A service-based kill chain analysis construction method
By constructing a service resource pool, performing local situation updates, and reducing the dimensionality of directed hypergraphs, combined with service-oriented twin optimization, the real-time problem of kill chain generation under frequent situational changes and state space catastrophes was solved, achieving efficient kill chain generation.
Patent Information
- Application Number
- CN202610066236.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-19
- Publication Date
- 2026-05-29
- Estimated Expiration
- 2046-01-19
AI Technical Summary
Existing kill chain generation methods struggle to meet real-time requirements under conditions of frequent battlefield situational shifts and state space catastrophes, leading to accumulated solution delays and an inability to effectively cope with complex and ever-changing combat missions.
By constructing a service-based kill chain analysis method, and utilizing service resource pools, local situation updates, directed hypergraph dimensionality reduction, and service twin optimization, we can reduce cold starts during situation transitions, decrease latency accumulation, and ensure the real-time generation of kill chains.
It effectively mitigates state space catastrophe, reduces invalid computation and cold start frequency in link construction, ensures that the kill chain generation speed meets the real-time requirements of actual combat, and improves the efficiency of kill chain construction under frequent situational changes.
Smart Images

Figure CN121543455B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a service-based kill chain analysis construction method, belonging to the field of collaborative combat technology. Background Technology
[0002] The kill chain is a product of information warfare. It typically refers to the closed-loop process of detecting a target, aiming at the target, engaging the enemy, and assessing the outcome of the engagement. It plays a crucial role in the organization and overall advancement of combat operations.
[0003] A Chinese patent application with publication number CN119249878A discloses a method for constructing a kill chain generation model, including the following steps: defining entities, instantiating the attributes of entities, acquiring human experience and battlefield intelligence, and constructing a constraint relationship knowledge graph based on human experience and battlefield intelligence; constructing a kill chain generation agent based on the attributes of entities; abstracting the constraint relationship knowledge graph into an action space mask; training the kill chain generation agent through a deep reinforcement learning algorithm and the action space mask to obtain a kill chain generation model.
[0004] Although existing technologies have solved the problems of long computation time and slow decision-making in traditional methods, and are conducive to realizing combat missions with complex and ever-changing battlefield situations, they have not considered the synergistic effects of frequent situational changes and state space catastrophes. Specifically, frequent situational changes in the battlefield environment will trigger multiple cold starts in the generation of kill chains. As the number of equipment increases, the possible combinations of kill chains grow exponentially, forming state space catastrophes. The superposition of the two leads to the continuous accumulation of solution delays, eventually exceeding the time threshold for dynamic reconstruction of kill chains, making it difficult to meet the core requirement of real-time kill chain generation in actual combat. Summary of the Invention
[0005] To address the shortcomings of existing technologies, the present invention aims to provide a service-based kill chain analysis construction method. By standardizing services, updating the situation locally, and reducing constraints through pruning, the method curbs state space catastrophes. It also reduces the cold start of situation switching through local hot replacement, twin optimization, and template reuse. The two methods work together to reduce latency accumulation and solve the real-time problem.
[0006] To achieve the above objectives, the present invention provides the following technical solution:
[0007] A service-based kill chain analysis construction method includes:
[0008] A service resource pool is built based on service points, and a service discovery mechanism is set up to update the service resource pool;
[0009] Based on the fused situation set, situation switching events are identified to locally update the fused situation set, generate a situation awareness set, and construct a three-level situation view.
[0010] Construct directed hyperedges, combine them with functional category constraints to generate a directed hypergraph, generate a dimensionality-reduced state space based on multidimensional constraints, and select an initial combat candidate set from the service resource pool.
[0011] Configure the management node, send mission invitation information to the initial combat candidate set in combination with sub-tasks, generate response schemes and perform adaptation screening, generate the initial kill chain, and perform failure verification.
[0012] A service twin is constructed to deduce the initial kill chain, and the execution delay of the computing end is used for elimination and reorganization or delay optimization.
[0013] Build service composition templates and establish a performance feedback mechanism to score the service composition templates to classify them into high-efficiency and low-efficiency templates.
[0014] Specifically, the steps for building a service resource pool include:
[0015] Obtain a comprehensive list of equipment across the entire combat system and categorize it.
[0016] Each equipment entity is converted into a service point, and an interface architecture is set up for each service point.
[0017] Configure master and slave nodes, and set up service discovery mechanisms, including active reporting and passive detection;
[0018] For each service point, obtain the operational status coefficient, task success rate, and resource redundancy rate, and calculate the health score of the service point.
[0019] Based on historical combat data, supplement the association logic of service points to build a service resource pool.
[0020] Specifically, the steps for identifying situational change events include:
[0021] A three-dimensional fusion framework is constructed, with real-time data from the service resource pool as the core layer, target data and environmental interference data collected by sensors in the combat system as the perception layer, and the initial combat mission as the constraint layer.
[0022] Based on the association logic of each service point in the service resource pool, historical matching and real-time adaptation are performed to generate candidate links;
[0023] For each candidate link, calculate the initial fitness score of the candidate link, and take the candidate link with the maximum initial fitness score as the initial link.
[0024] Based on the aforementioned 3D fusion framework and the initial link, a fusion situation set is generated, and format consistency and logical rationality are verified.
[0025] The configuration includes a dual-mode event recognition mechanism that combines real-time monitoring and periodic verification, and identifies situational switching events based on the fused situation set.
[0026] Specifically, the steps for identifying situational change events also include:
[0027] Based on the situational change event, the impact range coefficient, task urgency, and resource correlation are obtained, and the event urgency of the situational change event is calculated through weighted calculation.
[0028] Based on the configured situation classification criteria, an event list of the situation switching events is generated;
[0029] Based on the urgency of the events in the event list, a tiered triggering mechanism is established, and differentiated update strategies are matched to generate triggering instructions for the situational switching events.
[0030] Using association depth, functional importance, resource dependence, time sensitivity, and redundancy as evaluation indicators, and defining quantitative rules for each indicator, a link impact evaluation system is constructed.
[0031] Using candidate links as the evaluation object, calculate the link impact, set an impact threshold, and filter out links and associated service points whose impact exceeds the impact threshold to form an affected mapping table;
[0032] Configure local update rules and build an incremental update engine to generate a situational awareness set through data extraction, logic reconstruction, consistency verification, and integrated storage.
[0033] Specifically, the steps for constructing a three-level situation view include:
[0034] Based on the aforementioned situational awareness set, a target layer is constructed using a structure of target, identifier, and constraint.
[0035] A task layer is constructed based on task decomposition, subtask definition, and priority sorting.
[0036] The service point set is selected from the situational awareness set, and a service layer is constructed through the bidirectional mapping relationship between subtasks and service points to generate a three-level situational view;
[0037] Obtain the priority coefficient of the subtask, the adaptability between the service point and the subtask, and the service point collaboration compatibility. Calculate the hierarchical linkage adaptability through weighted calculation.
[0038] Set up a hierarchical linkage mechanism, including upward linkage, downward linkage, and cross-level linkage;
[0039] After the linkage adjustment, the hierarchical linkage adaptability of the new service point set is calculated, and the linkage adjustment is effectively determined by the adaptability threshold.
[0040] Specifically, the steps for generating the dimensionality-reduced state space include:
[0041] Based on the aforementioned three-level situational view, a three-layer node system for the directed hypergraph is defined, including target layer nodes, task layer nodes, and service layer nodes, and a set of directed hyperedge nodes is generated.
