integrated instruction sequence is prioritized to generate the adaptive adversarial instructions.

[0124] Decomposing the regional countermeasure schemes is the first step in generating adaptive countermeasure instructions. A set of regional countermeasure schemes is received, each corresponding to a target region. Taking a certain electromagnetic countermeasure as an example, the countermeasure schemes targeting region A include two main categories of countermeasure actions: interference coverage and signal deception. This regional countermeasure scheme is decomposed into four countermeasure sub-tasks: band 1 jamming (execution time sequence T1-T3), band 2 jamming (execution time sequence T2-T5), false signal deployment 1 (execution time sequence T3-T6), and false signal deployment 2 (execution time sequence T4-T8). Each countermeasure sub-task clearly defines the specific countermeasure actions and their corresponding execution time windows.

[0125] Feasibility assessment of the decomposed adversarial sub-tasks is a crucial step in ensuring that adversarial instructions comply with mission constraints. Each adversarial sub-task is evaluated based on the operational mission constraints. These constraints include: energy constraints (total power not exceeding 50kW per unit time), time constraints (tasks must be completed within 30 minutes), and resource constraints (the number of available adversarial devices is 8). Through evaluation, a set of candidate adversarial sub-tasks that meet the constraints is selected. In the example above, due to the energy constraint, the partial time overlap between frequency band 1 jamming and frequency band 2 jamming was excluded, and the modified execution sequence was retained. Ultimately, the candidate adversarial sub-tasks for area A include: frequency band 1 jamming (T1-T2), frequency band 2 jamming (T3-T5), false signal deployment 1 (T3-T6), and false signal deployment 2 (T5-T8).

[0126] The task allocation phase assigns candidate adversarial subtasks to the various collaborative units within the group collaborative model. The set of candidate adversarial subtasks is distributed among the corresponding collaborative units according to the target area. In this implementation, there are 10 collaborative units distributed across different geographical locations. Collaborative Unit 1 and Collaborative Unit 2 are responsible for the adversarial tasks in area A. They respectively receive: Collaborative Unit 1 receives frequency band 1 jamming and false signal deployment task 1; Collaborative Unit 2 receives frequency band 2 jamming and false signal deployment task 2. The task allocation process considers the geographical location, equipment capabilities, and current load of each collaborative unit to ensure the rationality of the task allocation.

[0127] The distributed decision-making mechanism is the core of the system. Each coordinating unit independently generates local countermeasure instructions based on the received countermeasure sub-tasks and operational task constraints. Taking coordinating unit 1 as an example, for the frequency band 1 jamming task, the local countermeasure instructions it generates include: jamming device parameter settings (center frequency set to 2.4GHz, bandwidth set to 20MHz, power set to 15kW), and execution times from T1 (actual time is 13:20:00) to T2 (13:25:00). For the false signal delivery task 1, the generated instructions include: signal characteristic parameters (signal type is radar pulse signal, pulse width is set to 3μs, repetition frequency is set to 1000Hz) and execution times from T3 (13:26:00) to T6 (13:35:00).

[0128] The command negotiation mechanism among coordinating units ensures the overall coordination and consistency of the countermeasures effect. A communication network based on a mesh topology is established, through which each coordinating unit sends its own local countermeasures commands to neighboring coordinating units and receives commands from neighboring units. In the above example, coordinating unit 1 sends its generated local countermeasures commands for frequency band 1 interference and false signal projection 1 to coordinating units 2 and 3 (responsible for adjacent area B); simultaneously, coordinating unit 1 receives local countermeasures commands for frequency band 2 interference and false signal projection 2 from coordinating unit 2, and local countermeasures commands for frequency band 3 interference from coordinating unit 3.

[0129] The conflict detection and timing adjustment process ensures the coordinated execution of countermeasures actions by each unit. Each coordinating unit performs conflict detection on its own commands based on the local countermeasures instructions received from neighboring units. The detection includes spectrum resource conflicts, timing conflicts, and energy allocation conflicts. Taking coordinating unit 1 as an example, it discovers that its planned false signal deployment 1 (T3-T6) partially overlaps with coordinating unit 2's band 2 interference (T3-T5) in the spectrum (around 2.5GHz), potentially reducing the interference effect. Therefore, coordinating unit 1 adjusts the frequency parameters of false signal deployment 1 to a center frequency of 2.7GHz, avoiding potential mutual interference. Simultaneously, it slightly delays the execution time by 2 minutes to ensure that it is time-staggered with coordinating unit 3's band 3 interference, optimizing overall energy allocation.

