CN116151492B - Auxiliary decision analysis system of combined combat system - Google Patents

The auxiliary decision analysis system of the combined combat system comprises a data acquisition module, a decision analysis module and a decision analysis module, wherein the data acquisition module is used for acquiring combat data and generating a knowledge graph according to the combat data; the scheme generating module is connected with the data acquisition module and used for intelligently generating an initial combat scheme according to the knowledge graph and the multi-objective optimization model; the scheme optimizing module is connected with the scheme generating module and is used for optimizing and adjusting the initial combat scheme according to the real-time battlefield situation to obtain the target combat scheme. The intelligent means such as knowledge maps, reinforcement learning and the like are scientifically applied to the association and generation of the combat scheme, a combat knowledge base is constructed, and the battlefield situation is accurately analyzed based on a knowledge reasoning technology; the battlefield scheme optimizing model is established, and the action tasks are reasonably planned by utilizing the technologies such as reinforcement learning and the like, so that the efficiency and effect of the battlefield scheme can be greatly improved.

Auxiliary decision analysis system of combined combat system

Technical Field

The invention belongs to the field of combat command, and particularly relates to an auxiliary decision analysis system of a combined combat system.

Background

The combined combat is a multi-element integrated large-system combat which is carried out in a multi-dimensional combat space, participates in multi-element combat forces and adopts various combat laws in a plurality of combat stages. In the combined combat, the participating forces are related to each other, are mutually dependent and are mutually synergistic, so that the functions of 'whole is larger than the sum of parts' show amplifying effect. The typical feature of the combined combat is that the combat effort is composed of two or more armies, with unified combined army commander and combined partnership, and unified combined combat plan for achieving the established common goal, combat jointly implemented under the unified plan.

With the development of science and technology, informatization becomes a key factor for determining war victory and defeat, and high efficiency and automation of an information system are important directions for the development of military informatization level. And directly influencing the effect of the combined combat according to the combat scheme formulated by the combat information. The number of elements, such as equipment, personnel and the like, involved in modern war is very complex, and the relationship among the elements is very complex, so that the node scale and the complexity grow exponentially in a combat system, and great challenges are brought to the efficient and high-quality generation of a combat scheme. Currently, big data and artificial intelligence technology are rapidly developed, and the development of an information-based society is effectively promoted. How to utilize the deep learning method to analyze big data information, and then assist the combined combat to formulate the combat scheme has important meaning.

Disclosure of Invention

In order to achieve the above object, the present invention provides the following solutions: a joint combat architecture auxiliary decision analysis system, comprising:

the data acquisition module is used for acquiring combat data and generating a knowledge graph according to the combat data;

the scheme generating module is connected with the data acquisition module and used for constructing a multi-objective optimization model and intelligently generating an initial combat scheme according to the knowledge graph and the multi-objective optimization model; and

And the scheme optimizing module is connected with the scheme generating module and is used for optimizing and adjusting the initial combat scheme according to the real-time battlefield situation to obtain a target combat scheme.

Preferably, the data acquisition module comprises a data acquisition unit, a data fusion unit, a data evaluation unit and a data updating unit;

the data acquisition unit is used for acquiring combat data of different data types, wherein the combat data comprises combat task information, combat target information, target force information, target equipment information and target action information of structured data types; combat mission information, combat target information, target force information, target equipment information and target action information of the semi-structured data type; combat mission information, combat target information, target force information, target equipment information, and target action information of unstructured data types;

the data fusion unit is used for carrying out knowledge extraction and knowledge cleaning on the acquired combat data with different data types and then carrying out data fusion so as to obtain target combat data with clear association;

the data evaluation unit is used for presetting a confidence threshold value, carrying out reliability quantitative evaluation on the target combat data, and eliminating data with the reliability smaller than the confidence threshold value;

the data updating unit is used for storing the target combat data subjected to the credibility quantitative evaluation into a knowledge graph so as to generate the knowledge graph or update the knowledge graph in real time.

Preferably, the knowledge extraction by the data fusion unit includes entity extraction, relationship extraction and semantic attribute extraction;

the entity extraction is used for constructing entity dictionary, knowledge rule or knowledge feature from the structured data and the semi-structured data, and extracting entity conforming to a certain class;

the relational extraction is used to mine knowledge from unstructured data;

the semantic attribute extraction is used for adding a space-time description method to the resource description framework, constructing an enhanced resource description framework triplet suitable for space-time indexing and query, and extracting the semantic attribute based on the enhanced resource description framework triplet.