[0042] Design internal collaborative hyperedges and cross-layer adaptive hyperedges to generate a set of directed hyperedges;
[0043] Set multi-dimensional constraints, including functional category constraints, health constraints, timing constraints, and resource redundancy constraints;
[0044] For each hyperedge in the directed hyperedge set, the functional category matching degree, health compliance degree, timing compliance degree, and resource redundancy adaptability are obtained through the quantitative standard of multi-dimensional constraints. The constraint satisfaction degree of each hyperedge is calculated, and invalid hyperedges are screened out.
[0045] The final weight of the hyperedge is calculated based on the product of the constraint satisfaction and the hyperedge weight, so as to select the hyperedges to be pruned and the directed hypergraph after constraint embedding.
[0046] Specifically, the steps for generating the dimensionality-reduced state space also include:
[0047] Based on the affected mapping table and the situational switching event, the affected region in the directed hypergraph is located, a three-level pruning priority rule is set, and the affected region and its associated hyperedges are pruned in priority order, and consistency verification is performed.
[0048] Based on the pruned directed hypergraph, a state space description system is constructed, including path dimension, constraint dimension, and efficiency dimension, to generate a dimensionality-reduced state space.
[0049] Based on functional category constraints, a preliminary candidate pool is selected from the service resource pool;
[0050] Substitute the service points in the preliminary candidate pool into the reduced state space to check whether the service points are adapted to the temporal relationship and link constraints of the directed superedge, and generate a verification candidate pool.
[0051] The service points in the verification candidate pool are sorted and categorized hierarchically to generate an initial combat candidate set.
[0052] Specifically, the steps for generating the initial kill chain include:
[0053] Based on the task layer of the three-level situational view, management nodes are configured for subtasks, and a set of management nodes is generated by combining the bidirectional mapping relationship.
[0054] Using the management node as the task inviter, task invitation information is constructed;
[0055] Based on the initial set of combat candidates, mission invitations are distributed in a targeted manner through management nodes;
[0056] Set response time limits, collect valid response solutions, and generate a set of response solutions for the subtask;
[0057] Acquire functional matching degree, performance compliance degree, state compatibility degree, and computing power adaptability;
[0058] The comprehensive rating score for each response plan is calculated using failure weight, capability adaptability, execution cost, and estimated completion time as rating indicators.
[0059] The response scheme set is sorted in descending order, and the optimal service point is selected by combining the reduced state space. The optimal service points of all subtasks are integrated, and the optimal service point combination is generated based on the directed hypergraph.
[0060] Based on the logic of optimal service point combination and combat process, an initial kill chain is generated, and real-time failure verification is performed based on the service resource pool.
[0061] Specifically, the steps for deducing the initial kill chain include:
[0062] Extract the optimal service point combination, collaborative timing relationship, data interaction path and sub-task constraints, and access multi-source synchronized data to build a service twin;
[0063] Based on the service twin, the execution flow of the initial kill chain is simulated to obtain various latency parameters and the end-to-end total latency of the initial kill chain is obtained.
[0064] By using service-oriented twins to simulate latency changes in various scenarios, multi-scenario latency simulations are generated.
[0065] If the total end-to-end latency exceeds the real-time threshold, an optimization decision is triggered; otherwise, the initial kill chain is retained.
[0066] For optimization decisions, the predicted delay value in the multi-scenario delay simulation is compared with the real-time threshold to determine whether a predicted delay occurs.
[0067] If the number of predicted delays exceeds the simulation limit, elimination and reorganization will be carried out, triggering a re-selection of the optimal service point combination; otherwise, delay optimization will be performed to select alternative service points.
[0068] Based on the optimized decision results, an optimized kill chain is generated, and a second-order delay simulation is performed on the optimized kill chain based on the service twin.
[0069] Specifically, the steps of the performance feedback mechanism include:
[0070] Based on the situation awareness set, situation features are extracted and standardized to generate a situation feature set.
[0071] Build a service composition template, including a template identification area, a service composition configuration area, and a performance benchmark area;
[0072] The service combination template is subjected to compliance verification, and an index table of situation characteristics-template ID is established to generate a self-evolving kill chain knowledge base;
[0073] Establish a real-time energy efficiency data acquisition mechanism to collect real-time energy efficiency data and deviation data, and calculate template quality scores;
[0074] Set a threshold for classification, and classify templates into levels based on the template quality score, including efficient templates and inefficient templates. Efficient templates are marked as priority for reuse, and inefficient templates are marked as to be eliminated.
[0075] Initiate self-evolution optimization, review and remove inefficient templates, and perform constraint optimization on efficient templates;
[0076] When faced with a new combat mission, extract the feature vector of the new combat mission, call the index table, filter out the candidate set of suitable templates, and select the optimal suitable template according to the comprehensive quality score.
[0077] The optimal adaptation template is instantiated based on the real-time status data of the service resource pool.
[0078] The beneficial effects of this invention are:
[0079] The distributed service resource pool standardizes combat entities into service points, and combined with dynamic monitoring, ensures that information and physical status are synchronized, providing a well-organized data foundation for link construction. The three-level situational view, through high-low layer decoupling and local updates, avoids global reconstruction caused by situational shifts, buffering the impact of frequent changes. Directed hypergraph modeling and multi-dimensional constraint embedding transform disordered combinations into ordered choices of finite functional categories. Combined with constraint propagation algorithms to prune ineffective combinations, it effectively compresses the exponentially growing state space, mitigating state space catastrophe. Through targeted invitations and rating screening of management nodes, combined with local hot replacement of failed service points, it reduces the number of links. The system addresses invalid computations and cold start frequency in path construction; it uses forward-looking extrapolation of service twins to optimize latency risks in advance, ensuring that the link's real-time performance meets standards; it binds situational characteristics and optimizes the link through a self-evolving kill chain knowledge base; and it avoids repeated full-process computation for new tasks through efficient template reuse and rapid instantiation, fundamentally reducing trial-and-error latency and cold start accumulation. It curbs state space catastrophes through structured modeling and constraint pruning, and reduces cold starts caused by frequent situational switching through local updates, hot replacements, and template reuse. The two work together to reduce the accumulation of solution latency, ensuring that the kill chain generation speed meets the real-time requirements of actual combat. Attached Figure Description
[0080] Figure 1 Flowchart of a service-based kill chain analysis construction method;
[0081] Figure 2 This is a flowchart illustrating the process of identifying situational change events in this invention;
[0082] Figure 3 This is a flowchart of constructing a three-level situation view in this invention;
[0083] Figure 4 This is a flowchart of the process for generating the dimensionality-reduced state space in this invention. Detailed Implementation
[0084] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.
[0085] refer to Figures 1 to 4 As shown in the figure, this embodiment introduces a service-based kill chain analysis construction method, including:
[0086] Step S1: Construct a distributed service resource pool, converting various entities in the combat system into service points with standardized capability interfaces. Each service point stores capability descriptions, real-time status parameters, and service quality attributes. Simultaneously, a service discovery mechanism is configured in the service resource pool. Distributed monitoring is used to perceive the dynamic changes of all service points in real time, including dynamic access, real-time status reporting, and deregistration. This ensures that the service point information in the service resource pool remains synchronized with the physical equipment status and records the service point association logic formed during the historical kill chain construction process, reserving a data foundation for link relationship identification. Here, entities are related equipment in the combat system, such as reconnaissance equipment, decision-making equipment, and strike equipment. Capability descriptions are the core performance parameters of the entity, such as the detection range of reconnaissance equipment and the range of strike equipment. Real-time status parameters are the dynamic operational data of the entity, such as current location, available payload, operational status, and network connection quality. Service quality attributes include response latency, mission success rate, and resource consumption rate.
