A low altitude defense platform deployment system and method

Abstract

This application discloses a low-altitude defense platform deployment system and method. The system includes: a data acquisition module for acquiring relevant data of the defense area and information on available platforms; a capability quantification module for uniformly quantifying the capability parameters of each platform to obtain standardized capability descriptions; an AI agent search module, comprising multiple parallel-working AI agents and a shared memory area; each AI agent iteratively executes the generation of candidate deployment schemes, calculation of constraint violations, and evaluation of comprehensive performance scores, and updates the locally optimal deployment scheme based on the evaluation results, exchanging scheme information through the shared memory area to guide subsequent searches; and a scheme determination module for determining the globally optimal deployment scheme when termination conditions are met. This application improves the comprehensive protection effectiveness and planning efficiency of heterogeneous defense platform deployment schemes through capability quantification and multi-agent parallel collaborative search, achieving automated global optimization deployment under complex constraints.

Description

Technical Field

[0001] This application relates to the field of low-altitude defense technology, and in particular to a low-altitude defense platform deployment system and method. Background Technology

[0002] With the rapid development and widespread adoption of low-altitude aircraft technology in the civilian sector, low-altitude security incidents such as illegal intrusion and unauthorized flights are occurring frequently, making the need for low-altitude security protection for critical locations such as airports, nuclear power plants, and important military and political sites increasingly urgent. In constructing a low-altitude defense system, it is necessary to configure various heterogeneous defense platforms, including detection and interception platforms, based on the protection requirements of the defense scenarios. The deployment scheme of these defense platforms directly determines the overall protection capability and emergency response efficiency of the entire low-altitude defense system.

[0003] Currently, the deployment planning of low-altitude defense platforms in the industry mostly adopts manual experience-based planning, static pre-plan matching for typical scenarios, or traditional heuristic optimization algorithms. In practical engineering applications, the above-mentioned existing solutions have intractable technical defects: On the one hand, different types of heterogeneous defense platforms vary significantly in their working principles, performance parameters, and scope of action. Existing technologies lack a unified quantitative evaluation benchmark for capabilities, resulting in inaccurate assessments of the effectiveness of deployment solutions. Moreover, the generated solutions often have non-compliance issues during actual implementation, requiring repeated manual adjustments, which is inefficient. On the other hand, the search space for feasible solutions to deployment solutions increases exponentially with the scale of the problem. Existing optimization methods are inefficient in solving these solutions, easily getting trapped in local optima, making it difficult to obtain a deployment solution with better global performance within a reasonable time, and failing to meet the rapid response requirements of emergency protection scenarios. Summary of the Invention

[0004] Based on the above problems, this application provides a low-altitude defense platform deployment system and method, aiming to improve the comprehensive protection effectiveness and planning efficiency of heterogeneous defense platform deployment schemes, and realize the automated and globally optimized deployment of multiple types of platforms under complex constraints.

[0005] The embodiments of this application disclose the following technical solutions: The first aspect of this application provides a low-altitude defense platform deployment system, the system comprising: The data acquisition module is used to acquire data related to the defense area and information on available platforms; The capability quantification module is used to perform unified quantification processing on the capability parameters of each platform in the available platform information to obtain a standardized capability description. The AI ​​agent search module includes multiple AI agents working in parallel and a shared memory area; Each of the AI ​​agents is used to iteratively perform the following operations until a preset termination condition is met: Based on its own search strategy, the current local optimal deployment scheme, the relevant data of the defense area, and the standardized capability description, a large language model is used to generate candidate deployment schemes for the available platform information. The constraint violation degree of the candidate deployment scheme is calculated based on the preset constraint conditions, and when the constraint violation degree is less than or equal to the preset violation degree threshold, the comprehensive performance score of the candidate deployment scheme is calculated in combination with the target evaluation index. When the overall performance score of the candidate deployment scheme is better than the overall performance score of the current locally optimal deployment scheme, the candidate deployment scheme is updated to the current locally optimal deployment scheme, and the updated locally optimal deployment scheme is written into the shared memory area. At the same time, the deployment scheme information of other AI agents is obtained from the shared memory area to guide the generation of the next candidate deployment scheme. The scheme determination module is used to determine the globally optimal deployment scheme based on the information in the shared memory area when the preset termination condition is met.

[0006] In an optional implementation, the data related to the defense area includes geographical information of the defense area, information on protected targets, and threat intelligence information; the system also includes a data preprocessing module, used for: Based on the geographic information of the defense area, the defense area is gridded, dividing the defense area into grids with a preset precision, and the center point of each grid is used as a candidate deployment location to obtain a set of candidate deployment locations; Based on the protection target information, key protection areas are determined; The direction of the threat source is determined based on the threat intelligence information.

[0007] In an optional implementation, the step of generating candidate deployment schemes for the available platform information using a large language model based on its own search strategy, the current locally optimal deployment scheme, the relevant data of the defense area, and the standardized capability description includes: The search strategy, the current local optimal deployment scheme, the set of candidate deployment locations, the key protection area, the direction of the threat source, and the standardized capability description are integrated into natural language instructions, which are then input into the large language model to obtain platform location adjustment suggestions for the current local optimal deployment scheme. Based on the platform location adjustment suggestions, deployment locations are selected for each platform from the set of candidate deployment locations to form the candidate deployment scheme.

[0008] In an optional implementation, the standardized capability is described as a standardized capability vector, which includes detection capability value, interception capability value, response speed value, and terrain adaptation value. The detection capability value is obtained by normalizing the detection distance, detection accuracy, and detection field of view. The interception capability value is obtained by normalizing the interception distance, interception success rate, and capacity parameters. The response speed value is obtained by normalizing the reaction time and maneuver speed. The terrain adaptability value is obtained by evaluating the installation conditions and line-of-sight requirements.

[0009] In the optional implementation, the preset constraints include terrain occlusion constraints, platform capability constraints, electromagnetic compatibility constraints, and resource supply constraints. The terrain occlusion constraint requires that the line-of-sight between the platform's deployment location and the defense area be greater than or equal to a preset line-of-sight threshold. The platform capability constraint requires that the distance between the deployment location of the platform used for detection and the boundary of the defense area be less than or equal to the platform's maximum detection range. The electromagnetic compatibility constraints require that the spacing between platforms operating in the same frequency band be greater than or equal to a preset minimum spacing. The resource supply constraints require that the platform be deployed within the coverage area of ​​a pre-defined infrastructure. The calculation of the constraint violation degree of the candidate deployment scheme based on preset constraint conditions includes: The degree of violation of the terrain occlusion constraint, platform capability constraint, electromagnetic compatibility constraint, and resource supply constraint by the candidate deployment scheme is weighted and summed according to a preset constraint weight coefficient to obtain the constraint violation degree.

[0010] In optional implementations, the target evaluation metrics include coverage, average response time, and overall interception probability. The calculation of the overall performance score of the candidate deployment scheme by combining the target evaluation indicators includes: The comprehensive performance score is obtained by weighting and summing the coverage rate, the reciprocal normalized value of the average response time, and the comprehensive interception probability according to the preset target weight coefficient, and then subtracting the penalty term composed of the constraint violation degree and the constraint violation penalty coefficient.

[0011] In an optional implementation, the number of AI agents in the AI ​​agent search module is dynamically determined based on the number of platforms in the available platform information and the size of the candidate deployment location set; The AI ​​agent search module also includes a strategy allocation unit, which is used to allocate the search strategy to each AI agent during the initialization phase; the search strategy is used to generate an initial deployment plan and to continuously guide the generation of candidate deployment plans in subsequent iterations.

[0012] In the optional implementation, the self-search strategy includes at least one of the following: random strategy, greedy strategy, and clustering strategy; The random strategy involves randomly assigning locations from the set of candidate deployment locations; the greedy strategy prioritizes deployment based on the direction of the threat source and / or the key protection areas; and the clustering strategy involves grouping and deploying according to platform functions.

