Knowledge and data driving-based blue army task planning method

Abstract

The invention discloses a blue army task planning method based on knowledge and data driving, and relates to the technical field of military simulation modeling and computer simulation, and the blue army task planning method comprises a sensing layer, a fusion layer, a reasoning layer, a planning layer, an optimization layer and an output layer. Bidirectional interaction between the layers is achieved through data interfaces, and deep fusion of knowledge and data in the whole process is ensured; by constructing a dynamically updated knowledge graph and a knowledge-data mapping mechanism, domain knowledge can guide data modeling in the whole process, meanwhile, the data can optimize a knowledge system in real time, the limitation of a single driving mode is solved, and through testing, in a complex military drill scene, the knowledge and data can be fully fused. The matching degree of the fusion-driven planning scheme and the actual situation reaches 92%, is improved by 25% compared with a knowledge-driven scheme, and is improved by 30% compared with a data-driven scheme.

Description

Technical Field

[0001] This invention belongs to the field of military simulation modeling and computer simulation technology, and in particular relates to a knowledge- and data-driven blue force mission planning method. Background Technology

[0002] In fields such as military training, cybersecurity, and emergency drills, the role of the "enemy" (or opposing force) is becoming increasingly important. The scientific and rational nature of its mission planning directly determines the effectiveness of the exercise and the validity of the defense's capability assessment. Traditional blue force mission planning mainly relies on human experience. Planners formulate plans based on their knowledge of the domain and understanding of historical cases. This model has obvious limitations: on the one hand, manual planning is inefficient and cannot adapt to the rapid planning needs of multiple tasks, resources, and constraints in complex scenarios; on the other hand, human decision-making is easily influenced by subjective experience biases, making it difficult to fully consider changes in the situation and potential risks, resulting in insufficient adaptability and optimality of the planning scheme.

[0003] With the development of information technology, task planning technologies based on single-driven models have gradually emerged. Among them, knowledge-driven planning technology generates planning schemes by constructing domain rule bases or knowledge graphs and using reasoning mechanisms, which can ensure the logic and domain compliance of the schemes. However, this technology does not make sufficient use of dynamic data and is difficult to flexibly adjust the schemes according to real-time situations (such as adjustments to Red Army deployments or changes in environmental parameters). Data-driven planning technology relies on machine learning algorithms to analyze historical data and can achieve predictive optimization of the schemes. However, this technology lacks deep integration with domain expertise, and the reliability of the planning results drops significantly when historical data is insufficient or the scenario changes abruptly.

[0004] Currently, complex adversarial scenarios place higher demands on the blue force's mission planning: it requires the logical support of domain knowledge, the ability to mine patterns from historical data, and the dynamic response capability to real-time situations. Existing single-driven planning technologies can no longer meet these requirements. Therefore, developing a blue force mission planning technology that integrates knowledge and data as dual drivers has become a key direction for improving the level of blue force mission planning. Summary of the Invention

[0005] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution:

[0006] This invention is a knowledge- and data-driven blue team mission planning method, comprising a perception layer, a fusion layer, a reasoning layer, a planning layer, an optimization layer, and an output layer;

[0007] Each layer interacts bidirectionally through data interfaces, ensuring deep integration of knowledge and data throughout the entire process.

[0008] Furthermore, the perception layer is responsible for data acquisition, including static data and dynamic data. Static data includes domain knowledge data and historical task data; dynamic data is real-time situational awareness data. The perception layer includes a multi-source data collector and a data preprocessing module.

[0009] The fusion layer is responsible for realizing the fusion processing of knowledge and data, and the fusion layer includes a knowledge graph construction and update module, a data feature extraction module, and a fusion decision module.

[0010] Knowledge Graph Construction and Update Module: Based on static domain knowledge data collected by the perception layer, an initial KG is constructed through technologies such as entity recognition, relation extraction, and attribute annotation. At the same time, by utilizing the patterns in dynamic situational data and historical task data, the entity relationships in the KG are dynamically updated through ML algorithms (such as adding the attribute "success rate of a certain tactic in a specific environment").