[0087] Step S2: Based on the real-time status data of the service resource pool, target data and initial combat tasks in the actual combat scenario are integrated. Initial link relationships are constructed through matching logic to generate a fused situation set. Situation correlation analysis is used to identify situation switching events that affect the construction of the kill chain, and the affected service points and corresponding link relationships are screened to locally update the fused situation set, thereby generating a locally updated situation awareness set. A three-level situation view is constructed, including the target layer, task layer, and service layer, to achieve decoupling of high-level combat tasks and underlying service capabilities. When the target or situation changes, the affected demand range is first mapped through the task layer, and then the corresponding service points are reorganized to avoid global cascading reconstruction caused by changes in underlying details. This effectively buffers the impact of frequent situation switching on kill chain generation and provides accurate demand boundaries and data support for subsequent service combinations.
[0088] Step S3: Based on the initial combat mission, requirements are refined to generate the combat flow. The combat flow is modeled as a directed hyperedge, with the vertices of the directed hyperedge representing service points of different functional types in the three-level situational view, such as reconnaissance service points, decision-making service points, and strike service points. The direction represents the temporal relationship of the combat flow. At the same time, functional category constraints are defined to transform the disordered arrangement and combination of service points in the service resource pool into an ordered selection within a finite number of functional categories. Combining the directed hyperedge and functional category constraints, a structured directed hypergraph is constructed, and multi-dimensional constraints are built to be integrated into the directed hypergraph to transform the kill chain construction into a dynamic constraint satisfaction problem. Invalid combinations are pruned through the constraint propagation algorithm to generate a reduced-dimensional state space. Simultaneously, service points that meet the functional category constraints and domain knowledge constraints are selected from the service resource pool to form the initial combat candidate set. Among them, the functional type is determined based on the entity category, such as the service points corresponding to reconnaissance equipment being reconnaissance service points and the service points corresponding to strike equipment being strike service points.
[0089] Step S4: Based on the reduced state space, configure management nodes at the task layer of the three-level situational view. The management nodes act as task inviters, sending task invitation information to service points in the initial combat candidate set using sub-tasks decomposed from the task layer. The capability compatibility between each service point and sub-task is calculated based on the service resource pool. Response plans are fed back within a preset response time limit, including capability compatibility, execution cost, and estimated completion time. A rating function for the management nodes is constructed based on failure weights and capability compatibility to adapt and filter response plans, determine the optimal service point combination, and generate the initial kill chain, including the optimal service point combination and its corresponding link sequence. Failure verification is performed on the initial kill chain. If a service point failure is detected, local re-bidding is conducted based on the directed superedge where the failed service point is located, and hot replacement of the failed service point is performed. The management node is the core of sub-task coordination and scheduling, dynamically assigned by the task layer according to the sub-task type. For example, the management node corresponding to a reconnaissance sub-task is a service point with reconnaissance task scheduling capabilities, responsible for sub-task allocation and service point adaptation. Sub-tasks are generated by decomposing the initial combat task and functional category constraints.
[0090] Step S5: Based on the initial kill chain, construct a service twin that is synchronously mapped with physical combat and perform forward calculations to simulate the execution process of the initial kill chain. The calculation is delayed to perform elimination and reorganization or delay optimization. Elimination and reorganization is used to eliminate the initial kill chain and redetermine the optimal combination of service points, while delay optimization is used to replace risky service points with candidate service points to generate an optimized kill chain and ensure that the real-time performance of the kill chain meets the requirements.
[0091] Step S6: Extract situational features from the situational awareness database, bind the optimized kill chain with situational features and the energy efficiency data of the service twin, generate service combination templates, store them in the self-evolving kill chain knowledge base, and establish an efficiency feedback mechanism. Based on real-time energy efficiency data, score the service combination templates, classify them into high-efficiency and low-efficiency templates, eliminate low-efficiency templates, and optimize the parameter configuration of high-efficiency templates. When facing a new combat mission, call the situational features of each service combination template in the kill chain knowledge base, perform feature matching to select suitable service combination templates, and quickly instantiate them by combining the real-time status data of the current service resource pool, so as to achieve near-instantaneous kill chain generation, reduce redundant calculations and trial-and-error delays, and fundamentally improve the efficiency of kill chain construction under frequent situational switching.
[0092] Specifically, the steps for building a service resource pool include:
[0093] Obtain a full-domain equipment list for the combat system, classify it based on functional dimensions such as reconnaissance, decision-making, and strike, and save the basic data of each equipment's entire life cycle to the full-domain equipment list, such as factory performance parameters, combat deployment records, historical fault data, and maintenance cycle information. At the same time, import core parameters from the preset tactical rule base, including equipment coordination priority, mission adaptation threshold, and resource allocation limit.
[0094] Based on the equipment classification results, each equipment entity is converted into an independent service point, and an interface architecture is set for each service point, including a capability description interface, a status interaction interface, and a service quality interface. Linear normalization is used to map performance parameters to the same dimension to achieve capability comparability across equipment types. Among them, the capability description interface is used to obtain the core combat capability parameters of the service point, the status interaction interface is used to receive real-time status parameters and perform unified format conversion, and the service quality interface generates standardized scores based on the quantitative indicators and calculation rules of service quality attributes.
[0095] Based on a distributed redundancy design of master and slave nodes, the command center is the master node, and slave nodes are deployed in the combat area according to geographical partitions. Each slave node is configured with a cache library to cache access data of service points within its jurisdiction, while the master node stores relevant data of service points across the entire domain. Both the master and slave nodes are servers, serving as hardware server carriers for the distributed architecture, and are used to manage service points within their respective geographical areas.
[0096] The configuration includes a dual-mode service discovery mechanism with active reporting and passive detection. Active reporting involves automatically sending a registration request to the slave node after the service point starts up, carrying its own interface description and initial status data. After successful registration, the status update is reported according to a preset period to ensure the real-time status data. Passive detection involves the slave node periodically sending detection commands to service points within its jurisdiction that have not reported their status. If no response is received multiple times, the service point is marked as disconnected and synchronized to the master node to avoid invalid service points occupying resources.
[0097] For each service point, the operating status coefficient, task success rate, and resource redundancy rate are obtained. Through weighted calculation, the health score of the service point is calculated, which serves as the service point's waiting priority. Among them, the operating status coefficient is derived based on real-time status parameters, which correspond to normal operation, degraded operation, and operation failure. Different statuses correspond to different operating status coefficients. For example, the operating status coefficient corresponding to normal operation is 1. The task success rate is the ratio of the number of historical successes to the total number of executions. The resource redundancy rate is the ratio of the remaining resources to the rated resources. The weight coefficients are set by the tactical rule base.
[0098] Based on historical combat data from each service point, supplementary service point association logic is added, such as combat scenario type, combination of participating service points, collaborative timing relationship, task execution result, and constraint satisfaction status. Based on the storage structure of scenario classification and timing index, the association logic is classified according to combat scenario and indexed by task execution timestamp in each scenario. Combined with service point interface data, health score, and association logic, a service resource pool is constructed.
[0099] Specifically, the steps for identifying situational change events include:
[0100] Using real-time data from the service resource pool as the core layer, target data and environmental interference data collected by sensors in the combat system as the perception layer, and the initial combat mission as the constraint layer, a three-dimensional fusion framework is constructed to clarify the fusion priority and format conversion rules of data at each level, ensuring the orderly integration of multi-source data. Among them, target data includes target type, trajectory, priority, and threat level, and the initial combat mission includes combat objectives, time windows, and damage requirements.