[0013] In an optional implementation, the system further includes a dynamic adjustment module, which, when it detects an update to the data related to the defense area or a change in the available platform information, takes the currently deployed globally optimal deployment scheme as the current locally optimal deployment scheme and calls the AI ​​intelligent agent search module to perform the iterative operation to generate an incremental adjustment scheme.

[0014] The second aspect of this application provides a method for deploying a low-altitude defense platform, applied to the low-altitude defense platform deployment system described in any implementation of the first aspect, the method comprising: The data acquisition module obtains relevant data on the defense area and information on available platforms. The capability parameters of each platform in the available platform information are uniformly quantified by the capability quantification module to obtain a standardized capability description. Through the AI ​​agent search module, perform the following operations iteratively until the preset termination condition is met: Based on its own search strategy, the current local optimal deployment scheme, the relevant data of the defense area, and the standardized capability description, a large language model is used to generate candidate deployment schemes for the available platform information. The constraint violation degree of the candidate deployment scheme is calculated based on the preset constraint conditions, and when the constraint violation degree is less than or equal to the preset violation degree threshold, the comprehensive performance score of the candidate deployment scheme is calculated in combination with the target evaluation index. When the overall performance score of the candidate deployment scheme is better than the overall performance score of the current local optimal deployment scheme, the candidate deployment scheme is updated to the current local optimal deployment scheme, and the updated local optimal deployment scheme is written into the shared memory area in the AI ​​agent search module. At the same time, the deployment scheme information of other AI agents is obtained from the shared memory area to guide the generation of the next candidate deployment scheme. When the preset termination conditions are met, the scheme determination module determines the globally optimal deployment scheme based on the information in the shared memory area.

[0015] Compared with the prior art, this application has the following beneficial effects: The low-altitude defense platform deployment system proposed in this application includes: a data acquisition module for acquiring relevant data of the defense area and information on available platforms; a capability quantification module for uniformly quantifying the capability parameters of each platform in the available platform information to obtain a standardized capability description; and an AI agent search module, including multiple parallel AI agents and a shared memory area. Each AI agent iteratively performs the following operations until a preset termination condition is met: based on its own search strategy, the current locally optimal deployment scheme, relevant data of the defense area, and the standardized capability description, it uses a large language model to generate candidate deployment schemes for the available platform information; and based on preset constraints... The constraint violation degree of candidate deployment schemes is calculated, and when the constraint violation degree is less than or equal to a preset violation degree threshold, the comprehensive performance score of the candidate deployment scheme is calculated in combination with the target evaluation index. When the comprehensive performance score of the candidate deployment scheme is better than the comprehensive performance score of the current local optimal deployment scheme, the candidate deployment scheme is updated to the current local optimal deployment scheme, and the updated local optimal deployment scheme is written into the shared memory area. At the same time, the deployment scheme information of other AI agents is obtained from the shared memory area to guide the generation of the next candidate deployment scheme. The scheme determination module is used to determine the global optimal deployment scheme based on the information in the shared memory area when the preset termination condition is met. As can be seen, the technical solution of this application uses a capability quantification module to uniformly quantify the capability parameters of various heterogeneous platforms, obtaining standardized capability descriptions and establishing a unified quantitative evaluation benchmark. This enables different types of platforms to conduct collaborative performance evaluation on the same dimension. Simultaneously, during the iterative generation of candidate deployment schemes, constraint violation calculations and comprehensive performance score evaluations are performed concurrently. This ensures that the final globally optimal deployment scheme achieves multi-objective comprehensive optimization while meeting compliance requirements, effectively improving the comprehensive protection performance of collaborative deployment schemes for multiple types of heterogeneous platforms. This avoids the problems of inaccurate performance evaluations and repeated manual adjustments required for non-compliant schemes caused by the lack of unified quantitative standards in traditional solutions. Furthermore, this application utilizes multiple AI agents working in parallel within the AI ​​agent search module. Each AI agent searches within the deployment space based on its own search strategy and exchanges its local optimal deployment scheme information through a shared memory area. This achieves collaborative optimization among multiple agents, significantly improving search efficiency, shortening scheme generation time, and reducing the risk of getting trapped in local optima. It can obtain deployment schemes with better global performance within a reasonable time, improving the planning efficiency and global optimization capabilities of large-scale deployment schemes. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A schematic diagram of a low-altitude defense platform deployment system provided in this application embodiment; Figure 2 A flowchart illustrating a low-altitude defense platform deployment method provided in this application embodiment. Detailed Implementation

[0018] As described earlier, with the rapid development and widespread adoption of low-altitude aircraft technology in the civilian sector, low-altitude security incidents such as illegal intrusion and unauthorized flights are occurring frequently, making the need for low-altitude security protection for critical locations such as airports, nuclear power plants, and important military and political sites increasingly urgent. In constructing a low-altitude defense system, it is necessary to configure various heterogeneous defense platforms, including detection and interception platforms, according to the protection requirements of the defense scenarios. The deployment scheme of these defense platforms directly determines the overall protection capability and emergency response efficiency of the entire low-altitude defense system.

[0019] Currently, the deployment planning of low-altitude defense platforms in the industry mostly adopts manual experience-based planning, static pre-plan matching for typical scenarios, or traditional heuristic optimization algorithms. In practical engineering applications, the above-mentioned existing solutions have intractable technical defects: On the one hand, different types of heterogeneous defense platforms vary significantly in their working principles, performance parameters, and scope of action. Existing technologies lack a unified quantitative evaluation benchmark for capabilities, resulting in inaccurate assessments of the effectiveness of deployment solutions. Moreover, the generated solutions often have non-compliance issues during actual implementation, requiring repeated manual adjustments, which is inefficient. On the other hand, the search space for feasible solutions to deployment solutions increases exponentially with the scale of the problem. Existing optimization methods are inefficient in solving these solutions, easily getting trapped in local optima, making it difficult to obtain a deployment solution with better global performance within a reasonable time, and failing to meet the rapid response requirements of emergency protection scenarios.

[0020] To address the aforementioned problems, the inventors have proposed a low-altitude defense platform deployment system and method after research.

[0021] The low-altitude defense platform deployment system includes: a data acquisition module for acquiring relevant data of the defense area and information on available platforms; a capability quantification module for uniformly quantifying the capability parameters of each platform in the available platform information to obtain standardized capability descriptions; and an AI agent search module, comprising multiple parallel-working AI agents and a shared memory area. Each AI agent iteratively performs the following operations until a preset termination condition is met: based on its own search strategy, the current locally optimal deployment scheme, relevant data of the defense area, and standardized capability descriptions, it uses a large language model to generate candidate deployment schemes for available platform information; and it calculates candidate deployment schemes based on preset constraints. The system selects the constraint violation degree of the deployment scheme and calculates the comprehensive performance score of the candidate deployment scheme in combination with the target evaluation index when the constraint violation degree is less than or equal to the preset violation degree threshold. When the comprehensive performance score of the candidate deployment scheme is better than the comprehensive performance score of the current local optimal deployment scheme, the candidate deployment scheme is updated to the current local optimal deployment scheme, and the updated local optimal deployment scheme is written into the shared memory area. At the same time, the deployment scheme information of other AI agents is obtained from the shared memory area to guide the generation of the next candidate deployment scheme. The scheme determination module is used to determine the global optimal deployment scheme based on the information in the shared memory area when the preset termination condition is met.