[0011] Data feature extraction module: Extracts features from historical and real-time data collected by the perception layer, such as key features like "task type", "resource input", and "frequency of situational change", to provide data support for subsequent reasoning and planning;

[0012] Fusion Decision Module: Establishes a mapping relationship between knowledge and data, and associates and matches the extracted data features with entities and relationships in the KG to form a fused knowledge-data set;

[0013] The reasoning layer is responsible for reasoning through the fused knowledge-data set to generate constraints and candidate solutions for task planning; the reasoning layer includes a knowledge reasoning engine and a data prediction engine.

[0014] Knowledge Reasoning Engine: It adopts a combination of rule-based reasoning (RBR) and case-based reasoning (CBR), and uses the rule and case knowledge in the KG to reason out the basic constraints of task planning (such as resource limits, tactical applicability conditions, time windows, etc.).

[0015] Data prediction engine: Utilizes ML algorithms (such as LSTM, random forest, etc.) to predict the trend of situational changes in fused data, such as predicting possible deployment adjustments of the Red Army, changing trends of environmental parameters, etc., to provide dynamic constraints for planning;

[0016] The planning layer is responsible for generating an initial blue team mission planning scheme based on the constraints output by the inference layer. The mission planning engine (TPE) within the planning layer employs a three-level process of "objective decomposition - resource allocation - step planning," including:

[0017] Target decomposition: The overall mission objective of the Blue Force is decomposed into several sub-objectives (such as "breaking through defenses" being decomposed into sub-objectives such as "reconnaissance and location", "jamming and suppression", and "assault operation").

[0018] Resource allocation: Based on equipment knowledge in the KG and resource allocation patterns in historical data, allocate corresponding equipment, personnel, time and other resources to each sub-objective;

[0019] Step planning: Based on the logical relationship and time constraints of the sub-goals, formulate the execution steps and sequence of each sub-goal to form an initial planning scheme;

[0020] The optimization layer is responsible for multi-objective optimization of the initial planning scheme to improve its optimality. The multi-objective optimization (MOP) engine in the optimization layer uses the NSGA-III algorithm, with "maximizing task success rate," "minimizing resource consumption," and "maximizing concealment" as core optimization objectives. It constructs an optimization objective function, including:

[0021] Task success rate: Calculated based on the execution success rate of similar solutions in historical data and real-time situation prediction results;

[0022] Resource consumption: Total input and estimated loss of various resources in the statistical plan;

[0023] Concealment: Assessment based on knowledge of tactical concealment in the KG and data on the impact of the real-time environment on signals and actions;

[0024] Through iterative optimization using algorithms, a set of solutions consisting of multiple non-inferior optimal solutions is generated.

[0025] The output layer is used to display, evaluate, and output the optimized solution; the output layer includes a solution visualization module, a solution evaluation module, and a solution output interface;

[0026] The scheme visualization module is used to display the sub-objectives, execution steps, resource allocation, and other contents of the planning scheme in the form of charts, time series diagrams, etc.

[0027] The scheme evaluation module is used to evaluate the feasibility and effectiveness of the schemes in the scheme set through simulation (combining the tactical rules in KG and the situational changes predicted by data), and output an evaluation report;

[0028] The solution output interface is used to output the optimal solution (selected according to the evaluation report) in a standardized format (such as XML, JSON), which can be connected to external systems such as command systems and exercise platforms.