[0101] Based on the association logic of each service point in the service resource pool, the matching degree calculation is used to filter out the historical association logic that matches the current initial mission and target type of the operation, so as to perform historical matching and reduce the cost of building from scratch during cold start. Combined with the real-time data in the service resource pool, the historical association logic is adapted and adjusted, such as replacing the failed or unqualified service points and supplementing the currently available high-quality service points, thereby generating multiple candidate links that conform to the current resource status. Each candidate link includes the service point collaborative combination and time sequence relationship.
[0102] For each candidate link, the initial fit of the candidate link is calculated by weighted calculation based on the compatibility between service point capabilities and task constraints, the average health score of all service points in the link, and the similarity between historical links and the current scenario. The candidate link with the maximum initial fit is taken as the initial link. The compatibility between service point capabilities and task constraints is derived from the degree of matching between the service point capability description and the task requirements. For example, if the reconnaissance range covers the target area, it is set to 1; if it partially covers, it is between 0 and 1; and if it does not cover, it is set to 0. The similarity between historical links and the current scenario is obtained by using the cosine similarity algorithm through vector comparison of combat scenario features, target type, and task constraints.
[0103] Based on the three-dimensional fusion framework and initial link, a fusion situation set is generated, and the format consistency and logical rationality are checked to remove contradictory data and invalid information, thereby ensuring the reliability of the fusion data.
[0104] The configuration includes a dual-mode event recognition mechanism of real-time monitoring and periodic verification, and identifies situational change events based on a fused situational awareness set. Real-time monitoring monitors dynamic data within the fused situational awareness set; if the change in monitored data exceeds a preset monitoring threshold (e.g., a service point's operational status changes from normal to faulty, or a target trajectory deviation exceeds the reconnaissance range), immediate recognition is triggered. Periodic verification performs consistency checks on static data at preset intervals; if static changes occur, such as mission adjustments or service point capability degradation, delayed recognition is triggered. Dynamic data includes real-time service point status, target trajectory, and environmental interference intensity; static data includes mission constraints and service point core capabilities; and the preset interval is dynamically adjusted based on the operational rhythm.
[0105] Configure situational classification standards, including impact scope, urgency, and event type. Simultaneously, set judgment thresholds and trigger conditions for various situational transition events. Based on the identified situational transition events, obtain the impact scope coefficient, task urgency, and resource correlation. Through weighted calculation, calculate the event urgency of the situational transition event. Combined with the situational classification standards, generate an event list of current situational transition events, including event type, impact scope, urgency, trigger time, and associated data identifiers. The impact scope includes global, regional, and local; the urgency includes level one, level two, and level three; the event type includes service-side, target-side, task-side, and environment-side; the impact scope coefficient is quantified based on the impact scope (e.g., global = 1, regional = 0.6, local = 0.2); the task urgency is the ratio of remaining time to total task time; the resource correlation is determined by whether the event involves core service points / critical targets (e.g., 1 for those involving, 0.3 for those not involving); and the weighting coefficient is dynamically calculated based on the combat scenarios in the tactical rule base.
[0106] Based on the urgency of the event list, a hierarchical triggering mechanism is established to match differentiated update strategies. Level 1 events directly trigger the highest priority update, Level 2 events trigger regional-level updates, and Level 3 events are only recorded without actively triggering updates, thus avoiding computational delays caused by excessive updates and generating triggering instructions for current situation switching events.
[0107] The evaluation indicators are depth of association, functional importance, resource dependence, time sensitivity, and redundancy. Quantification rules for each indicator are defined, and a link impact evaluation system is constructed through indicator weights. Since the initial link is only used for priority execution, other candidate links are still potential resources and may become effective links after changes in the situation. Candidate links that have not undergone adaptation screening are used as evaluation objects, and the link impact of each candidate link is calculated. Among them, the indicator weights are set by the tactical rule base according to the type of combat mission. Depth of association includes direct association, indirect association, and no association. Functional importance includes supporting core missions, supporting secondary missions, and no mission support. Resource dependence includes involving scarce service points and involving ordinary service points. Time sensitivity is quantified based on time windows, and redundancy is quantified by the number of alternative links.
[0108] Based on the priority of combat missions, an impact threshold is set, and links and associated service points with an impact level higher than the impact threshold are screened out to form an affected mapping table, clarifying the key service points, collaborative relationships and causes of impact for each affected link;
[0109] Configure local update rules to update target data in the affected mapping table without affecting unaffected service points and links. Only update the data change fields, do not repeatedly update unchanged information, only reconstruct the affected collaboration logic, and do not change the global data structure.
[0110] An incremental update engine is built, including: extracting real-time status data of affected service points and collaborative relationship data of related links from the fusion situation set; adjusting the service point combination or collaborative timing of links based on the cause of the event, such as service point failure or target trajectory deviation; performing logical consistency verification between the updated local data and the unupdated data to ensure that the capabilities of the newly replaced service points meet the task constraints; merging the updated local data with the original fusion situation set to generate a partially updated situation awareness set; and establishing an update log recording mechanism to record key information of each update, providing a basis for subsequent performance tracking and template optimization, narrowing the update scope, and avoiding cold start delays.
[0111] Specifically, the steps for constructing a three-level situation view include:
[0112] Based on the target data collected from situational awareness, the core attributes of combat targets are identified, including type, location, threat level, and constraints, including damage requirements and time windows. A three-dimensional description structure of target, identifier, and constraints is adopted to construct the target layer.
[0113] Based on the constraints of the initial mission and target layer, and based on mission decomposition, sub-task definition, and priority ranking, the initial mission is broken down into independent sub-task units, such as reconnaissance and positioning, fire decision-making, and precision strike. The functional requirements, execution sequence, and constraint thresholds of each sub-task are clarified, thereby constructing the mission layer.
[0114] Based on the functional requirements of sub-tasks, a set of suitable service points is selected from the situational awareness set, a two-way mapping relationship between sub-tasks and service points is established, the adaptation priority of each service point is clarified, and the functional category labels of service points, such as reconnaissance, decision-making, and strike, are marked to construct a service layer and finally generate a three-level situational view; among them, the adaptation priority is obtained based on the health score and service quality attribute ranking.
[0115] Obtain the priority ranking of subtasks in the task layer, including core subtasks, important subtasks, and general subtasks. Different priorities correspond to different quantitative values, thereby generating priority coefficients for subtasks. Based on the service point capability descriptions of the service layer and the functional requirements of subtasks in the task layer, obtain the adaptability between service points and subtasks, including fully satisfied, partially satisfied, and unsatisfied. Different satisfaction levels correspond to different adaptability quantification values. Using the Jaccard similarity coefficient, calculate the service point collaboration compatibility based on the historical collaboration records of service point association logic. By weighting the priority coefficient, adaptability, and collaboration compatibility, calculate the hierarchical linkage adaptability, quantitatively evaluate the degree of collaboration adaptability of goals, tasks, and services after hierarchical adjustment, and avoid logical conflicts or insufficient adaptability caused by hierarchical adjustment.