[0022] This application's technical solution uses a capability quantification module to uniformly quantify the capability parameters of various heterogeneous platforms, obtaining standardized capability descriptions and establishing a unified quantitative evaluation benchmark. This enables different types of platforms to conduct collaborative performance evaluations on the same dimension. Simultaneously, during the iterative generation of candidate deployment schemes, constraint violation calculations and comprehensive performance score evaluations are performed concurrently. This ensures that the final globally optimal deployment scheme achieves multi-objective comprehensive optimization while meeting compliance requirements, effectively improving the comprehensive protection performance of collaborative deployment schemes for multiple types of heterogeneous platforms. This avoids the problems of inaccurate performance evaluations and repeated manual adjustments required for non-compliant schemes caused by the lack of unified quantitative standards in traditional solutions. Furthermore, this application utilizes multiple AI agents working in parallel within the AI ​​agent search module. Each AI agent searches within the deployment space based on its own search strategy and exchanges information on its locally optimal deployment schemes through a shared memory area. This achieves collaborative optimization among multiple agents, significantly improving search efficiency, shortening scheme generation time, and reducing the risk of getting trapped in local optima. It can obtain globally superior deployment schemes within a reasonable timeframe, improving the planning efficiency and global optimization capabilities of large-scale deployment schemes.

[0023] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0024] See Figure 1 This figure is a schematic diagram of a low-altitude defense platform deployment system structure provided in an embodiment of this application. Figure 1 As shown, the low-altitude defense platform deployment system provided in this application embodiment includes: a data acquisition module 101, a capability quantification module 102, an AI agent search module 103, and a scheme determination module 104. The AI ​​agent search module 103 includes multiple AI agents 1031 working in parallel and a shared memory area 1032.

[0025] The data acquisition module 101 is used to acquire data related to the defense area and information on available platforms.

[0026] In this application embodiment, the defense area related data refers to various basic data related to the defense scenario required for deploying a low-altitude defense system, such as the geographical range of the defense area, terrain features, location and level of key targets to be protected, and the direction and intensity of threat sources.

[0027] Available platform information refers to the list of various types of defense platforms that are currently available for deployment and their technical parameters, such as platform type, quantity, detection range, interception range, response time, etc.

[0028] In one example implementation, when constructing an airport low-altitude defense system, the data acquisition module 101 receives boundary coordinates and elevation data of the airport and its surrounding 15-kilometer radius as relevant data for the defense area, and acquires a list of available platforms, including 5 S-band radars, 8 sets of optoelectronic equipment, 12 interceptor drones and 3 sets of laser weapons, as available platform information.

[0029] In this embodiment, the data acquisition module 101 can receive geographic information data in GeoJSON or Shapefile format through a file upload interface, and receive structured data of platform lists and threat intelligence through a form or API interface.

[0030] This application embodiment uses the data acquisition module 101 to uniformly aggregate basic data and platform information of defense scenarios from different sources and in different formats into the system, providing a data foundation for subsequent capability quantification and intelligent search.

[0031] The capability quantification module 102 is used to uniformly quantify the capability parameters of each platform in the available platform information to obtain a standardized capability description.

[0032] In the embodiments of this application, capability parameters refer to the original performance indicators of each platform in its functional dimensions, such as the detection range and detection accuracy of radar, the interception range and interception success rate of intercepting drones, and the reaction time of laser weapons.

[0033] Standardized capability description refers to a quantitative description obtained by normalizing the original parameters with different physical meanings and dimensions, which can be compared and calculated on a unified scale.

[0034] In this embodiment of the application, the capability quantification module 102 traverses the list of available platforms and, for each platform, calculates a set of standardized values ​​with unified dimensions and consistent value ranges by integrating the original parameters of multiple dimensions such as detection, interception, and response, as a standardized capability description of the platform.

[0035] In one example implementation, a radar with a detection range of 8 kilometers and an optoelectronic device with a detection range of 3 kilometers are processed by the capability quantization module 102, and their respective detection capabilities are converted into standardized values ​​in the range of 0 to 1, so that the capabilities of the two different detection platforms can be directly compared.

[0036] This application embodiment solves the problem of heterogeneous platforms being unable to be uniformly measured due to differences in working principles and performance indicators through the capability quantification module 102, and establishes a unified quantitative evaluation benchmark for subsequent collaborative deployment optimization of AI agents across different platform types.

[0037] The AI ​​agent search module 103 includes multiple AI agents 1031 that work in parallel and a shared memory area 1032.

[0038] In the embodiments of this application, an AI agent refers to an intelligent entity with autonomous search capabilities. Each AI agent can independently generate, modify, and evaluate deployment plans.

[0039] The shared memory area is a storage space that can be read and written by all AI agents, used to exchange and share information on excellent deployment schemes discovered during the search process among multiple AI agents.

[0040] In this embodiment, multiple AI agents operate simultaneously, each exploring the space of possible deployment solutions from different starting points or using different search methods. Excellent solutions discovered by each AI agent during the search process are stored in a shared memory area. Simultaneously, each AI agent also reads solution information discovered by other agents from the shared memory area to improve its own search.

[0041] In one example implementation, the system creates 16 AI agents that work simultaneously. A shared memory area stores several currently discovered optimal deployment schemes. In each iteration, each agent contributes its own discoveries to the shared memory area and also obtains the results of other agents.

[0042] Each AI agent is used to iteratively perform the following operations until a preset termination condition is met: S101. Based on its own search strategy, the current local optimal deployment scheme, relevant data of the defense area and standardized capability description, it uses a large language model to generate candidate deployment schemes for available platform information.

[0043] In the embodiments of this application, the self-search strategy refers to the search behavior rules followed by each AI agent that are different from those of other agents. It determines how the agent tends to explore the solution space.

[0044] The current locally optimal deployment scheme refers to the best deployment scheme identified by the AI ​​agent in the iterations to date after performance evaluation.

[0045] Candidate deployment schemes refer to new deployment schemes generated by the large language model in this iteration that are to be evaluated.

[0046] In this embodiment, the AI ​​agent provides its own search strategy, the specific content of the current local optimal deployment scheme, the characteristics of the defense area-related data, and the standardized capability descriptions of each platform to the large language model. The large language model then uses this contextual information to reason and generate a new deployment scheme as a candidate.

[0047] In one example implementation, an AI agent inputs the state information "there is a blind spot in the coverage of the current plan on the north side of the defense area" and a standardized capability description into a large language model. After inference, the model outputs a new deployment plan that adjusts a certain radar to the north as a candidate deployment plan.

[0048] This application's embodiments, by introducing the reasoning capabilities of a large language model, enable each AI agent to understand the complex requirements of the defense scenario and generate targeted and reasonable candidate solutions, rather than blindly making random attempts.

[0049] S102. Calculate the constraint violation degree of the candidate deployment scheme based on the preset constraint conditions, and calculate the comprehensive performance score of the candidate deployment scheme in combination with the target evaluation index when the constraint violation degree is less than or equal to the preset violation degree threshold.

[0050] In the embodiments of this application, the preset constraints refer to various engineering restrictions that the deployment scheme must meet when it is actually implemented.

[0051] Constraint violation is a quantitative measure of the degree to which a candidate deployment scheme violates the above constraints.

[0052] Target evaluation metrics refer to various performance indicators used to measure the effectiveness of a deployment solution in providing protection.

[0053] The overall performance score is a single score used to evaluate the overall quality of a plan after comprehensively calculating the evaluation indicators of various objectives.

[0054] In this embodiment, after the AI ​​agent generates a candidate deployment scheme, it first checks whether the scheme meets various preset constraints and calculates a constraint violation value. If the violation degree is within an acceptable range (i.e., does not exceed the preset violation threshold), the various performance indicators of the scheme are calculated, and a performance score is obtained. If the violation degree is too high, the scheme is determined to be infeasible and discarded.

[0055] In one example implementation, if the deployment location of a radar in a candidate solution cannot effectively cover the target area due to terrain obstruction, and the calculated constraint violation exceeds a preset threshold, the solution is directly filtered out and does not enter the subsequent performance evaluation stage.

[0056] This application embodiment ensures that only solutions that simultaneously meet compliance requirements and possess high efficiency can be retained as candidates by simultaneously performing constraint checks and performance evaluations after the solution is generated, thus avoiding invalid solutions occupying subsequent computing resources.