Preferably, the knowledge cleansing by the data fusion unit includes entity association, entity disambiguation, and entity alignment;

the entity association is used for establishing a correlation relationship between the soldier entities;

the entity disambiguation is used for solving the ambiguity problem generated by the entity with the same name;

the entity alignment is used for mining and analyzing the same entity for each entity in the multi-source heterogeneous information;

the credibility quantitative evaluation performed by the data evaluation unit is evaluated based on the accuracy and coverage rate of knowledge;

the accuracy is used for evaluating the accuracy of the knowledge, and the coverage rate is used for evaluating the coverage degree of the knowledge for a certain field.

Preferably, the scheme generating module comprises a model building unit, a data processing unit and a scheme generating unit;

the model construction unit is used for constructing the multi-objective optimization model and the reinforcement learning model;

the data processing unit is used for carrying out target combination sequencing on the target combat data in the knowledge graph to obtain a sequencing result;

the scheme generating unit is used for carrying out multi-objective overall optimal processing on the sequencing results based on the multi-objective optimization model, and generating the initial combat scheme according to the optimal processing results.

Preferably, the data processing unit comprises a target classifying unit, a target analyzing unit and a target sorting unit;

the target classifying unit is used for classifying the hit targets in the target combat data after summarizing to obtain a classifying result;

the target analysis unit is used for determining target combinations after single-target value analysis and multi-target value analysis are carried out according to the classification result;

the target ranking unit is used for ranking targets according to a combat plan or battlefield situation, comprehensively considering the task correlation degree, situation influence degree, fire threat degree, target defense degree and attack difficulty degree to provide a target hitting initial list, and carrying out existing capability assessment on the target combination by combining weather, topography, existing weapons, reaction time and attack limiting factors to determine the ranking result of target hitting.

Preferably, the classification result is classified into a fixed target and a movable target according to the target maneuvering characteristics, into a soft target, a light guard target, a heavy guard target and a firm target according to the target guard characteristics, and into a point target, a line target and a plane target according to the target distribution characteristics.

Preferably, the scheme optimizing module comprises a scheme optimizing unit and a scheme evaluating unit;

the scheme optimizing unit is used for carrying out multi-round optimization of a multi-objective function on the initial combat scheme based on the reinforcement learning model, and analyzing an optimizing result into combat schemes of a combat layer and a tactical layer through a scheme generating engine;

the scheme evaluation unit is used for comprehensively evaluating the scheme quality of the combat schemes of the battle layer and the tactical layer based on the analytic hierarchy process to obtain the target combat scheme.

Preferably, the scheme optimizing unit further comprises a target clustering unit, a fire distributing unit and a bullet matching unit;

the target clustering unit is used for classifying the combat targets and simplifying task allocation;

the fire power distribution unit is used for distributing task clusters to fire power alliances, and in each fire power alliance, fire points in the fire power alliances are interacted with target tasks through the reinforcement learning model to obtain an optimal strategy of target task distribution;

and the bullet matching unit is used for performing bullet matching on the combat target according to the optimal strategy.

Preferably, the scheme evaluation unit includes a first evaluation unit, a second evaluation unit, and a third evaluation unit;

the first evaluation unit is used for constructing an evaluation framework according to the combat influencing factors; the combat influencing factors comprise a battlefield environment, a hitting task list and combat intents; wherein the battlefield environment comprises a battlefield situation and a natural environment; the battlefield situation comprises a fire resource deployment situation and a potential target individual situation; the hitting task list comprises hitting tasks and task sequences; the striking task comprises a time attribute, a fire resource and a striking object; the time attributes include execution time and duration; the fire resources comprise battlefield positions, resource types and resource quantity; the hit objects comprise combat capability and combat effect;

the second evaluation unit is used for inputting the combat schemes of the combat layer and the tactical layer into an evaluation framework to perform index cleaning, normalization processing, index weight extraction, parameter matrix judgment and scheme sorting, so as to obtain a scheme sorting result;

the third evaluation unit is used for performing scheme evaluation according to the scheme sorting result to obtain a scheme evaluation result, judging whether the scheme evaluation result is ideal or not, executing a scheme if the scheme evaluation result is ideal, and storing the scheme into a knowledge graph to facilitate subsequent calling; if not, returning the scheme to the scheme optimizing unit to continue optimizing until the scheme evaluation result is ideal, and outputting to obtain the target combat scheme.