[0130] Finally, the coordinated local adversarial instructions from all collaborative units are integrated and prioritized to generate the final adaptive adversarial instructions. During integration, all local adversarial instructions are arranged chronologically and prioritized based on the importance of the target area, adversarial effectiveness assessment, and resource utilization efficiency. In the above scenario, the final integrated instruction sequence for area A is: Band 1 jamming (T1-T2), Band 2 jamming (T3-T5), False signal deployment 1 (adjusted from T3+2min to T6+2min), and False signal deployment 2 (T5-T8). Each instruction is assigned a priority identifier, with Band 1 jamming given the highest priority (priority value 9 out of 10) because it targets the most critical target within area A and is executed first in the time sequence.

[0131] This method, through distributed decision-making and collaborative adjustment mechanisms, enables adaptive adversarial command generation for multi-target regions in complex adversarial environments, effectively solving the problems of slow response speed and poor adaptability of traditional centralized decision-making in dynamic environments.

[0132] In one optional implementation, the adaptive adversarial command drives the adversarial device to perform noise suppression operations on the target area, and collects adversarial effect evaluation information. The adversarial effect evaluation information is then used to dynamically update the local adversarial strategy and the global adversarial knowledge base, including:

[0133] The adaptive countermeasure instruction is parsed, and the countermeasure device control parameters and execution time are extracted. A timing trigger signal for the countermeasure device is generated based on the execution time. Based on the timing trigger signal and the countermeasure device control parameters, the countermeasure device is driven to perform noise suppression operation on the target area according to the predetermined execution time.

[0134] During the noise suppression operation, noise signal change data of the target area is collected in real time, and the countermeasure effect evaluation index is calculated based on the noise signal change data. The countermeasure effect evaluation index includes noise suppression rate and countermeasure response delay.

[0135] The adversarial effect evaluation index is compared with the expected adversarial effect in the adaptive adversarial instruction to generate adversarial effect deviation information; adversarial effect evaluation information is constructed based on the adversarial effect deviation information, which includes adversarial effect deviation information, noise signal change data and execution time correlation records;

[0136] The adversarial effect deviation information and noise signal change data in the adversarial effect evaluation information are fed back to each cooperative unit in the group cooperative model; each cooperative unit corrects the adversarial parameter set and adversarial action sequence according to the received adversarial effect deviation information to obtain the updated local adversarial strategy.

[0137] Each collaborative unit transmits and merges the updated local adversarial strategy through an information interaction topology, replaces and updates the adversarial strategy in the global adversarial knowledge base, and completes the dynamic update of the local adversarial strategy and the global adversarial knowledge base.

[0138] In this embodiment, the process of using adaptive adversarial commands to drive the adversarial device to perform noise suppression operations on the target area, collect adversarial effect evaluation information, and dynamically update the local adversarial strategy and the global adversarial knowledge base using the adversarial effect evaluation information is as follows.

[0139] First, the received adaptive countermeasure instructions are parsed by the parser module. These instructions contain key control information, such as the countermeasure device's operating frequency being set to the range of 200Hz-1200Hz, the signal strength parameter being set to an adjustable range of 45dB-75dB, and the execution time being from 10:30:00 to 10:45:00 on October 15, 2023. After extracting these control parameters, the parser module converts the execution time into an internal timestamp format, generating a sequence of timing trigger signals accurate to the millisecond level. For example, the execution interval is subdivided into trigger points every 50 milliseconds, forming a trigger signal sequence of 18,000 timing trigger points.

[0140] Based on the generated timing trigger signals and extracted control parameters, the abstract control parameters are converted into low-level control commands that the device can recognize through the instruction conversion module. After receiving these commands, the countermeasure device control module activates the countermeasure device according to the predetermined execution time. In actual operation, the countermeasure device adopts intelligent phase adjustment technology. When the noise frequency in the target area is detected to be 800Hz and the intensity is 68dB, it automatically generates a cancellation signal with opposite phase and matching amplitude to form directional acoustic wave countermeasures and perform noise suppression operation on the target area.

[0141] During the noise suppression operation, an acoustic sensor array deployed at four key locations in the target area collects ambient acoustic data in real time. Each sensor records the sound signal at a sampling rate of 96kHz. The raw data is then processed by a signal preprocessing unit for noise reduction and normalization. The effectiveness evaluation index is calculated based on the processed noise signal change data. Specifically, the noise suppression rate is calculated as the percentage change in signal strength before and after suppression. For example, after the operation, the noise in the target area decreased from 68dB to 23dB, resulting in a suppression rate of 66.2%. The countermeasure response delay is calculated by measuring the time difference between noise detection and the effective suppression effect; in this case, the response delay is 78 milliseconds.