[0057] S103. When the overall performance score of the candidate deployment scheme is better than the overall performance score of the current local optimal deployment scheme, the candidate deployment scheme is updated to the current local optimal deployment scheme, and the updated local optimal deployment scheme is written into the shared memory area. At the same time, the deployment scheme information of other AI agents is obtained from the shared memory area to guide the generation of the next candidate deployment scheme.

[0058] In this embodiment, when the overall performance score of a newly generated candidate deployment scheme exceeds the score of the local optimum previously held by the AI ​​agent, the agent replaces its own local optimum with the new scheme. Subsequently, it writes the new scheme into a shared memory area for other agents to reference, and simultaneously reads optimal or excellent scheme fragments discovered by other agents from the shared memory area, absorbing the search experience of other agents.

[0059] In one example implementation, AI agent #1 generates a new solution that is better than its local optimum in the 50th iteration and immediately writes it into the shared memory. Simultaneously, it reads an excellent configuration of radar deployment locations shared by AI agent #8 from the shared memory and incorporates this configuration into its next search round.

[0060] This application embodiment enables the rapid dissemination and combination of excellent solution elements within the group through information sharing and collaboration among intelligent agents. This not only accelerates convergence but also avoids the problem of each intelligent agent getting trapped in local optima when searching independently.

[0061] The scheme determination module 104 is used to determine the globally optimal deployment scheme based on the information in the shared memory area when the preset termination conditions are met.

[0062] In this embodiment of the application, when the iterative search of all AI agents meets the preset termination condition, the scheme determination module 104 reads the locally optimal schemes contributed by all AI agents from the shared memory area, and selects the one with the highest score as the final globally optimal deployment scheme for output.

[0063] In one example implementation, after 500 iterations, the convergence condition is met. The scheme determination module 104 obtains the optimal schemes of 16 AI agents from the shared memory area, compares the comprehensive performance scores of each scheme, and selects the scheme with the highest score as the globally optimal deployment scheme.

[0064] In this embodiment, the scheme determination module 104 summarizes the results of multi-path parallel search and selects the globally optimal solution based on a unified and quantitative performance score standard, thus ensuring the comprehensive performance and decision quality of the final output deployment scheme.

[0065] This application's embodiments utilize a capability quantification module to uniformly quantify the capability parameters of various heterogeneous platforms, obtaining standardized capability descriptions and establishing a unified quantitative evaluation benchmark. This enables different types of platforms to conduct collaborative performance evaluations on the same dimension. Simultaneously, during the iterative generation of candidate deployment schemes, constraint violation calculations and comprehensive performance score evaluations are performed concurrently. This ensures that the final globally optimal deployment scheme achieves multi-objective comprehensive optimization while meeting compliance requirements, effectively improving the overall protection performance of collaborative deployment schemes for multiple types of heterogeneous platforms. This avoids the problems of inaccurate performance evaluations and repeated manual adjustments required for non-compliant schemes caused by the lack of unified quantitative standards in traditional schemes. Furthermore, this application utilizes multiple AI agents working in parallel within the AI ​​agent search module. Each AI agent searches within the deployment space based on its own search strategy and exchanges its locally optimal deployment scheme information through a shared memory area. This achieves collaborative optimization among multiple agents, significantly improving search efficiency, shortening scheme generation time, and reducing the risk of getting trapped in local optima. It can obtain globally superior deployment schemes within a reasonable timeframe, improving the planning efficiency and global optimization capabilities of large-scale deployment schemes.

[0066] In one alternative implementation, the defense area-related data includes geographical information of the defense area, information on protected targets, and threat intelligence information.

[0067] In this application embodiment, the geographic information of the defense area refers to data describing the spatial extent and terrain features of the defense area, such as the boundary coordinates and digital elevation model data of the defense area. The information on protected targets refers to data such as the type, location coordinates, and importance level of key areas requiring priority protection. The threat intelligence information refers to characteristic data about potential threats, such as the direction of threat origin, threat intensity level, and threat type (e.g., fixed-wing UAVs, multi-rotor UAVs, etc.).

[0068] To preprocess the raw data for subsequent search operations, the low-altitude defense platform deployment system also includes a data preprocessing module, used for: The defense area is gridded based on the geographic information of the defense area, dividing the defense area into grids with a preset precision, and the center point of each grid is used as a candidate deployment location to obtain a set of candidate deployment locations.

[0069] In this embodiment, meshing refers to discretizing the continuous defense area space into a finite number of mesh cells, so as to transform the infinite continuous deployment location selection problem into a finite discrete combinatorial optimization problem. The candidate deployment location set refers to the finite set of coordinates consisting of all mesh center points, which are available for platform deployment location selection.

[0070] In one example implementation, the defense area is approximately 700 square kilometers, divided into approximately 2800 grids using a 500m x 500m grid. The coordinates of the center point of each grid form a set of candidate deployment locations. During subsequent AI agent search, the deployment locations for each platform are selected from this set.

[0071] Key protection areas are identified based on information about the protection targets.

[0072] In this embodiment of the application, the key protection area refers to the sub-area within the defense area that needs to be prioritized for protection effectiveness, based on the key location type and importance level in the protection target information.

[0073] In one example implementation, for an airport scenario, key locations such as takeoff and landing routes, terminals, and control towers are marked as key protection areas based on the protection target information. Deployment schemes covering these areas will receive higher evaluations in subsequent scheme assessments.

[0074] Determine the source of the threat based on threat intelligence information.

[0075] In this embodiment of the application, the threat source direction refers to the sector in the direction from which a potential intrusion threat is most likely to enter the defense area, as determined by threat intelligence analysis.

[0076] In one example implementation, if threat intelligence indicates that the threat is mainly entering from the east, north, and west, these three directions are identified as the sources of the threat, and the AI ​​agent is guided to deploy defense resources in these directions during subsequent searches.

[0077] In one alternative implementation, based on its own search strategy, the current locally optimal deployment scheme, relevant data on the defense zone, and standardized capability descriptions, a large language model is used to generate candidate deployment schemes for available platform information, including: The system integrates its own search strategy, the current local optimal deployment scheme, the set of candidate deployment locations, key protection areas, threat source directions, and standardized capability descriptions into natural language commands, inputs them into a large language model, and obtains platform location adjustment suggestions based on the current local optimal deployment scheme.

[0078] In this application embodiment, natural language instructions refer to converting various structured information (including strategy type, current deployment scheme content, candidate location range, key protection area location, threat direction, platform capability description, etc.) in the current search state into text descriptions that can be understood and reasoned by a large language model.

[0079] Platform location adjustment suggestions refer to textual suggestions from large language models, based on input natural language instructions and after inference, regarding how the deployment location of one or more platforms should be adjusted.

[0080] In one example implementation, the AI ​​agent integrates status information such as "Currently, a greedy strategy is adopted, the key protection area is the airport runway, the main threat comes from the north, and the No. 1 radar is deployed at a certain coordinate and the coverage to the north is insufficient" into a natural language command, inputs it into a large language model, and the model returns an adjustment suggestion: "It is recommended to move the No. 1 radar to the north to expand the coverage to the north."

[0081] Based on the platform location adjustment suggestions, deployment locations are selected for each platform from the set of candidate deployment locations to form candidate deployment plans.

[0082] In this embodiment, the AI ​​agent parses the adjustment suggestion text returned by the large language model and maps the position adjustment intention contained therein to the selection or replacement operation of specific coordinate points in the candidate deployment position set, thus forming a new complete deployment scheme.

[0083] In one example implementation, the large language model returns "move Radar 1 approximately 500 meters north". Based on this, the AI ​​agent searches for a grid point located approximately 500 meters north of the current location of Radar 1 in the candidate deployment location set, selects this point as the new deployment location of Radar 1, and combines it with the original locations of other platforms to form a new candidate deployment scheme.