Compared with the prior art, the invention has the following advantages and technical effects:

the invention takes the knowledge graph as the overall knowledge source to provide the knowledge support of the bottom layer for the scheme execution; and (3) taking multi-objective overall optimization of the battle scheme as an optimization target, adopting a reinforcement learning method to perform multi-round optimization on the multi-objective function, analyzing the optimization result into battle schemes of a battle layer and a tactical layer through a scheme generation engine, and completing association and intelligent generation of the battle schemes. The intelligent means such as knowledge maps, reinforcement learning and the like are scientifically applied to the association and generation of the combat scheme, a combat knowledge base is constructed, and the battlefield situation is accurately analyzed based on the knowledge reasoning technology; the battlefield scheme optimizing model is established, and the action tasks are reasonably planned by utilizing the technologies such as reinforcement learning and the like, so that the efficiency and effect of the battlefield scheme can be greatly improved.

Drawings

The accompanying drawings, which are included to provide a further understanding of the application, illustrate and explain the application and are not to be construed as limiting the application. In the drawings:

FIG. 1 is a schematic diagram of a system architecture according to an embodiment of the present invention;

FIG. 2 is a diagram of a construction flow chart of a combat plan knowledge graph according to an embodiment of the present invention;

FIG. 3 is a diagram of a multi-objective optimization model of a combat scheme according to an embodiment of the present invention;

FIG. 4 is a reinforcement learning based solution optimization diagram of an embodiment of the present invention;

FIG. 5 is a diagram of a battle plan evaluation criterion according to an embodiment of the present invention;

fig. 6 is a diagram of a battle plan evaluation method based on an analytic hierarchy process in an embodiment of the present invention.

Detailed Description

It should be noted that, in the case of no conflict, the embodiments and features in the embodiments may be combined with each other. The present application will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.

It should be noted that the steps illustrated in the flowcharts of the figures may be performed in a computer system such as a set of computer executable instructions, and that although a logical order is illustrated in the flowcharts, in some cases the steps illustrated or described may be performed in an order other than that illustrated herein.

As shown in fig. 1, the auxiliary decision analysis system of the combined combat system provided by the invention comprises,

the data acquisition module is used for acquiring combat data and generating a knowledge graph according to the combat data;

the scheme generating module is connected with the data acquisition module and used for constructing a multi-objective optimization model and intelligently generating an initial combat scheme according to the knowledge graph and the multi-objective optimization model; and

And the scheme optimizing module is connected with the scheme generating module and is used for optimizing and adjusting the initial combat scheme according to the real-time battlefield situation to obtain a target combat scheme.

Furthermore, the system comprises three layers, namely a data acquisition layer, a scheme optimization layer and a scheme generation layer, wherein each layer has different functions, and the association and intelligent generation of the combat plan are completed together through the interaction of data, information and knowledge.

The knowledge map layer is mainly responsible for constructing a knowledge base for supporting the generation of a combat plan, and has the main functions of knowledge acquisition, knowledge fusion, knowledge evaluation, knowledge reasoning and the like. The generation of a combat plan as a decision-making activity with a very high knowledge density, requires a high-quality knowledge base to be constructed on the one hand and to have the ability to dynamically update on the other hand. Therefore, the knowledge graph layer of the system not only needs to construct a high-quality knowledge graph and provides knowledge support for target sequencing, model construction and reinforcement learning optimization of the scheme optimization layer, but also keeps contact with the scheme generation layer, and the generated and evaluated reasonable scheme is stored into the knowledge graph as knowledge, so that the knowledge graph can be updated.

The knowledge graph of the system can be regarded as a normalized and engineered knowledge base, and the construction and the use of the knowledge graph follow a certain engineering rule. Specifically, knowledge units are required to be extracted from various combat data such as tasks, targets, forces, equipment, actions and the like to form the bottommost knowledge resources; then, knowledge fusion is carried out on knowledge extracted from the multi-source data, the problems of repeated knowledge and ambiguous association between knowledge are solved, and a combat plan knowledge base is constructed through knowledge fusion and entity linking; and finally, carrying out quantitative evaluation on the reliability of the knowledge after fusion, discarding the knowledge with lower confidence, and guaranteeing the quality of a knowledge base, wherein the specific flow is shown in figure 2.

Specifically, the data acquisition module may include a data acquisition unit, a data fusion unit, a data evaluation unit, and a data update unit.