[0142] The measured adversarial effect evaluation metrics were compared and analyzed with the expected adversarial effect in the adaptive adversarial command. The expected noise suppression rate was set at 70% and the expected response delay was 50 milliseconds, while the actual measured suppression rate was 66.2% and the response delay was 78 milliseconds. Based on this, adversarial effect deviation information was generated: suppression rate deviation -3.8% and response delay deviation +28 milliseconds. Subsequently, a complete adversarial effect evaluation information data package was constructed, including the above deviation information, noise signal change data records (including noise intensity and spectrum change data at 18,000 time points), and detailed execution time correlation records, forming a structured JSON format data package with a size of approximately 8.5MB.

[0143] The constructed adversarial effect evaluation information is distributed to each collaborative unit in the swarm collaborative model through a secure encrypted channel. In this embodiment, the swarm collaborative model consists of 8 collaborative units, each responsible for optimizing adversarial strategies for different frequency bands or scenarios. After receiving the adversarial effect deviation data in the mid-frequency band (600Hz-900Hz), the third collaborative unit finds that the noise suppression efficiency at the 800Hz frequency point is insufficient. It corrects the phase adjustment parameters through a built-in parameter optimization algorithm, adjusting the phase compensation value from the original 175 degrees to 182 degrees and increasing the signal gain by 2.3dB. After simulation verification, it is expected to improve the suppression rate to 69.5%. At the same time, the fifth collaborative unit optimizes the signal processing flow based on the response delay deviation data, reduces the filter order, shortens the signal processing delay by 21 milliseconds, and is expected to reduce the total response delay to 57 milliseconds. After these parameter and strategy adjustments, each collaborative unit generates an updated local adversarial strategy.

[0144] The updated local adversarial strategies are transmitted and integrated through an information exchange topology. This topology employs a fully connected approach to ensure efficient information sharing among the coordinating units. The phase adjustment optimization strategy of the third coordinating unit and the delay optimization strategy of the fifth coordinating unit are integrated at the central fusion node. After conflict detection confirms that there is no mutual exclusion, these optimized strategies are applied to the global adversarial knowledge base. The strategy entry with ID GKB-20231015-123 in the global adversarial knowledge base is updated. The new strategy includes the optimized parameter set and execution sequence, and the confidence score of the updated strategy is increased from 85 to 91, completing the dynamic update of the local adversarial strategies and the global adversarial knowledge base. The updated knowledge base entry is timestamped and a change log is recorded to support subsequent strategy backtracking and effect analysis.

[0145] This invention relates to a biomimetic swarm cooperative intelligent anti-noise system for combat environments, comprising:

[0146] The first unit is used to collect multidimensional noise signals in the combat environment and perform time-frequency domain decomposition on the multidimensional noise signals to obtain a noise feature matrix;

[0147] The second unit is used to construct a group collaboration model based on a biomimetic mechanism. The group collaboration model contains multiple collaboration units. Each collaboration unit establishes a local adversarial strategy based on the feature components in the noise feature matrix and forms a global adversarial knowledge base through information interaction between collaboration units.

[0148] The third unit is used to spatially map the adversarial strategies in the global adversarial knowledge base to the target areas in the combat environment based on the type of feature components in the noise feature matrix, so as to obtain a set of regional adversarial schemes.

[0149] The fourth unit is used to generate adaptive countermeasure instructions based on the set of regional confrontation schemes and operational mission constraints, through the distributed decision-making mechanism of each cooperative unit in the group cooperation model.

[0150] The fifth unit is used to drive the adversarial device to perform noise suppression operations on the target area based on the adaptive adversarial command, and to collect adversarial effect evaluation information. The adversarial effect evaluation information is used to dynamically update the local adversarial strategy and the global adversarial knowledge base to form a closed-loop optimization.