[0029] Furthermore, the workflow of the knowledge graph construction and update module is as follows:

[0030] First, the static domain knowledge data (such as tactical manuals and equipment manuals) collected by the perception layer is parsed. Entity recognition (such as "tank" and "ambush tactics") is achieved through the BERT model, and relation extraction (such as "tanks are suitable for ambush tactics") is achieved through dependency parsing. Combined with manually labeled attribute information (such as "tank speed: 60km / h"), an initial knowledge base (KG) is constructed. Second, dynamic data (such as "ambush tactic success rate 80%" in the current exercise) and historical task data (such as "tank formation success rate in hilly terrain 75%" in the past 100 exercises) from the perception layer are received in real time. The XGBoost algorithm is used to mine the correlation patterns in the data and generate knowledge update suggestions (such as adding the attribute "75% formation success rate in hilly terrain" to "ambush tactics"). Finally, the update suggestions are verified through a knowledge verification mechanism (combining a domain expert rule base and historical data consistency verification). After successful verification, the KG is updated to ensure the accuracy and timeliness of the KG.

[0031] Furthermore, the task planning engine workflow is as follows:

[0032] The target decomposition module, based on the "task-subtask" relationship knowledge in the KG and historical task decomposition cases, uses the CBR algorithm to automatically decompose the overall target. The resource allocation module establishes a "resource-subtarget" matching model, which uses equipment performance knowledge in the KG as constraints and historical resource allocation data as training samples to output a resource allocation scheme through a neural network algorithm. The step planning module, based on the "subtask logical relationship" knowledge in the KG (such as "reconnaissance and positioning" must precede "assault action") and time constraints in real-time situational data (such as "Red Army troop rotation time window: 14:00-14:30"), generates subtask execution steps through a time-series planning algorithm, and finally integrates them to form an initial planning scheme.

[0033] Includes the following steps:

[0034] (1) The perception layer starts the multi-source data collector to collect static domain knowledge data, historical task data and real-time situational awareness data. After being processed by the data preprocessing module, the data is transmitted to the fusion layer.

[0035] (2) The fusion layer constructs and updates the KG through the knowledge graph construction and update module, extracts data features through the data feature extraction module, and then realizes the fusion of knowledge and data through the fusion decision module to generate a fusion dataset;

[0036] (3) The knowledge reasoning engine of the reasoning layer performs reasoning based on KG to generate basic constraints; the data prediction engine performs situation prediction based on the fused dataset to generate dynamic constraints; the two types of constraints are jointly transmitted to the planning layer.

[0037] (4) The TPE at the planning layer generates an initial blue team mission planning scheme based on the constraints through target decomposition, resource allocation, and step planning;

[0038] (5) The MOP engine of the optimization layer takes the initial scheme as input, generates a set of non-dominated optimal schemes through a multi-objective optimization algorithm, and transmits it to the output layer;

[0039] (6) The output layer evaluates the set of schemes through the scheme evaluation module, displays the results in conjunction with the scheme visualization module, and finally outputs the optimal planning scheme through the scheme output interface.

[0040] The present invention has the following beneficial effects:

[0041] 1. This invention achieves deep integration of knowledge and data: by constructing a dynamically updated knowledge graph and a knowledge-data mapping mechanism, domain knowledge guides data modeling throughout the process, while data can optimize the knowledge system in real time, overcoming the limitations of a single-driven model. Testing shows that in complex military exercise scenarios, the fusion-driven planning scheme achieves a 92% match with the actual situation, a 25% improvement over the knowledge-driven scheme and a 30% improvement over the data-driven scheme.

[0042] 2. This invention enhances the dynamic adaptability of the planning scheme: relying on real-time situational awareness data and a data prediction engine, it can quickly capture changes in the situation and adjust the planning scheme accordingly. In cybersecurity drills, the response time for scheme adjustments is reduced to less than 5 seconds, an improvement of 83% compared to existing schemes (more than 30 seconds), effectively responding to the Red Army's dynamic defense adjustments.

[0043] 3. This invention optimizes multi-objective performance: Through the NSGA-Ⅲ multi-objective optimization algorithm, the planning scheme achieves global optimality in terms of task success rate, resource consumption, concealment and other indicators.