[0116] A hierarchical linkage mechanism is established, including upward linkage, downward linkage, and cross-layer linkage. Upward linkage triggers adaptation adjustments only for the corresponding subtasks when the status of a service point changes at the service layer. This is done by replacing the service point with a candidate service point from the bidirectional mapping relationship. The effectiveness of the linkage adjustment is determined by the hierarchical linkage adaptation degree and the adaptation threshold after replacement. If the linkage adjustment is ineffective, candidate service points are re-selected without affecting other subtasks or the target layer. Downward linkage is used to map to the corresponding subtask through the task layer when the constraints at the target layer change, and then triggers service point reorganization at the service layer. The service point combination with the highest hierarchical linkage adaptation degree and meeting the adaptation threshold is the final service point reorganization solution. If none of the combinations meet the requirements, feedback is sent to the task layer to readjust the subtask requirements or back to the service layer to expand the selection range. Cross-layer linkage is used to feed back to the target layer for constraint reassessment when the status of a core subtask changes, until the hierarchical linkage adaptation degree of the selected service point combination meets the adaptation threshold, ensuring a dynamic balance between target achievement and resource adaptation.
[0117] Specifically, the steps for generating the dimensionality-reduced state space include:
[0118] Based on a three-level situational awareness view, a three-layer node system of a directed hypergraph is defined to achieve accurate mapping with the three-level situational awareness view. This system includes target layer nodes, task layer nodes, and service layer nodes. Each layer of nodes encapsulates the core attributes and identifiers of its corresponding level, generating a set of directed hyperedge nodes. Specifically, target layer nodes are centered on combat targets, encapsulating target attributes and constraints, with each target corresponding to a unique node identifier. Task layer nodes are based on decomposed sub-tasks, encapsulating sub-task attributes, with node identifiers corresponding one-to-one with sub-task IDs. Service layer nodes are based on service points, encapsulating core information of service points and reusing the unique identifier of each service point.
[0119] Design internal collaborative hyperedges and cross-layer adaptable hyperedges to characterize the relationships within and between layers, generating a set of directed hyperedges. Internal collaborative hyperedges are used to describe the relationship logic within the same layer, such as the temporal dependency relationship between reconnaissance subtasks and decision subtasks in the task layer. The direction of the hyperedge reflects the relationship sequence or priority. Cross-layer adaptable hyperedges are used to describe the mapping relationship between different layers, such as the requirement mapping between target layer nodes and task layer nodes. The hyperedge weight is a quantified value of the relationship adaptation.
[0120] Multidimensional constraints are set to convert the graph into quantitative constraints for directed hypergraphs, including functional category constraints, health constraints, temporal constraints, and resource redundancy constraints. Among them, functional category constraints are based on equipment function classification and limit the adaptation boundary between service layer nodes and task layer nodes. For example, reconnaissance service points can only adapt to reconnaissance sub-task nodes. Health constraints are based on the health score and health threshold of service points to select candidate service points. Temporal constraints are based on the execution sequence of sub-tasks and limit the direction and association order of collaborative hyperedges within the task layer. For example, the preceding sub-task node must point to the following sub-task node. Resource redundancy constraints are based on redundancy indicators to prioritize hyperedges without alternative service points, providing a basis for subsequent pruning.
[0121] For each hyperedge in the directed hyperedge set, the functional category matching degree, health compliance degree, timing compliance degree, and resource redundancy adaptability degree are obtained through the quantitative standards corresponding to the multi-dimensional constraints. The constraint satisfaction degree of each hyperedge is calculated by weighted summation using the weights set by the tactical rule base. If a hyperedge violates any constraint, the constraint satisfaction degree is directly recorded as 0 and marked as an invalid hyperedge. Among them, the functional category matching degree is quantified based on whether the functions in the functional category constraints match, including complete matching, partial matching, and non-matching. Different matching states correspond to different quantitative values of functional category matching degree. The health compliance degree is quantified based on whether the health constraint is met. The health compliance degree is 1 when the threshold condition is met and 0 when it is not met. The timing compliance degree is quantified based on whether the timing constraint is met. The timing compliance degree is 1 when the timing requirement is met and 0 when it is not met. The resource redundancy adaptability degree is quantified based on the number of alternative links.
[0122] Based on the product of constraint satisfaction and hyperedge weight, the final weight of the hyperedge is calculated. Hyperedges with a final weight lower than the set constraint threshold are temporarily stored as hyperedges to be pruned. Thus, the directed hypergraph after constraint embedding is constrained to include a set of directed hyperedge nodes, a set of hyperedges with final weights, and marked as valid / pruned states.
[0123] Based on the affected mapping table and situational switching events, the affected regions in the directed hypergraph are located. A three-level pruning priority rule is set, and the affected regions and their associated hyperedges are pruned according to priority. The hypergraph structure of unaffected regions remains unchanged, achieving incremental processing. At the same time, the consistency of the pruned directed hypergraph is checked to ensure that key nodes still have suitable associated paths. If a logical break occurs, key redundant hyperedges are backtracked and retained to avoid affecting subsequent service point combinations. The affected regions are determined by the affected service layer nodes, associated task layer nodes, and corresponding hyperedges. The three-level pruning priority rule includes first-level pruning, second-level pruning, and third-level pruning. First-level pruning directly removes invalid hyperedges and completely removes paths that violate the core rules. Second-level pruning deletes hyperedges to be pruned to reduce low-fitness paths. Third-level pruning is used to evaluate the path redundancy of the remaining hyperedges, uses the shortest path algorithm to identify duplicate paths, retains only the path with the highest weight, and removes redundant paths.
[0124] Based on the pruned directed hypergraph, the dimensionality-reduced state space information is extracted, and a state space description system is constructed, including path dimension, constraint dimension, and efficiency dimension. The path dimension records the complete path of target-task-service formed by all effective hyperedge combinations. The constraint dimension marks the constraint satisfaction and core constraint adaptation of each path. The efficiency dimension is associated with the average health and collaborative compatibility of the service points corresponding to each path, and finally generates the dimensionality-reduced state space.
[0125] Based on functional category constraints, service points that are consistent with the functional category of directed hyperedge vertices are selected from the service resource pool. For example, only reconnaissance service points are selected for hyperedge reconnaissance vertices, and only decision service points are selected for decision vertices, forming a preliminary candidate pool divided by functional category.
[0126] Substitute the service points in the initial candidate pool into the reduced-dimensional state space to verify whether the service points are adapted to the temporal relationship and stage constraints of the directed hyperedge. Only service points that meet the adaptation criteria are retained, thereby generating a verification candidate pool. Among them, the temporal relationship verification is used to confirm whether the response speed and data transmission delay of the service points meet the temporal requirements of the combat process, such as whether reconnaissance service points can provide target data before decision-making service points are started. The stage constraint verification is used to verify whether the service points can meet the stage constraints of the corresponding vertices of the hyperedge, such as whether strike service points can achieve the target damage accuracy requirements.
[0127] The service points in the verification candidate pool are hierarchically classified and organized according to the functional category vertices of the directed hyperedges and the operational process sequence. First, the service points are divided into subsets according to functional categories, such as reconnaissance candidate subsets and decision-making candidate subsets. The core information of the service points in each subset is marked, such as capability parameters, health score, collaborative compatibility level, and adaptive hyperedge position. Based on the three-dimensional mapping relationship of functional category, service point, and constraint satisfaction, a constraint mapping table is established to clarify the constraint compliance details of each service point, thus forming the final initial operational candidate set.
[0128] Specifically, the steps for generating the initial kill chain include:
[0129] Based on the task layer data of the three-level situation view, a dedicated management node is configured for each decomposed subtask. The management node corresponds one-to-one with the subtask, encapsulates the core requirements of the subtask and the associated directed hyperedge information, and reuses the bidirectional mapping relationship to clarify the service point invitation scope boundary of the management node, thereby avoiding aimless invitations and generating a set of management nodes.
[0130] Using management nodes as task inviters, standardized task invitation information is constructed, including subtask identifiers, functional requirement details, constraints, and response format requirements, to ensure that the task invitation information accurately matches the subtask requirements and multi-dimensional constraints.