[0151] A third aspect of the present invention provides an electronic device, comprising:

[0152] processor;

[0153] Memory used to store processor-executable instructions;

[0154] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0155] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0156] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0157] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A biomimetic swarm cooperative intelligent countermeasure method for combat environment noise, characterized in that, include: Collect multidimensional noise signals from the combat environment and perform time-frequency domain decomposition on the multidimensional noise signals to obtain a noise feature matrix; A group collaboration model is constructed based on a biomimetic mechanism. The group collaboration model contains multiple collaboration units. Each collaboration unit establishes a local adversarial strategy based on the feature components in the noise feature matrix, and forms a global adversarial knowledge base through information interaction between collaboration units. Based on the feature component types in the noise feature matrix, the adversarial strategies in the global adversarial knowledge base are spatially mapped to the target areas in the combat environment to obtain a set of regional adversarial schemes; Based on the set of regional confrontation schemes and the constraints of the combat mission, adaptive confrontation instructions are generated through the distributed decision-making mechanism of each collaborative unit in the group collaboration model. Based on the adaptive adversarial command, the adversarial device performs noise suppression operations on the target area and collects adversarial effect evaluation information. The adversarial effect evaluation information is used to dynamically update the local adversarial strategy and the global adversarial knowledge base to form a closed-loop optimization.

2. The method according to claim 1, characterized in that, Each collaborative unit establishes a local adversarial strategy based on the feature components in the noise feature matrix, and forms a global adversarial knowledge base through information exchange between collaborative units, including: Extract feature components of different frequency bands from the noise feature matrix, and divide the multiple cooperative units into multiple cooperative levels according to the energy distribution characteristics of the feature components; Within each collaborative level, each collaborative unit constructs a local feature space based on the assigned feature components, and iteratively generates an initial local adversarial strategy in the local feature space through a biomimetic evolution mechanism. Establish an information interaction topology structure between collaborative units, wherein the information interaction topology structure defines the information transmission path and collaborative weight between each collaborative unit; Each collaborative unit transmits the initial local adversarial strategy to its neighboring collaborative units through the information interaction topology, and receives local adversarial strategies from its neighboring collaborative units. Based on the received local adversarial strategies and the collaborative weights, the initial local adversarial strategy is fused and corrected to obtain the corrected local adversarial strategy. The modified local adversarial strategies are aggregated according to the cooperation level, and the global adversarial knowledge base is formed through the strategy consistency constraint mechanism between levels.

3. The method according to claim 2, characterized in that, Within each collaborative level, each collaborative unit constructs a local feature space based on its assigned feature components, and iteratively generates an initial local adversarial strategy within this local feature space through a biomimetic evolutionary mechanism, including: Each collaborative unit constructs a local feature space based on the assigned feature components. The local feature space includes feature dimensions and feature boundaries. The feature dimensions are determined by the vector dimensions of the feature components, and the feature boundaries are determined by the value range of each component in the feature components. Multiple candidate adversarial strategies are initialized in the local feature space, and each candidate adversarial strategy includes a set of adversarial parameters and a sequence of adversarial actions. A biomimetic fitness evaluation function is constructed, which calculates the fitness value of each candidate adversarial strategy based on the feature components and the combat mission objective. The multiple candidate adversarial strategies are iteratively optimized using a biomimetic evolution mechanism. The iterative optimization process includes: selecting candidate adversarial strategies based on the fitness value, mutating the set of adversarial parameters of the selected candidate adversarial strategies, and recombining the adversarial action sequence of the selected candidate adversarial strategies to generate a new generation of candidate adversarial strategies. Repeat the iterative optimization process until the fitness value of the new generation of candidate adversarial strategies meets the preset convergence condition, and determine the candidate adversarial strategy that meets the preset convergence condition as the initial local adversarial strategy.

4. The method according to claim 1, characterized in that, Based on the feature component types in the noise feature matrix, the adversarial strategies in the global adversarial knowledge base are spatially mapped to target areas in the operational environment to obtain a set of regional adversarial schemes, including: The feature component type identifier is extracted from the noise feature matrix. The feature component type identifier represents the frequency domain attribute and energy distribution attribute of each feature component. Acquire spatial information of a target area in the combat environment, wherein the spatial information of the target area includes the spatial coordinates and spatial extent of the target area; Establish a mapping association rule between feature component types and target regions, wherein the mapping association rule defines the correspondence between feature component type identifiers and target region spatial information; According to the mapping association rules, the feature component type identifiers associated with each adversarial strategy in the global adversarial knowledge base are matched with the target region spatial information to determine the applicable target region for each adversarial strategy. For each target region, an adversarial strategy that is applicable to the target region and matches the target region is selected from the global adversarial knowledge base to form a set of candidate adversarial strategies for the target region; Based on the feature component type identifier of each adversarial strategy in the candidate adversarial strategy set and the spatial range of the target area, calculate the spatial coverage of each adversarial strategy in the target area. Based on the spatial coverage, the adversarial strategies in the candidate adversarial strategy set are combined and optimized to generate a regional adversarial scheme for the target area. The regional adversarial schemes for all target areas are then summarized to form the regional adversarial scheme set.