[0044] 4. This invention significantly improves planning efficiency: the automated full-process processing replaces the core steps of manual planning, greatly improving efficiency and reducing errors caused by human subjective bias.

[0045] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is the overall system architecture diagram;

[0048] Figure 2 It is a flowchart of knowledge graph construction and updating;

[0049] Figure 3 This is a diagram of the internal logic structure of the task planning engine;

[0050] Figure 4 This is a flowchart of the multi-objective optimization engine. Detailed Implementation

[0051] 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.

[0052] Please see Figure 1-4 As shown, this invention is a knowledge- and data-driven blue team mission planning method, comprising a six-layer architecture of "perception-fusion-reasoning-planning-optimization-output," as follows: Figure 1 As shown, each layer interacts bidirectionally through data interfaces, ensuring deep integration of knowledge and data throughout the entire process.

[0053] (1) Perception Layer: Responsible for data acquisition, including static and dynamic data. Static data includes domain knowledge data (such as tactical manuals, equipment parameters, rule bases, etc.) and historical mission data (such as past exercise plans, execution results, evaluation reports, etc.); dynamic data is real-time situational awareness data (such as Red Army deployment data, environmental monitoring data, Blue Army equipment status data, etc.). The core modules include a multi-source data acquisition unit and a data preprocessing module (which realizes data cleaning, format conversion, deduplication, etc.).

[0054] (2) Fusion Layer: This layer enables the fusion processing of knowledge and data and is the core support layer of this invention. The core modules include a knowledge graph construction and update module, a data feature extraction module, and a fusion decision module.

[0055] Knowledge Graph Construction and Update Module: Based on static domain knowledge data collected by the perception layer, an initial KG is constructed through technologies such as entity recognition, relation extraction, and attribute annotation. At the same time, by utilizing the patterns in dynamic situational data and historical task data, the entity relationships in the KG are dynamically updated through ML algorithms (such as adding the attribute "success rate of a certain tactic in a specific environment").

[0056] Data feature extraction module: Extracts features from historical and real-time data collected by the perception layer, such as key features like "task type", "resource input", and "frequency of situational change", to provide data support for subsequent reasoning and planning.

[0057] Fusion Decision Module: Establishes a mapping relationship between knowledge and data, and associates and matches the extracted data features with entities and relationships in the KG to form a fused knowledge-data set.

[0058] (3) Reasoning layer: Based on the fused knowledge-data set, reasoning is performed to generate constraints and candidate solutions for task planning. The core modules include the knowledge reasoning engine and the data prediction engine.

[0059] Knowledge Reasoning Engine: It adopts a combination of rule-based reasoning (RBR) and case-based reasoning (CBR), and uses the rule and case knowledge in the KG to reason out the basic constraints of task planning (such as resource limits, tactical applicability conditions, time windows, etc.).

[0060] Data prediction engine: Utilizes ML algorithms (such as LSTM, random forest, etc.) to predict the trend of situational changes in fused data, such as predicting possible deployment adjustments of the Red Army, changing trends of environmental parameters, etc., to provide dynamic constraints for planning.

[0061] (4) Planning Layer: Based on the constraints output by the reasoning layer, an initial blue team task planning scheme is generated. The core module is the Task Planning Engine (TPE), which adopts a three-level process of "goal decomposition - resource allocation - step planning":

[0062] Target decomposition: The overall mission objective of the Blue Force is decomposed into several sub-objectives (such as "breaking through defenses" being decomposed into sub-objectives such as "reconnaissance and location", "jamming and suppression", and "assault operation").

[0063] Resource Allocation: Based on equipment knowledge in the KG and resource allocation patterns in historical data, allocate corresponding equipment, personnel, time, and other resources to each sub-objective. Step Planning: According to the logical relationships and time constraints of the sub-objectives, formulate the execution steps and their sequence for each sub-objective, forming an initial planning scheme.