[0084] In one alternative implementation, the normalized capability is described as a normalized capability vector, which includes detection capability values, interception capability values, response speed values, and terrain adaptation values.

[0085] In this embodiment, the detection capability value refers to a normalized measure of the platform's target detection capability, the interception capability value refers to a normalized measure of the platform's threat interception capability, the response speed value refers to a normalized measure of the platform's reaction speed from detection to response, and the terrain adaptability value refers to a comprehensive score of the platform's deployment adaptability under different terrain conditions. These four dimensions together constitute a standardized vector describing the platform's overall capabilities.

[0086] The detection capability value is obtained by normalizing the detection distance, detection accuracy, and detection field of view.

[0087] In this embodiment, the detection distance refers to the maximum distance at which the platform can effectively detect a target; the detection accuracy refers to the error range of the platform's target position measurement; and the detection field of view refers to the angular range covered by the platform's detection. By comprehensively considering these three factors and performing normalization, a detection capability value in the range of 0 to 1 is obtained, denoted as D_i. Its calculation formula is as follows: D_i=(Rd_i / Rd_max)×(Pa_max / Pa_i)×(θ_i / 180°); Where Rd_i is the detection range of platform i, and Rd_max is the maximum detection range among all platforms; Pa_i is the detection accuracy (error value) of platform i, and Pa_max is the maximum detection error (i.e., the lowest accuracy) among all platforms. Using Pa_max / Pa_i makes the platform with higher accuracy (smaller error) score higher; θ_i is the detection field of view of platform i, in degrees.

[0088] In one example implementation, a radar has a detection range of 8 kilometers (Rd_i=8km, Rd_max=8km), a detection accuracy of ±50 meters (Pa_i=50m), a field of view of 360°, and a maximum detection error of 50 meters (Pa_max=50m) among all platforms. The calculated value is Di = (8 / 8) × (50 / 50) × (360 / 180) = 1 × 1 × 2 = 2. After normalization to the 0-1 range, the detection capability value is obtained. An optoelectronic device has a detection range of 3 kilometers, a detection accuracy of ±10 meters (higher accuracy than the radar), and a field of view of 120°. The calculated value is Di = (3 / 8) × (50 / 10) × (120 / 180) = 0.375 × 5 × 0.667 ≈ 1.25. After normalization, its detection capability value is obtained. Using the above formula, the optoelectronic device with higher detection accuracy achieves a higher score in the accuracy dimension.

[0089] The interception capability value is obtained by normalizing the interception distance, interception success rate, and capacity parameters.

[0090] In this embodiment, interception distance refers to the maximum distance at which the platform can effectively intercept threats; interception success rate refers to the probability that the platform successfully destroys or interferes with a target within the effective interception distance; and capacity parameter refers to a measure of the platform's sustained combat capability, such as the number of munitions or energy reserves. The interception capability value is denoted as I_i, and by comprehensively considering these three factors and performing normalization, a value within the range of 0 to 1 is obtained. The calculation formula is as follows: I_i=(Rint_i / Rint_max)×Ps_i×(C_i / C_max); Where Rint_i is the interception distance of platform i, and Rint_max is the maximum interception distance among all platforms; Ps_i is the interception success rate of platform i; C_i is the capacity parameter of platform i, and C_max is the maximum capacity parameter among all platforms. For detection platforms that do not have interception capabilities, their interception capability value I_i = 0.

[0091] In one example implementation, the interception distance of an interceptor drone is 5 kilometers (Rint_max=5km), the interception success rate is 80% (Ps_i=0.8), and the ammunition capacity is 6 rounds (C_max=6 rounds). The calculated value is I_i=(5 / 5)×0.8×(6 / 6)=1×0.8×1=0.8.

[0092] The response speed value is obtained by normalizing the reaction time and maneuver speed.

[0093] In this embodiment, reaction time refers to the time required for the platform to complete preparation for handling after receiving the target instruction, and maneuverability refers to the speed at which the mobile platform reaches the designated deployment location. The response speed value is denoted as S_i, and by comprehensively considering these two factors and performing normalization, a value within the range of 0 to 1 is obtained. Its calculation formula is as follows: S_i=1 / (1+Tr_i / Tr_ref)×(Vm_i / Vm_max); Where Tr_i is the reaction time of platform i, Tr_ref is the reference reaction time (e.g., 60 seconds); Vm_i is the maneuver speed of platform i, and Vm_max is the maximum maneuver speed among all platforms. The platform with the shorter reaction time and the greater maneuver speed has a higher response speed value.

[0094] In one example implementation, the reaction time of a laser weapon is 1 second (Tr_i=1s, Tr_ref=60s), and the maneuver speed is a fixed value. The calculated value is S_i=1 / (1+1 / 60)×1≈0.98.

[0095] The terrain adaptability value is obtained by evaluating the installation conditions and line-of-sight requirements.

[0096] In this embodiment, the installation condition requirements refer to the platform's requirements for the physical conditions of the deployment site, such as area, flatness, load-bearing capacity, and power supply. The line-of-sight requirement refers to the platform's need for good visibility during operation. The terrain adaptability value is denoted as T_i, which is determined by evaluating both requirements using an expert scoring method. The value ranges from 0 to 1, with a value closer to 1 indicating stronger adaptability to terrain conditions.

[0097] In one example implementation, the interceptor drone has low site requirements, can take off and land vertically, and has a terrain adaptability score of 0.8 (T_i). The large radar requires flat, solid ground and a stable power supply, and has a terrain adaptability score of 0.7 (T_i).

[0098] In one alternative implementation, the preset constraints include terrain occlusion constraints, platform capability constraints, electromagnetic compatibility constraints, and resource supply constraints.

[0099] Terrain occlusion constraints require that the line-of-sight between the platform's deployment location and the defense area be greater than or equal to a preset line-of-sight threshold.

[0100] In this embodiment, terrain occlusion constraints are used to ensure that a platform deployed at a certain location can effectively detect or intercept targets within its area of ​​responsibility without being obstructed by terrain features such as mountains and buildings. Line-of-sight rate refers to the proportion of unobstructed lines of sight in the platform's detection direction.

[0101] In one example implementation, the system performs line-of-sight analysis based on a digital elevation model, calculating the visibility between the platform deployment location and each sampling point within the defense area, requiring the visibility rate to reach or exceed a preset threshold. If a radar deployment location is obstructed by hills within the defense area, and the calculated visibility rate is below the threshold, the deployment location is determined to violate terrain obstruction constraints.

[0102] Platform capability constraints require that the distance between the deployment location of the platform used for detection and the boundary of the defense zone be less than or equal to the platform's maximum detection range.

[0103] In this embodiment of the application, this constraint ensures that the deployed detection platform can cover the edge of the defense area within its effective working range, and will not have a detection coverage blind spot due to excessive deployment distance.

[0104] In one example implementation, a radar with a maximum detection range of 5 kilometers is deployed at a location 6 kilometers away from the nearest boundary of the defense zone, which violates the platform capability constraints.

[0105] Electromagnetic compatibility constraints require that the spacing between platforms operating in the same frequency band be greater than or equal to a preset minimum spacing.

[0106] In this embodiment of the application, this constraint is used to prevent electronic devices operating in the same or similar frequency bands from generating mutual electromagnetic interference due to being deployed too close together, thus affecting their normal operation.

[0107] In one example implementation, if the distance between the deployment locations of two radars operating in the same frequency band is less than a preset minimum distance, the deployment scheme violates electromagnetic compatibility constraints.

[0108] Resource supply constraints require the platform to be deployed within the pre-defined infrastructure coverage area.

[0109] In this embodiment of the application, this constraint ensures that the platform deployment location has the basic support conditions required for normal operation, such as power supply, communication, and road access.