[0131] Based on the initial combat candidate set, task invitations are distributed in a targeted manner through management nodes. Invitations are only sent to service points in the initial combat candidate set that match the function of the sub-task. For example, reconnaissance sub-task invitations are only sent to reconnaissance candidate service points to avoid distribution to irrelevant service points and to narrow the response range.
[0132] Response time limits are set based on the urgency of subtasks, requiring invited service points to provide response solutions within the time limit, including capability adaptability, execution cost, and estimated completion time. After the response time limit expires, all valid response solutions are collected through the management node to form a set of response solutions specific to the corresponding subtask. Capability adaptability is derived based on the matching between the service point's core capabilities and the subtask requirements. Execution cost is calculated by combining the service point's own operational losses and resource consumption costs. Estimated completion time is estimated based on the service point's own response speed, data processing efficiency, and current load status.
[0133] For each response scheme, based on the real-time capability data of service points in the service resource pool and the requirements of sub-tasks, the functional matching degree, performance compliance degree, and state compatibility degree are obtained. Through weighted calculation, the capability adaptability is calculated. Among them, the functional matching degree is quantified based on the degree to which the core functions of the service point cover the requirements of the sub-tasks, including complete coverage, partial coverage, and no coverage; the performance compliance degree is quantified based on the degree to which the performance indicators of the service point meet the threshold requirements of the sub-tasks, including complete compliance, partial compliance, and no compliance; and the state compatibility degree is quantified based on the stability of the real-time operating state of the service point, including no conflict, slight fluctuation, and abnormal state.
[0134] A comprehensive rating function is constructed for the management node. For the response plan corresponding to the management node, the comprehensive rating score of each response plan is calculated by weighted summation using failure weight, capability adaptability, execution cost, and estimated completion time as rating indicators. Among them, the failure weight is derived based on the service point health score. The higher the health score, the lower the failure weight.
[0135] For the response solution set, it is sorted in descending order based on the comprehensive rating score. Combined with the constraints in the dimensionality-reduced state space, the service point with the best comprehensive rating score is selected and bound to the corresponding subtask to generate the optimal service set. The optimal service sets of all subtasks are integrated and linked together based on the temporal relationship of the directed hypergraph to form a complete optimal service point combination. The collaboration order and data interaction path are clearly defined, taking into account adaptability, reliability and execution efficiency. At the same time, the temporal constraints after dimensionality reduction are anchored to ensure that the service point combination is both optimal and compliant, and to avoid invalid combinations from entering the subsequent process.
[0136] Based on the logic of optimal service point combination and combat process, a structured initial kill chain is generated, which clarifies the link's start conditions, service point coordination order, data transmission interface, and termination judgment criteria. Based on the real-time status data of the service resource pool called by the management node, the operating status, health score, and network connection status of each service point in the link are checked one by one to perform real-time failure verification. If all service points are valid, the initial kill chain is confirmed to be effective and the link information is solidified. If a failed service point is detected, a local re-tender is immediately carried out based on the directed hyperedge where the failed service point is located to achieve hot replacement of the failed service point.
[0137] Specifically, the steps for deducing the initial kill chain include:
[0138] Based on the initial kill chain, the optimal service point combination, collaborative temporal relationship, data interaction path and sub-task constraints are extracted as the core framework for building the service twin, ensuring that the service twin's structure is completely matched with the initial chain.
[0139] By accessing multi-source synchronized data, real-time mapping between physical warfare and service twins is achieved. This includes service point status data, battlefield environment data, and constraint rule data. Based on the three-layer mapping relationship of service points, links, and environment, a service twin is constructed, and a physical synchronization mechanism is configured. Based on a preset update cycle, the updated data of the service resource pool and the situational awareness center are synchronized to ensure that the status of the service twin is consistent with the physical warfare system in real time. Among them, service point status data includes the real-time operating status and performance parameters of the service resource pool; battlefield environment data includes environmental interference data and target dynamic data from the situational awareness center; constraint rule data includes functional category constraints and multi-dimensional constraints; and the service twin includes a service point twin layer, a link twin layer, and an environment twin layer. The service point twin layer replicates the physical characteristics, operating status, and capability boundaries of each service point and supports real-time synchronous updates of status parameters. The link twin layer restores the collaborative timing and data transmission logic of the initial kill chain and simulates the interaction process between service points. The environment twin layer maps the dynamic battlefield environment and simulates the impact of environmental factors on service point operation and data transmission.
[0140] Based on service twins, the execution flow of the initial kill chain is simulated. The twins of each service point are triggered sequentially according to the collaborative timing, recreating the entire process of subtask execution, data transmission, and collaborative interaction. The interference impact of the battlefield environment on each stage is simulated synchronously. Various latency quantities are obtained through end-to-end latency quantification standards, including the processing latency of service points, data transmission latency, collaborative waiting latency, and environmental interference latency. The latency of each stage is accumulated according to the execution flow to obtain the end-to-end total latency of the initial kill chain. Among them, the processing latency is derived from the real-time processing rate of the service point and the complexity of the subtask; the data transmission latency is derived from the transmission distance, bandwidth, and interference intensity; the collaborative waiting latency is derived from the timing constraints and the completion time of the preceding stage; and the environmental interference latency is derived from the environmental interference intensity and the anti-interference capability of the service point.
[0141] Based on the changing trends of the battlefield environment and the evolution of service point status, service twins are used to simulate latency changes in various potential scenarios, identify latency risk points, and generate multi-scenario latency simulations, including latency prediction values under different scenarios.
[0142] The system compares the total end-to-end latency with the real-time threshold in the tactical rule base to determine whether the real-time performance of the initial kill chain meets the standard. If the total end-to-end latency exceeds the real-time threshold, there is a risk of latency exceeding the limit, triggering an optimization decision. If the total end-to-end latency does not exceed the real-time threshold, there is no risk of latency exceeding the limit, and the initial kill chain is directly retained.
[0143] For optimization decisions, the predicted delay values in multi-scenario delay simulations are compared with real-time thresholds to determine whether predicted delays occur. If the number of predicted delays exceeds the preset simulation limit, it is determined that the overall link adaptation is insufficient, and elimination and reorganization are carried out. The initial kill chain is marked as an invalid link, and the configuration of the management node is fed back to trigger the re-selection of the optimal service point combination. Using the initial combat candidate set and the comprehensive rating function, the weight of the comprehensive rating function is adjusted based on the delay simulation results, such as increasing the weight of the estimated completion time, and the optimal service point combination is re-determined.
[0144] If the number of predicted delays does not exceed the preset simulation limit, it is determined that the local risk is controllable. Delay optimization is performed, the risk service points corresponding to the delay risk points are located, their functional categories, capability requirements, and constraints are extracted, and alternative service points that are consistent with the functional categories of the risk service points and meet the constraint requirements are selected from the initial operational candidate set. Based on capability adaptability and delay simulation data, alternative service points that meet the adaptability standards and have lower delays are selected as replacement targets.
[0145] Based on the optimization decision results, an optimized kill chain is generated. If elimination and reorganization are performed, the optimal service point combination that has been re-selected will be assembled into a new kill chain according to the temporal constraints and link construction rules of the directed hypergraph. If delay optimization is performed, the candidate service points will replace the risk service points in the initial kill chain, and the collaborative temporal sequence and data interaction path of the link will be updated synchronously to ensure the logical continuity of the link after replacement.