[0064] (5) Optimization Layer: Performs multi-objective optimization on the initial planning scheme to improve its optimality. The core module is the Multi-Objective Optimization (MOP) engine, which adopts the NSGA-III algorithm. The core optimization objectives are "maximizing task success rate", "minimizing resource consumption", and "maximizing concealment". The optimization objective function is constructed as follows:

[0065] Task success rate: Calculated based on the execution success rate of similar solutions in historical data and real-time situation prediction results.

[0066] Resource consumption: The total amount of input and estimated loss of various resources in the statistical plan.

[0067] Concealment: Assessment based on knowledge of tactical concealment in the KG and data on the impact of the real-time environment on signals and actions.

[0068] Through iterative optimization using algorithms, a set of solutions consisting of multiple non-inferior optimal solutions is generated.

[0069] (6) Output layer: Implements the display, evaluation and output of the optimized solution. The core modules include the solution visualization module, the solution evaluation module and the solution output interface.

[0070] The solution visualization module displays the sub-goals, execution steps, resource allocation, and other content of the planning solution in the form of charts, time series diagrams, etc.

[0071] The scheme evaluation module evaluates the feasibility and effectiveness of the schemes in the scheme set through simulation (combining the tactical rules in the KG and the situational changes predicted by data) and outputs an evaluation report.

[0072] Solution output interface: Outputs the optimal solution (selected based on the evaluation report) in a standardized format (such as XML, JSON), which can be connected to external systems such as command systems and exercise platforms.

[0073] (II) Detailed Design of Core Modules

[0074] 1. Knowledge Graph Construction and Update Module

[0075] This module is the core of the fusion layer, and its workflow is as follows: Figure 2 As shown, the process involves several steps. First, static domain knowledge data (such as tactical manuals and equipment manuals) collected by the perception layer is parsed. Entity recognition (e.g., "tank" and "ambush tactics") is achieved using the BERT model, and relation extraction (e.g., "tanks are suitable for ambush tactics") is performed using dependency parsing. An initial knowledge base (KG) is constructed by combining manually labeled attribute information (e.g., "tank speed: 60km / h"). Second, dynamic data (e.g., "ambush tactic success rate 80%" in the current exercise) and historical task data (e.g., "tank formation success rate in hilly terrain 75%" in the past 100 exercises) from the perception layer are received in real time. The XGBoost algorithm is used to mine the correlation patterns in the data and generate knowledge update suggestions (e.g., adding the attribute "75% formation success rate in hilly terrain" to "ambush tactics"). Finally, the update suggestions are verified through a knowledge verification mechanism (combining a domain expert rule base and historical data consistency checks). Once verification is successful, the KG is updated to ensure its accuracy and timeliness.

[0076] 2. Task Planning Engine (TPE)

[0077] TPE employs a hierarchical planning strategy, and its internal logical structure is as follows: Figure 3 As shown. The target decomposition module, based on the "task-subtask" relationship knowledge in the KG and historical task decomposition cases, automatically decomposes the overall target using the CBR algorithm; the resource allocation module establishes a "resource-subtarget" matching model, which uses equipment performance knowledge in the KG as constraints and historical resource allocation data as training samples, and outputs a resource allocation scheme through a neural network algorithm; the step planning module, based on the "subtask logical relationship" knowledge in the KG (such as "reconnaissance and positioning" must precede "assault action") and time constraints in real-time situational data (such as "Red Army troop rotation time window: 14:00-14:30"), generates subtask execution steps through a time-series planning algorithm, and finally integrates them to form an initial planning scheme.

[0078] 3. Multi-Objective Optimization (MOP) Engine

[0079] The optimization process of the MOP engine is as follows: Figure 4 As shown, the initial schemes output by the planning layer are used as the initial population for optimization. Secondly, using "task success rate," "resource consumption," and "stealth" as objective functions, the objective function value for each individual (scheme) is calculated based on the fusion data provided by the fusion layer. Then, the NSGA-III algorithm is used for selection, crossover, and mutation operations to iteratively generate a new population. Simultaneously, "tactical feasibility rules" from the KG framework (such as "prohibiting the simultaneous use of two mutually exclusive tactics") are introduced as constraints to verify the feasibility of individuals during the iteration process, eliminating violating schemes. Finally, when the maximum number of iterations is reached or the objective function value converges, the set of non-dominated optimal schemes is output.