[0110] In one example implementation, if a platform is to be deployed in an area without power line coverage and without road access, then the deployment location violates resource supply constraints.

[0111] The constraint violation degree of candidate deployment schemes is calculated based on preset constraints, including: The degree of violation of the candidate deployment schemes against terrain occlusion constraints, platform capability constraints, electromagnetic compatibility constraints, and resource supply constraints is weighted and summed according to preset constraint weight coefficients to obtain the constraint violation degree.

[0112] In this embodiment of the application, the degree of constraint violation is denoted as V(X), and its calculation formula is as follows: V(X) = Σw_j × v_j(X); Where v_j(X) represents the degree of violation of the j-th type of constraint by the candidate deployment scheme X (0 indicates complete satisfaction, and a positive value indicates a violation), and w_j is the preset constraint weight coefficient for this type of constraint. The degree of violation of each constraint is first quantified into a numerical value, and then the violation degree values ​​are weighted and summed according to their respective weight coefficients to obtain a comprehensive constraint violation value.

[0113] In one example implementation, the degree of violation of terrain occlusion constraint v_1=0.1, the degree of violation of electromagnetic compatibility constraint v_3=0.2, and the other constraints are satisfied. The weights of each constraint are w_1=0.3 and w_3=0.4, respectively. Then the degree of constraint violation V(X)=0.3×0.1+0.4×0.2=0.03+0.08=0.11.

[0114] In one alternative implementation, the target evaluation metrics include coverage, average response time, and overall intercept probability.

[0115] In this embodiment, coverage rate refers to the proportion of the area within the defense zone effectively covered by at least one detection platform to the total area of ​​the defense zone. Average response time refers to the average time required from initial detection to completion of interception when a simulated threat enters the defense zone from different directions. Overall interception probability refers to the combined success probability of interception against a single threat target, achieved through the coordinated action of all interception platforms capable of covering that target.

[0116] In the embodiments of this application, coverage reflects the spatial coverage completeness of the defense scheme, average response time reflects the temporal response efficiency of the defense scheme, and comprehensive interception probability reflects the final interception effectiveness of the defense scheme. The three together evaluate the comprehensive protection effectiveness of a deployment scheme from different dimensions.

[0117] In one example implementation, coverage is denoted as C(X), average response time as T_resp(X), and overall interception probability as P_kill(X). The specific calculation methods for these three are as follows: The defense area is divided into evaluation grids. For each grid, if there is at least one detection platform whose distance to the grid is less than its detection range and there is no terrain obstruction, then the grid is considered to be effectively covered. The proportion of covered grids to the total number of grids is the coverage rate C(X).

[0118] Multiple threat entry points are uniformly sampled at the boundary of the defense zone. For each entry point, the sum of the detection time of the nearest detection platform and the response time of the nearest interception platform is calculated. The average value of all sampled points is taken as the average response time T_resp(X).

[0119] For a threat, if there are n interception platforms that can cover it, and the interception success rate of each platform is P_i, then the overall interception probability P_kill(X) = 1 - ∏(1 - P_i).

[0120] The overall performance score of candidate deployment schemes is calculated by combining the target evaluation indicators, including: The comprehensive performance score is obtained by summing the coverage rate, the inverse normalized value of the average response time, and the comprehensive interception probability according to the preset target weight coefficient, and then subtracting the penalty term consisting of the constraint violation degree and the constraint violation penalty coefficient.

[0121] In this embodiment of the application, the overall performance score is denoted as Score(X), and its calculation formula is as follows: Score(X)=α×C(X)+β×(T_ref / T_resp(X))+γ×P_kill(X)-λ×V(X); Where α, β, and γ are the preset target weight coefficients for each objective, satisfying α + β + γ = 1; T_ref is the reference response time, used to normalize the average response time; λ is the constraint violation penalty coefficient; and V(X) is the constraint violation degree of candidate deployment scheme X. In the calculation, the average response time is first converted to its reciprocal form to align with other positive indicators and then normalized. Then, the three performance indicators are weighted and summed according to their weight coefficients. Finally, a penalty term consisting of the constraint violation degree multiplied by the penalty coefficient is subtracted from this weighted sum to obtain the final comprehensive performance score. The introduction of the penalty term ensures that even if a scheme performs well on the performance indicators, it will still be penalized if there is a constraint violation, thus guiding the search towards convergence in the feasible region.

[0122] In one example implementation, a balanced weighting pattern is used (α=0.4, β=0.3, γ=0.3), T_ref=60 seconds, λ=0.2. A certain solution has a coverage rate C(X)=0.92, an average response time T_resp(X)=45 seconds, a comprehensive interception probability P_kill(X)=0.88, and a constraint violation rate V(X)=0.05. Therefore, the comprehensive performance score Score(X)=0.4×0.92+0.3×(60 / 45)+0.3×0.88-0.2×0.05=1.022.

[0123] In one alternative implementation, to match the parallel search capability of the AI ​​agent with the problem size, the number of AI agents in the AI ​​agent search module is dynamically determined based on the number of platforms in the available platform information and the size of the candidate deployment location set.

[0124] In this embodiment, when there are many available platforms and a large set of candidate deployment locations, the solution space grows exponentially, requiring more AI agents to cover a wider search area; when the problem size is small, fewer AI agents can be configured to efficiently complete the search, avoiding unnecessary waste of computing resources.

[0125] In one example implementation, for small-scale deployments with fewer than 10 platforms, 8 to 16 AI agents are configured; for medium-scale deployments with 10 to 20 platforms, 16 to 32 AI agents are configured; and for large-scale deployments with more than 20 platforms, 32 to 64 AI agents are configured.

[0126] The AI ​​agent search module also includes a strategy allocation unit, which assigns a search strategy to each AI agent during the initialization phase. The agent's own search strategy is used to generate an initial deployment plan and continues to guide the generation of candidate deployment plans in subsequent iterations.

[0127] In the embodiments of this application, the strategy allocation unit specifies the type of search strategy to be followed by each AI agent before the search begins. Different AI agents adopt different search strategies, which helps to explore the solution space from multiple directions and in multiple ways, increasing the diversity and coverage of the search.

[0128] In one example implementation, the strategy allocation unit divides the 16 AI agents into three groups and assigns different search strategies to each group, enabling each group to start the search from different initial schemes and search paths.

[0129] In one alternative implementation, in order to enable the initial solution to cover multiple representative regions in the solution space, the self-search strategy includes at least one of the following: a random strategy, a greedy strategy, and a clustering strategy.

[0130] Among them, the random strategy randomly assigns locations from the set of candidate deployment locations, the greedy strategy prioritizes deployment based on the direction of threat sources and / or key protection areas, and the clustering strategy groups deployments according to platform functions.

[0131] In this embodiment, the random strategy generates an initial plan by randomly selecting a location for each platform from the set of candidate deployment locations. The advantage is that the initial plan is widely distributed and does not depend on prior knowledge.

[0132] The greedy strategy prioritizes deploying the most capable platforms near the source of the threat and key protection areas, and then gradually deploys the remaining platforms. Its advantage lies in the strong targeted protection of the initial solution.

[0133] The clustering strategy groups the platforms according to their detection and interception functions. Each group is deployed relatively centrally in space to achieve intra-group coordination. Its advantage lies in taking into account the tactical coordination between platforms with different functions.

[0134] In one example implementation, an AI agent employing a greedy strategy, upon initialization, prioritizes deploying the radar with the highest detection capability in a candidate location on the high-threat north side, close to the runway—a key protection area—before deploying the remaining platforms sequentially. An AI agent employing a clustering strategy, upon initialization, concentrates detection platforms in core observation positions within the defense area and interception platforms in the outer defense perimeter.

[0135] In one alternative implementation, in order to respond to changes in the situation after actual deployment, the low-altitude defense platform deployment system also includes a dynamic adjustment module. When the system detects that the relevant data of the defense area has been updated or the available platform information has changed, it takes the currently deployed global optimal deployment scheme as the current local optimal deployment scheme and calls the AI ​​intelligent agent search module to perform iterative operations to generate an incremental adjustment scheme.