[0146] Based on the service twin, a second delay simulation is performed on the optimized kill chain to verify whether the end-to-end delay meets the real-time threshold requirements. After the second verification is passed, the core information of the optimized kill chain is solidified, including service point combination, coordination timing, data transmission interface, and replacement rules, forming a structured link solution that can be directly executed.
[0147] Specifically, the steps of the performance feedback mechanism include:
[0148] Based on the situational awareness set, situational features are extracted, including task features, target features, environmental features, and resource features. The extracted situational features are then standardized, including classification and coding, quantization and normalization, and feature filtering, to generate a standardized situational feature set. Among them, task features include the core attributes of the combat mission, such as mission type, target damage requirements, and time window constraints, reflecting the core requirements of the mission; target features include key target information, such as target type, threat level, and real-time trajectory characteristics, reflecting the differentiated attributes of the target; environmental features include battlefield environmental elements, such as jamming type, jamming intensity, and weather conditions, characterizing the impact of the environment on combat; and resource features include the overall status of the service resource pool, such as the distribution of service point functional categories, the availability status of core service points, and resource redundancy, reflecting resource supply capabilities.
[0149] Construct a structured service composition template to achieve deep binding of scenarios, links, and efficiency. This includes: a template identification area that records the template ID and generation time, and associates it with a standardized situation feature set as a scenario label; a service composition configuration area that replicates and optimizes the core information of the kill chain, such as service point combination, collaboration timing, and constraint parameters; and an efficiency benchmark area that records the energy efficiency data of the service twin as a quality benchmark.
[0150] The service composition templates are validated for compliance to ensure that they match the situation characteristics and that the service composition logic is complete. After the validation is passed, they are stored in the self-evolving kill chain knowledge base. An index table of situation characteristics and template IDs is established and stored according to four-dimensional characteristics to support fast retrieval, thereby generating the self-evolving kill chain knowledge base.
[0151] Establish a real-time energy efficiency data collection mechanism to collect real-time energy efficiency data of the actual execution of the optimized kill chain, as well as deviation data between service twin simulation and actual execution. Quantitatively score the data from four dimensions: execution efficiency, resource efficiency, stability, and adaptation accuracy. The template quality score is obtained by weighted summation according to the weights set in the tactical rule base.
[0152] Set a threshold for classification and classify templates into different levels based on their quality scores, including efficient templates and inefficient templates. Efficient templates are marked as priority for reuse, while inefficient templates are marked as to be eliminated. This generates a template quality rating report that clarifies the score composition and level identifier for each template.
[0153] Initiate self-evolution optimization to review inefficient templates and confirm that multiple executions are due to inefficiency rather than accidental factors. Remove the reviewed inefficient templates from the self-evolutionary kill chain knowledge base, update the index table, and delete the corresponding index entries. For the shortcomings of efficient templates, combine real-time performance feedback data and constraint rules to adjust the template's service point filtering parameters, collaborative timing parameters, or constraint thresholds. The optimized efficient templates are then re-stored in the kill chain knowledge base, the template version number and performance benchmark data are updated, and the historical iteration records of the original templates are retained to support backtracking queries.
[0154] When faced with a new combat mission, the situational characteristics of the new combat mission are extracted and a standardized feature vector is generated. The index table of the kill chain knowledge base is called, and the cosine similarity algorithm is used to calculate the matching degree between the new mission features and all efficient template features. The candidate set of suitable templates is screened out, and the optimal suitable template is selected according to the comprehensive quality score.
[0155] Based on real-time status data from the service resource pool, the optimal adaptation template is quickly instantiated, replacing failed or busy service points, fine-tuning the collaboration timing and data interaction path to ensure adaptation to the current resource reality, and verifying whether the instantiated service combination meets the constraints of the new task and collaboration compatibility. If the verification passes, a new task kill chain that can be executed directly is generated. If it fails, the next template in the adaptation template candidate set is selected and the process is repeated to avoid repeated calculations and trial and error, achieving near-instantaneous link generation and fundamentally improving the efficiency of kill chain construction under frequent situational changes.
[0156] Working principle and effects:
[0157] Combat entities are standardized as service points. Distributed monitoring synchronizes their status, and a three-level situational awareness view is used to achieve decoupling between high and low layers and local updates. This buffers the impact of situational transitions and avoids global reconstruction. The combat process is modeled with a directed hypergraph, embedding multi-dimensional constraints and pruning invalid combinations through algorithms to curb the exponential growth of the state space. Through targeted invitations and rating screening of management nodes, coupled with local hot replacement of failed service points, invalid calculations are reduced. Service twins are used to proactively predict and optimize latency, ensuring the real-time performance of the links. Optimized links and situational features are bound as templates and stored in a self-evolving kill chain knowledge base. New tasks are quickly instantiated through feature matching to avoid redundant calculations. Structured modeling and constraint pruning mitigate state space catastrophes, and local updates, hot replacements, and template reuse reduce cold start accumulation, collaboratively reducing solution latency and meeting the real-time requirements of actual combat.
[0158] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.
Claims
1. A service-based kill chain analysis construction method, characterized in that, include: Convert various entities in the combat system into service points, build a service resource pool, and configure a service discovery mechanism in the service resource pool to update the service resource pool; Based on the real-time status data of the service resource pool, target data and initial combat tasks in the actual combat scenario are integrated, and a fused situation set is generated through matching logic. Situational correlation analysis is used to identify situational switching events, so as to locally update the fused situation set and generate a situational awareness set. Based on the situation awareness set, a three-level situation view is constructed, including the target layer, the task layer, and the service layer. When the target or situation changes, the scope of the affected demand is mapped through the task layer, and then the reorganization of the corresponding service points is triggered. Based on the initial combat mission, requirements are extracted to generate the combat process. The combat process is modeled as a directed hyperedge, and the vertices of the directed hyperedge are service points of different functional types in the three-level situation view. Combined with functional category constraints, a directed hypergraph is generated, and multi-dimensional constraints are constructed to be integrated into the directed hypergraph. Invalid combinations are pruned through the constraint propagation algorithm to generate a dimensionality-reduced state space, and an initial combat candidate set is selected from the service resource pool. Based on the reduced state space, the task layer is configured with management nodes based on the three-level situation view. The management nodes act as task inviters, and task invitation information is sent to the initial combat candidate set in combination with sub-tasks. Response schemes are generated and adapted and filtered, an initial kill chain is generated, and failure verification is performed. A service twin is constructed to deduce the initial kill chain, and the execution delay of the computing end is used for elimination and reorganization or delay optimization. Situational features are extracted from situational awareness data. The optimized kill chain is bound to situational features and the energy efficiency data of the service twin to generate service combination templates, which are then stored in the kill chain knowledge base. An efficiency feedback mechanism is established to score the service combination templates and classify them into high-efficiency and low-efficiency templates. When facing a new combat mission, the situational features of each service combination template in the kill chain knowledge base are called up, and feature matching is performed to select a suitable service combination template. The templates are then quickly instantiated by combining the real-time status data of the current service resource pool.
2. The service-based kill chain analysis construction method according to claim 1, characterized in that, The steps to build a service resource pool include: Obtain a comprehensive list of equipment across the entire combat system and categorize it. Each equipment entity is converted into a service point, and an interface architecture is set up for each service point. Configure master and slave nodes, and set up service discovery mechanisms, including active reporting and passive detection; For each service point, obtain the operational status coefficient, task success rate, and resource redundancy rate, and calculate the health score of the service point. Based on historical combat data, supplement the association logic of service points to build a service resource pool.