[0080] (III) Work Process

[0081] The complete workflow of the knowledge- and data-driven blue team mission planning technology of this invention is as follows:

[0082] (1) The perception layer starts the multi-source data collector to collect static domain knowledge data, historical task data and real-time situational awareness data. After being processed by the data preprocessing module, the data is transmitted to the fusion layer.

[0083] (2) The fusion layer constructs and updates the KG through the knowledge graph construction and update module, extracts data features through the data feature extraction module, and then realizes the fusion of knowledge and data through the fusion decision module to generate a fusion dataset.

[0084] (3) The knowledge reasoning engine of the reasoning layer performs reasoning based on KG to generate basic constraints; the data prediction engine performs situation prediction based on the fused dataset to generate dynamic constraints; the two types of constraints are jointly transmitted to the planning layer.

[0085] (4) The TPE at the planning layer generates an initial blue team mission planning scheme based on the constraints through target decomposition, resource allocation, and step planning.

[0086] (5) The MOP engine of the optimization layer takes the initial scheme as input, generates a set of non-dominated optimal schemes through a multi-objective optimization algorithm, and transmits it to the output layer.

[0087] (6) The output layer evaluates the set of schemes through the scheme evaluation module, displays the results in conjunction with the scheme visualization module, and finally outputs the optimal planning scheme through the scheme output interface.

[0088] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0089] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A knowledge- and data-driven blue team mission planning method, characterized in that: It includes a perception layer, a fusion layer, a reasoning layer, a planning layer, an optimization layer, and an output layer; Each layer interacts bidirectionally through data interfaces, ensuring deep integration of knowledge and data throughout the entire process.

2. The knowledge- and data-driven blue team mission planning method according to claim 1, characterized in that, The perception layer is responsible for data acquisition, including static data and dynamic data. Static data includes domain knowledge data and historical task data; dynamic data is real-time situational awareness data. The perception layer includes a multi-source data collector and a data preprocessing module. The fusion layer is responsible for realizing the fusion processing of knowledge and data, and the fusion layer includes a knowledge graph construction and update module, a data feature extraction module, and a fusion decision module. The knowledge graph construction and update module constructs an initial KG using static domain knowledge data collected based on the perception layer, through entity recognition, relation extraction, and attribute annotation techniques; simultaneously, it uses ML algorithms to dynamically update entity relationships in the KG based on patterns in dynamic situational data and historical task data. The data feature extraction module is used to extract features from historical and real-time data collected by the perception layer, providing data support for subsequent reasoning and planning. The fusion decision module is used to establish a mapping relationship between knowledge and data, and to associate and match the extracted data features with entities and relationships in the KG to form a fused knowledge-data set; The reasoning layer is responsible for reasoning through the fused knowledge-data set to generate constraints and candidate solutions for task planning; the reasoning layer includes a knowledge reasoning engine and a data prediction engine. The knowledge reasoning engine adopts a combination of rule-based reasoning and case-based reasoning, using the rule and case knowledge in the KG to deduce the basic constraints of task planning. The data prediction engine uses ML algorithms to predict the trend of situational changes in the fused data, providing dynamic constraints for planning; The planning layer is responsible for generating an initial blue team mission planning scheme based on the constraints output by the inference layer. The task planning engine in the planning layer adopts a three-level process of "goal decomposition - resource allocation - step planning", including: Target decomposition module: Decomposes the overall mission objective of the Blue Team into several sub-objectives; Resource allocation module: Based on equipment knowledge in KG and resource allocation patterns in historical data, allocate corresponding equipment, personnel, and time resources to each sub-objective; Step planning module: Based on the logical relationship and time constraints of the sub-goals, formulate the execution steps and sequence of each sub-goal to form an initial planning scheme; The optimization layer is responsible for multi-objective optimization of the initial planning scheme to improve its optimality. The multi-objective optimization engine in the optimization layer uses the NSGA-III algorithm, with "maximizing task success rate," "minimizing resource consumption," and "maximizing concealment" as optimization objectives. The optimization objective function is constructed, including: Task success rate: Calculated based on the execution success rate of similar solutions in historical data and real-time situation prediction results; Resource consumption: Total input and estimated loss of various resources in the statistical plan; Concealment: Assessment based on knowledge of tactical concealment in the KG and data on the impact of the real-time environment on signals and actions; Through iterative optimization using algorithms, a set of solutions consisting of multiple non-inferior optimal solutions is generated. The output layer is used to display, evaluate, and output the optimized solution; the output layer includes a solution visualization module, a solution evaluation module, and a solution output interface; The scheme visualization module is used to display the sub-objectives, execution steps, and resource allocation of the planning scheme in the form of charts and time series diagrams; The scheme evaluation module is used to evaluate the feasibility and effectiveness of the schemes in the scheme set through simulation and deduction, and output an evaluation report; The solution output interface is used to output the optimal solution in a standardized format, which can be connected to external systems such as command systems and exercise platforms.