[0136] In this embodiment, when the system detects changes such as threat intelligence updates (e.g., changes in the direction of threat sources), adjustments to the defense area, the addition of new deployable platforms, or failures of existing platforms, the dynamic adjustment module automatically initiates an incremental optimization process. This process uses the currently deployed solution as the starting point for the search, calls the AI ​​intelligent agent search module to perform local optimization within a limited search range, and quickly generates a new solution with minimal adjustments, rather than replanning from scratch.

[0137] In one example implementation, after the airport defense plan has been deployed and running for a period of time, threat intelligence indicates that the main threat direction has shifted from the east to the north. Upon detecting this update, the dynamic adjustment module uses the currently deployed plan as the initial solution and invokes the AI ​​agent search module to search within the constraints of allowing partial platform movement. An adjustment plan is generated in a short time, shifting some eastern defense resources to the north. After the adjustment, the probability of interception on the north side is significantly improved, and the plan generation time is greatly shortened compared to a complete replanning.

[0138] In an alternative implementation, to facilitate commanders' intuitive understanding and use of the deployment plan, the low-altitude defense platform deployment system also includes a plan output module, which outputs the globally optimal deployment plan in the form of a visual deployment diagram or a structured deployment list.

[0139] In this embodiment, the visual deployment map refers to a situational map on a terrain base map that graphically marks the deployment location, detection coverage, threat interception path, and other information of each platform. The structured deployment list refers to a list document that lists information such as platform number, type, deployment coordinates, orientation angle, and configuration parameters in a regular format such as a table.

[0140] In one example implementation, the visualization deployment map generated by the solution output module indicates the deployment location and coverage of all platforms with icons of different colors and semi-transparent areas, and draws the interception paths from each threat direction; at the same time, the structured deployment list output records the detailed information of each platform line by line.

[0141] This application's embodiments utilize a capability quantification module to uniformly quantify the capability parameters of various heterogeneous platforms, obtaining standardized capability descriptions and establishing a unified quantitative evaluation benchmark. This enables different types of platforms to conduct collaborative performance evaluations on the same dimension. Simultaneously, during the iterative generation of candidate deployment schemes, constraint violation calculations and comprehensive performance score evaluations are performed concurrently. This ensures that the final globally optimal deployment scheme achieves multi-objective comprehensive optimization while meeting compliance requirements, effectively improving the overall protection performance of collaborative deployment schemes for multiple types of heterogeneous platforms. This avoids the problems of inaccurate performance evaluations and repeated manual adjustments required for non-compliant schemes caused by the lack of unified quantitative standards in traditional schemes. Furthermore, this application utilizes multiple AI agents working in parallel within the AI ​​agent search module. Each AI agent searches within the deployment space based on its own search strategy and exchanges its locally optimal deployment scheme information through a shared memory area. This achieves collaborative optimization among multiple agents, significantly improving search efficiency, shortening scheme generation time, and reducing the risk of getting trapped in local optima. It can obtain globally superior deployment schemes within a reasonable timeframe, improving the planning efficiency and global optimization capabilities of large-scale deployment schemes.

[0142] Based on the low-altitude defense platform deployment system provided in the foregoing embodiments, this application also provides a low-altitude defense platform deployment method, which is applied to any of the low-altitude defense platform deployment systems in the above embodiments. Figure 2 This is a flowchart illustrating a low-altitude defense platform deployment method provided in an embodiment of this application. Figure 2 As shown, the method includes the following steps: S201. Obtain relevant data on the defense area and available platform information through the data acquisition module.

[0143] S202. The capability parameters of each platform in the available platform information are uniformly quantified through the capability quantification module to obtain a standardized capability description.

[0144] S203. Through the AI ​​agent search module, iteratively execute steps S2031-S2033 until the preset termination condition is met: S2031. Based on its own search strategy, the current local optimal deployment scheme, relevant data of the defense area and standardized capability description, a large language model is used to generate candidate deployment schemes for available platform information.

[0145] S2032. Calculate the constraint violation degree of the candidate deployment scheme based on the preset constraint conditions, and calculate the comprehensive performance score of the candidate deployment scheme in combination with the target evaluation index when the constraint violation degree is less than or equal to the preset violation degree threshold.

[0146] S2033. When the overall performance score of the candidate deployment scheme is better than the overall performance score of the current local optimal deployment scheme, the candidate deployment scheme is updated to the current local optimal deployment scheme, and the updated local optimal deployment scheme is written into the shared memory area in the AI ​​agent search module. At the same time, the deployment scheme information of other AI agents is obtained from the shared memory area to guide the generation of the next candidate deployment scheme.

[0147] S204. When the preset termination conditions are met, the scheme determination module determines the globally optimal deployment scheme based on the information in the shared memory area.

[0148] Optionally, the data related to the defense area includes geographic information of the defense area, information on protected targets, and threat intelligence information; the method also includes: The data preprocessing module performs gridding on the defense area based on the geographical information of the defense area, dividing the defense area into grids with a preset precision, and using the center point of each grid as a candidate deployment location to obtain a set of candidate deployment locations; key protection areas are determined based on protection target information; and the direction of threat sources is determined based on threat intelligence information.

[0149] Optionally, based on its own search strategy, the current locally optimal deployment scheme, relevant data of the defense area, and standardized capability descriptions, a large language model is used to generate candidate deployment schemes for available platform information, including: The system integrates its own search strategy, the current local optimal deployment scheme, the set of candidate deployment locations, key protection areas, threat source directions, and standardized capability descriptions into natural language commands, inputs them into a large language model, and obtains platform location adjustment suggestions based on the current local optimal deployment scheme.

[0150] Based on the platform location adjustment suggestions, deployment locations are selected for each platform from the set of candidate deployment locations to form candidate deployment plans.

[0151] Optionally, the standardized capability description is a standardized capability vector, which includes detection capability value, interception capability value, response speed value, and terrain adaptation value.

[0152] Among them, the detection capability value is obtained by normalizing the detection distance, detection accuracy and detection field of view; the interception capability value is obtained by normalizing the interception distance, interception success rate and capacity parameters; the response speed value is obtained by normalizing the reaction time and maneuver speed; and the terrain adaptability value is obtained by evaluating the installation conditions and line-of-sight requirements.

[0153] Optionally, the preset constraints include terrain occlusion constraints, platform capability constraints, electromagnetic compatibility constraints, and resource supply constraints.

[0154] Terrain occlusion constraints require that the line-of-sight between the platform's deployment location and the defense area be greater than or equal to a preset line-of-sight threshold.

[0155] Platform capability constraints require that the distance between the deployment location of the platform used for detection and the boundary of the defense zone be less than or equal to the platform's maximum detection range.

[0156] Electromagnetic compatibility constraints require that the spacing between platforms operating in the same frequency band be greater than or equal to a preset minimum spacing.

[0157] Resource supply constraints require the platform to be deployed within the pre-defined infrastructure coverage area.

[0158] The constraint violation degree of candidate deployment schemes is calculated based on preset constraints, including: The degree of violation of the candidate deployment schemes against terrain occlusion constraints, platform capability constraints, electromagnetic compatibility constraints, and resource supply constraints is weighted and summed according to preset constraint weight coefficients to obtain the constraint violation degree.

[0159] Optionally, target evaluation metrics include coverage, average response time, and overall intercept probability.

[0160] The overall performance score of candidate deployment schemes is calculated by combining the target evaluation indicators, including: The comprehensive performance score is obtained by summing the coverage rate, the inverse normalized value of the average response time, and the comprehensive interception probability according to the preset target weight coefficient, and then subtracting the penalty term consisting of the constraint violation degree and the constraint violation penalty coefficient.