3. The service-based kill chain analysis construction method according to claim 2, characterized in that, The steps to identify situational change events include: A three-dimensional fusion framework is constructed, with real-time data from the service resource pool as the core layer, target data and environmental interference data collected by sensors in the combat system as the perception layer, and the initial combat mission as the constraint layer. Based on the association logic of each service point in the service resource pool, historical matching and real-time adaptation are performed to generate candidate links; For each candidate link, calculate the initial fitness score of the candidate link, and take the candidate link with the maximum initial fitness score as the initial link. Based on the aforementioned 3D fusion framework and the initial link, a fusion situation set is generated, and format consistency and logical rationality are verified. The configuration includes a dual-mode event recognition mechanism that combines real-time monitoring and periodic verification, and identifies situational switching events based on the fused situation set.
4. The service-based kill chain analysis construction method according to claim 3, characterized in that, The steps for identifying situational change events also include: Based on the situational change event, the impact range coefficient, task urgency, and resource correlation are obtained, and the event urgency of the situational change event is calculated through weighted calculation. Based on the configured situation classification criteria, an event list of the situation switching events is generated; Based on the urgency of the events in the event list, a tiered triggering mechanism is established, and differentiated update strategies are matched to generate triggering instructions for the situational switching events. Using association depth, functional importance, resource dependence, time sensitivity, and redundancy as evaluation indicators, and defining quantitative rules for each indicator, a link impact evaluation system is constructed. Using candidate links as the evaluation object, calculate the link impact, set an impact threshold, and filter out links and associated service points whose impact exceeds the impact threshold to form an affected mapping table; Configure local update rules and build an incremental update engine to generate a situational awareness set through data extraction, logic reconstruction, consistency verification, and integrated storage.
5. The service-based kill chain analysis construction method according to claim 4, characterized in that, The steps to construct a three-level situation view include: Based on the aforementioned situational awareness set, a target layer is constructed using a structure of target, identifier, and constraint. A task layer is constructed based on task decomposition, subtask definition, and priority sorting. The service point set is selected from the situational awareness set, and a service layer is constructed through the bidirectional mapping relationship between subtasks and service points to generate a three-level situational view; Obtain the priority coefficient of the subtask, the adaptability between the service point and the subtask, and the service point collaboration compatibility. Calculate the hierarchical linkage adaptability through weighted calculation. Set up a hierarchical linkage mechanism, including upward linkage, downward linkage, and cross-level linkage; After the linkage adjustment, the hierarchical linkage adaptability of the new service point set is calculated, and the linkage adjustment is effectively determined by the adaptability threshold.
6. The service-based kill chain analysis construction method according to claim 5, characterized in that, The steps to generate a reduced-dimensional state space include: Based on the aforementioned three-level situational view, a three-layer node system for the directed hypergraph is defined, including target layer nodes, task layer nodes, and service layer nodes, and a set of directed hyperedge nodes is generated. Design internal collaborative hyperedges and cross-layer adaptive hyperedges to generate a set of directed hyperedges; Set multi-dimensional constraints, including functional category constraints, health constraints, timing constraints, and resource redundancy constraints; For each hyperedge in the directed hyperedge set, the functional category matching degree, health compliance degree, timing compliance degree, and resource redundancy adaptability are obtained through the quantitative standard of multi-dimensional constraints. The constraint satisfaction degree of each hyperedge is calculated, and invalid hyperedges are screened out. The final weight of the hyperedge is calculated based on the product of the constraint satisfaction and the hyperedge weight, so as to select the hyperedges to be pruned and the directed hypergraph after constraint embedding.
7. The service-based kill chain analysis construction method according to claim 6, characterized in that, The steps for generating a reduced-dimensional state space also include: Based on the affected mapping table and the situational switching event, the affected region in the directed hypergraph is located, a three-level pruning priority rule is set, and the affected region and its associated hyperedges are pruned in priority order, and consistency verification is performed. Based on the pruned directed hypergraph, a state space description system is constructed, including path dimension, constraint dimension, and efficiency dimension, to generate a dimensionality-reduced state space. Based on functional category constraints, a preliminary candidate pool is selected from the service resource pool; Substitute the service points in the preliminary candidate pool into the reduced state space to check whether the service points are adapted to the temporal relationship and link constraints of the directed superedge, and generate a verification candidate pool. The service points in the verification candidate pool are sorted and categorized hierarchically to generate an initial combat candidate set.
8. The service-based kill chain analysis construction method according to claim 7, characterized in that, The steps to generate the initial kill chain include: Based on the task layer of the three-level situational view, management nodes are configured for subtasks, and a set of management nodes is generated by combining the bidirectional mapping relationship. Using the management node as the task inviter, task invitation information is constructed; Based on the initial set of combat candidates, mission invitations are distributed in a targeted manner through management nodes; Set response time limits, collect valid response solutions, and generate a set of response solutions for the subtask; Acquire functional matching degree, performance compliance degree, state compatibility degree, and computing power adaptability; The comprehensive rating score for each response plan is calculated using failure weight, capability adaptability, execution cost, and estimated completion time as rating indicators. The response scheme set is sorted in descending order, and the optimal service point is selected by combining the reduced state space. The optimal service points of all subtasks are integrated, and the optimal service point combination is generated based on the directed hypergraph. Based on the logic of optimal service point combination and combat process, an initial kill chain is generated, and real-time failure verification is performed based on the service resource pool.
9. The service-based kill chain analysis construction method according to claim 8, characterized in that, The steps for deducing the initial kill chain include: Extract the optimal service point combination, collaborative timing relationship, data interaction path and sub-task constraints, and access multi-source synchronized data to build a service twin; Based on the service twin, the execution flow of the initial kill chain is simulated to obtain various latency parameters and the end-to-end total latency of the initial kill chain is obtained. By using service-oriented twins to simulate latency changes in various scenarios, multi-scenario latency simulations are generated. If the total end-to-end latency exceeds the real-time threshold, an optimization decision is triggered; otherwise, the initial kill chain is retained. For optimization decisions, the predicted delay value in the multi-scenario delay simulation is compared with the real-time threshold to determine whether a predicted delay occurs. If the number of predicted delays exceeds the simulation limit, elimination and reorganization will be carried out, triggering a re-selection of the optimal service point combination; otherwise, delay optimization will be performed to select alternative service points. Based on the optimized decision results, an optimized kill chain is generated, and a second-order delay simulation is performed on the optimized kill chain based on the service twin.
10. The service-based kill chain analysis construction method according to claim 9, characterized in that, The steps of a performance feedback mechanism include: Based on the situation awareness set, situation features are extracted and standardized to generate a situation feature set. Build a service composition template, including a template identification area, a service composition configuration area, and a performance benchmark area; The service combination template is subjected to compliance verification, and an index table of situation characteristics-template ID is established to generate a self-evolving kill chain knowledge base; Establish a real-time energy efficiency data acquisition mechanism to collect real-time energy efficiency data and deviation data, and calculate template quality scores; Set a threshold for classification, and classify templates into levels based on the template quality score, including efficient templates and inefficient templates. Efficient templates are marked as priority for reuse, and inefficient templates are marked as to be eliminated. Initiate self-evolution optimization, review and remove inefficient templates, and perform constraint optimization on efficient templates; When faced with a new combat mission, extract the feature vector of the new combat mission, call the index table, filter out the candidate set of suitable templates, and select the optimal suitable template according to the comprehensive quality score. The optimal adaptation template is instantiated based on the real-time status data of the service resource pool.
Citation Information
Patent Citations
Construction method of killing chain generation model
CN119249878A
Method and system for optimizing weapon equipment system based on killing chain
CN113255118A
Intelligent network awareness method for optimizing efficiency through step-by-step concurrent caching
CN114124503A