3. The knowledge- and data-driven blue team mission planning method according to claim 2, characterized in that, The workflow of the knowledge graph construction and update module is as follows: S1. Perform text parsing on the static domain knowledge data collected by the perception layer, realize entity recognition through the BERT model, realize relation extraction through dependency parsing, and construct the initial KG by combining manually labeled attribute information; S2. Receive dynamic data and historical task data from the perception layer in real time, and use the XGBoost algorithm to mine the correlation patterns in the data to generate knowledge update suggestions; S3. Verify the update suggestions through a knowledge verification mechanism. Once the verification is successful, update the KG to ensure the accuracy and timeliness of the KG.

4. The knowledge- and data-driven blue team mission planning method according to claim 2, characterized in that, The task planning engine workflow is as follows: The target decomposition module automatically decomposes the overall target based on the "task-subtask" relationship knowledge in the KG and historical task decomposition cases, using the CBR algorithm. The resource allocation module establishes a "resource-sub-target" matching model. This model is constrained by the equipment performance knowledge in the KG, uses historical resource allocation data as training samples, and outputs a resource allocation scheme through a neural network algorithm. The step planning module generates subtask execution steps based on the "subtask logical relationship" knowledge in KG and the time constraints in real-time situational data through a time-series planning algorithm, and finally integrates them to form an initial planning scheme.

5. A knowledge- and data-driven blue team mission planning method according to claim 1, comprising the following steps: (1) The perception layer starts the multi-source data collector to collect static domain knowledge data, historical task data and real-time situational awareness data. After being processed by the data preprocessing module, the data is transmitted to the fusion layer. (2) The fusion layer constructs and updates the KG through the knowledge graph construction and update module, extracts data features through the data feature extraction module, and then realizes the fusion of knowledge and data through the fusion decision module to generate a fusion dataset; (3) The knowledge reasoning engine of the reasoning layer performs reasoning based on KG and generates basic constraints; The data prediction engine performs situational prediction based on the fused dataset and generates dynamic constraints; Both types of constraints are transmitted to the planning layer; (4) The TPE at the planning layer generates an initial blue team mission planning scheme based on the constraints through target decomposition, resource allocation, and step planning; (5) The MOP engine of the optimization layer takes the initial scheme as input, generates a set of non-dominated optimal schemes through a multi-objective optimization algorithm, and transmits it to the output layer; (6) The output layer evaluates the set of schemes through the scheme evaluation module, displays the results in conjunction with the scheme visualization module, and finally outputs the optimal planning scheme through the scheme output interface.

Images