[0161] Optionally, the number of AI agents in the AI ​​agent search module is dynamically determined based on the number of platforms in the available platform information and the size of the candidate deployment location set.

[0162] The AI ​​agent search module also includes a strategy allocation unit, which is used to allocate a search strategy to each AI agent during the initialization phase. The search strategy is used to generate an initial deployment plan and to continuously guide the generation of candidate deployment plans in subsequent iterations.

[0163] Optionally, the search strategy itself includes at least one of the following: a random strategy, a greedy strategy, and a clustering strategy.

[0164] Among them, the random strategy randomly assigns locations from the set of candidate deployment locations, the greedy strategy prioritizes deployment based on the direction of threat sources and / or key protection areas, and the clustering strategy groups deployments according to platform functions.

[0165] Optionally, the method further includes: When the dynamic adjustment module detects updates to relevant data in the defense area or changes in available platform information, it takes the currently deployed global optimal deployment scheme as the current local optimal deployment scheme and calls the AI ​​intelligent agent search module to perform iterative operations to generate incremental adjustment schemes.

[0166] It should be noted that the various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for the method embodiments, since they are basically similar to the system embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the system embodiments. The system embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separate. The components indicated as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the solution in this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0167] The above description is merely one specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A low altitude defense platform deployment system, characterized by, The system comprises: a data acquisition module for acquiring defense area related data and available platform information; a capability quantification module for uniformly quantifying the capability parameters of each platform in the available platform information to obtain a standardized capability description; an AI agent search module comprising a plurality of parallel AI agents and a shared memory area; each AI agent is configured to iteratively perform the following operations until a preset termination condition is met: based on its own search strategy, the current local optimal deployment scheme, the defense area related data and the standardized capability description, a large language model is used to generate a candidate deployment scheme for the available platform information; based on the preset constraint condition, the constraint violation degree of the candidate deployment scheme is calculated, and when the constraint violation degree is less than or equal to a preset violation threshold, the comprehensive performance score of the candidate deployment scheme is calculated in combination with a target evaluation index; when the comprehensive performance score of the candidate deployment scheme is better than the comprehensive performance score of the current local optimal deployment scheme, the candidate deployment scheme is updated as the current local optimal deployment scheme, and the updated local optimal deployment scheme is written into the shared memory area, and the deployment scheme information of other AI agents is obtained from the shared memory area to guide the generation of the next candidate deployment scheme; a scheme determination module for determining a global optimal deployment scheme based on the information in the shared memory area when the preset termination condition is met.

2. The system of claim 1, wherein, The defense area related data includes defense area geographic information, defense target information and threat intelligence information; the system further comprises a data preprocessing module for: based on the defense area geographic information, the defense area is grid processed, the defense area is divided into grids of a preset precision, and the center point of each grid is taken as a candidate deployment position to obtain a candidate deployment position set; based on the defense target information, a key defense area is determined; based on the threat intelligence information, a threat source direction is determined.

3. The system of claim 2, wherein, The large language model is used to generate a candidate deployment scheme for the available platform information based on its own search strategy, the current local optimal deployment scheme, the defense area related data and the standardized capability description, comprising: the own search strategy, the current local optimal deployment scheme, the candidate deployment position set, the key defense area, the threat source direction and the standardized capability description are integrated into natural language instructions, which are input into the large language model to obtain platform position adjustment suggestions for the current local optimal deployment scheme; based on the platform position adjustment suggestions, deployment positions are selected for each platform from the candidate deployment position set to form the candidate deployment scheme.

4. The system of claim 1, wherein, The standardized capability description is a standardized capability vector, and the standardized capability vector comprises a detection capability value, an interception capability value, a response speed value and a terrain adaptation value; wherein the detection capability value is obtained by normalizing the detection distance, the detection accuracy and the detection field of view angle; the interception capability value is obtained by normalizing the interception distance, the interception success rate and the capacity parameter; the response speed value is obtained by normalizing the reaction time and the maneuvering speed; The terrain adaptability value is obtained by evaluating the installation conditions and line-of-sight requirements.

5. The system of claim 1, wherein, The preset constraints include terrain occlusion constraints, platform capability constraints, electromagnetic compatibility constraints, and resource supply constraints. The terrain occlusion constraint requires that the line-of-sight between the platform's deployment location and the defense area be greater than or equal to a preset line-of-sight threshold. The platform capability constraint requires that the distance between the deployment location of the platform used for detection and the boundary of the defense area be less than or equal to the platform's maximum detection range. The electromagnetic compatibility constraints require that the spacing between platforms operating in the same frequency band be greater than or equal to a preset minimum spacing. The resource supply constraints require that the platform be deployed within the coverage area of ​​a pre-defined infrastructure. The calculation of the constraint violation degree of the candidate deployment scheme based on preset constraint conditions includes: The degree of violation of the terrain occlusion constraint, platform capability constraint, electromagnetic compatibility constraint, and resource supply constraint by the candidate deployment scheme is weighted and summed according to a preset constraint weight coefficient to obtain the constraint violation degree.

6. The system of claim 1, wherein, The target evaluation metrics include coverage, average response time, and overall interception probability. The calculation of the overall performance score of the candidate deployment scheme by combining the target evaluation indicators includes: The comprehensive performance score is obtained by weighting and summing the coverage rate, the reciprocal normalized value of the average response time, and the comprehensive interception probability according to the preset target weight coefficient, and then subtracting the penalty term composed of the constraint violation degree and the constraint violation penalty coefficient.

7. The system of claim 2, wherein, The number of AI agents in the AI ​​agent search module is dynamically determined based on the number of platforms in the available platform information and the size of the candidate deployment location set; The AI ​​agent search module also includes a strategy allocation unit, which is used to allocate the search strategy to each AI agent during the initialization phase; the search strategy is used to generate an initial deployment plan and to continuously guide the generation of candidate deployment plans in subsequent iterations.

8. The system of claim 7, wherein, The self-search strategy includes at least one of the following: random strategy, greedy strategy, clustering strategy; The random strategy involves randomly assigning locations from the set of candidate deployment locations; the greedy strategy prioritizes deployment based on the direction of the threat source and / or the key protection areas; and the clustering strategy involves grouping and deploying according to platform functions.

9. The system of claim 1, wherein, It also includes a dynamic adjustment module, which, when it detects that the data related to the defense area has been updated or the information of the available platform has changed, takes the currently deployed global optimal deployment scheme as the current local optimal deployment scheme, and calls the AI ​​intelligent agent search module to perform iterative operations to generate an incremental adjustment scheme.

10. A low altitude defense platform deployment method, characterized by, The method, applied to the low-altitude defense platform deployment system according to any one of claims 1-9, comprises: The data acquisition module obtains relevant data on the defense area and information on available platforms. The capability parameters of each platform in the available platform information are uniformly quantified by the capability quantification module to obtain a standardized capability description. Through the AI ​​agent search module, perform the following operations iteratively until the preset termination condition is met: Based on its own search strategy, the current local optimal deployment scheme, the relevant data of the defense area, and the standardized capability description, a large language model is used to generate candidate deployment schemes for the available platform information. The constraint violation degree of the candidate deployment scheme is calculated based on the preset constraint conditions, and when the constraint violation degree is less than or equal to the preset violation degree threshold, the comprehensive performance score of the candidate deployment scheme is calculated in combination with the target evaluation index. When the overall performance score of the candidate deployment scheme is better than the overall performance score of the current local optimal deployment scheme, the candidate deployment scheme is updated to the current local optimal deployment scheme, and the updated local optimal deployment scheme is written into the shared memory area in the AI ​​agent search module. At the same time, the deployment scheme information of other AI agents is obtained from the shared memory area to guide the generation of the next candidate deployment scheme. When the preset termination conditions are met, the scheme determination module determines the globally optimal deployment scheme based on the information in the shared memory area.

Images