5. The method according to claim 1, characterized in that, Based on the aforementioned set of regional confrontation schemes and operational mission constraints, adaptive confrontation instructions are generated through the distributed decision-making mechanism of each collaborative unit in the group collaboration model, including: The regional confrontation scheme for each target region in the set of regional confrontation schemes is decomposed into multiple confrontation sub-tasks, and each confrontation sub-task includes confrontation actions and execution sequence. Based on the operational mission constraints, a feasibility assessment is performed on the multiple adversarial sub-missions, and a set of candidate adversarial sub-missions that meet the constraints is selected. The candidate adversarial subtask set is assigned to the corresponding collaborative units in the group collaborative model according to the target region, and each collaborative unit receives the assigned adversarial subtask; Each coordinating unit independently generates local adversarial instructions based on the received adversarial sub-tasks and the constraints of the combat mission through a distributed decision-making mechanism. The local adversarial instructions include the control parameters of the adversarial device and the execution time. A command negotiation mechanism is established between collaborative units. Each collaborative unit sends the local confrontation command to the adjacent collaborative unit through the command negotiation mechanism and receives the local confrontation command from the adjacent collaborative unit. Each coordinating unit performs conflict detection and timing adjustment on its own local adversarial commands based on the local adversarial commands received from neighboring coordinating units, and obtains coordinated local adversarial commands. The coordinated local adversarial instructions of all collaborative units are integrated, and the integrated instruction sequence is prioritized to generate the adaptive adversarial instructions.

6. The method according to claim 1, characterized in that, The adaptive adversarial command drives the adversarial device to perform noise suppression operations on the target area, and collects adversarial effect evaluation information. The adversarial effect evaluation information is then used to dynamically update the local adversarial strategy and the global adversarial knowledge base, including: The adaptive countermeasure instruction is parsed, and the countermeasure device control parameters and execution time are extracted. A timing trigger signal for the countermeasure device is generated based on the execution time. Based on the timing trigger signal and the countermeasure device control parameters, the countermeasure device is driven to perform noise suppression operation on the target area according to the predetermined execution time. During the noise suppression operation, noise signal change data of the target area is collected in real time, and the countermeasure effect evaluation index is calculated based on the noise signal change data. The countermeasure effect evaluation index includes noise suppression rate and countermeasure response delay. The adversarial effect evaluation index is compared with the expected adversarial effect in the adaptive adversarial instruction to generate adversarial effect deviation information; adversarial effect evaluation information is constructed based on the adversarial effect deviation information, which includes adversarial effect deviation information, noise signal change data and execution time correlation records; The adversarial effect deviation information and noise signal change data in the adversarial effect evaluation information are fed back to each cooperative unit in the group cooperative model; each cooperative unit corrects the adversarial parameter set and adversarial action sequence according to the received adversarial effect deviation information to obtain the updated local adversarial strategy. Each collaborative unit transmits and merges the updated local adversarial strategy through an information interaction topology, replaces and updates the adversarial strategy in the global adversarial knowledge base, and completes the dynamic update of the local adversarial strategy and the global adversarial knowledge base.

7. A biomimetic swarm cooperative intelligent combat environment noise countermeasure system, used to implement the method as described in any one of claims 1-6, characterized in that, include: The first unit is used to collect multidimensional noise signals in the combat environment and perform time-frequency domain decomposition on the multidimensional noise signals to obtain a noise feature matrix; The second unit is used to construct a group collaboration model based on a biomimetic mechanism. The group collaboration model contains multiple collaboration units. Each collaboration unit establishes a local adversarial strategy based on the feature components in the noise feature matrix and forms a global adversarial knowledge base through information interaction between collaboration units. The third unit is used to spatially map the adversarial strategies in the global adversarial knowledge base to the target areas in the combat environment based on the type of feature components in the noise feature matrix, so as to obtain a set of regional adversarial schemes. The fourth unit is used to generate adaptive countermeasure instructions based on the set of regional confrontation schemes and operational mission constraints, through the distributed decision-making mechanism of each cooperative unit in the group cooperation model. The fifth unit is used to drive the adversarial device to perform noise suppression operations on the target area based on the adaptive adversarial command, and to collect adversarial effect evaluation information. The adversarial effect evaluation information is used to dynamically update the local adversarial strategy and the global adversarial knowledge base to form a closed-loop optimization.

8. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 6.

Images