In knowledge sources of knowledge combat plan knowledge graphs, structured data, semi-structured data, and unstructured data can be classified according to the expression form of the data. Wherein the structured data is generally derived from the content of the structured expression, such as tables, normalized text, etc. Semi-structured data is derived from text or the like that has a certain structure, but is not fully structured. Unstructured data is derived from a variety of other sources. Namely, the data acquisition unit is used for acquiring combat data with different data types, wherein the combat data comprises combat task information, combat target information, target force information, target equipment information and target action information with structured data types; combat mission information, combat target information, target force information, target equipment information and target action information of the semi-structured data type; combat mission information, combat target information, target force information, target equipment information, target action information of unstructured data types.

The data fusion unit is used for carrying out knowledge extraction and knowledge cleaning on the acquired combat data with different data types and then carrying out data fusion so as to obtain target combat data with clear association.

The data evaluation unit is used for presetting a confidence threshold value, carrying out reliability quantitative evaluation on the target combat data, and eliminating the data with the reliability smaller than the confidence threshold value.

The data updating unit is used for storing the target combat data subjected to the credibility quantitative evaluation into a knowledge graph so as to generate the knowledge graph or update the knowledge graph in real time.

The combat plan knowledge can be expressed in the form of triples, i.e. "Subject", predicate (precursor), object "using a resource description framework (resource description framework, RDF), forming a large scale directed graph consisting of" point-edges ". The main body is a combat element and a combat entity, the predicate represents a method or technology for knowledge reasoning or information mining of the combat element and the combat entity, and the object is knowledge with attribute or resource description formed after reasoning. The standard RDF triples are not easy to express space-time information, and the space-time construction effect of the combat planning knowledge can be affected. To this end, the present system adds spatiotemporal description methods, such as temporal relationship predicates, spatial type statements, spatiotemporal relationship predicates, etc., to the RDF model to construct enhanced RDF triples suitable for spatiotemporal indexing and querying for semantic attribute extraction. Specifically, the knowledge extraction performed by the data fusion unit includes entity extraction, relationship extraction and semantic attribute extraction. Entity extraction is mainly to construct entity dictionary, knowledge rule or knowledge feature from structured and semi-structured data and extract entity conforming to a certain class. The knowledge rule of entity extraction is more definite, so that the extraction accuracy is higher, and the knowledge rule can also be applied to actual engineering. The relationship extraction then mines knowledge from unstructured data. The semantic attribute extraction is used for adding a space-time description method to the resource description framework, constructing an enhanced resource description framework triplet suitable for space-time indexing and query, and extracting the semantic attribute based on the enhanced resource description framework triplet.

The knowledge cleaning of the combat plan knowledge graph mainly comprises three parts, namely entity association, entity disambiguation and entity alignment. The entity association establishes a correlation relationship between the soldier entities. Entity alignment, also referred to as entity matching, refers to mining and analyzing the same entity for each entity in the multi-source heterogeneous information of the soldier. The entity disambiguation is specially used for solving the problem that the same-name entity generates ambiguity, and aims at solving the problem that a certain entity name in actual soldier information corresponds to a plurality of named entity objects, so that unification of different meanings of the same entity name in different scenes is completed.

The knowledge quality in the constructed knowledge graph is not completely the same, wherein some low-quality knowledge has little meaning on the generation of the fight plan and even affects the quality of the final fight plan, so the quality of the knowledge graph is evaluated. The evaluation process mainly comprises the steps of constructing a test set, providing a reference for knowledge evaluation, and then evaluating from the two angles of accuracy and coverage rate of the knowledge. The accuracy rate mainly evaluates the accuracy of the knowledge itself, and the coverage rate is used for evaluating the coverage degree of the knowledge to a certain field. Knowledge through knowledge evaluation can be applied to the generation of the combat plan. That is, the reliability quantitative evaluation is based on the accuracy and coverage of knowledge. The accuracy is used for evaluating the accuracy of the knowledge, and the coverage rate is used for evaluating the coverage degree of the knowledge for a certain field.

Further, the main objective of the system scheme generation layer is to generate an initial combat scheme, and main technologies include objective combination ordering, a multi-objective optimization model, a reinforcement learning model, a scheme generation engine and the like. Because battlefield situations change instantaneously, battle targets are various and have different values, the battle targets need to be combined and ordered, and high-value targets are selected from the combined and ordered targets to strike. After determining the hit targets, comprehensively considering the fire resources of the users, the matching degree of the fire resources and the hit targets, the fire resource application time and other factors, and constructing a multi-target optimization model of the fight scheme. Comprehensively considering factors such as optimization difficulty, time requirement and the like, adopting a deep reinforcement learning model to perform multi-objective optimization on an initial combat scheme, and improving the effect of multi-objective optimization by utilizing the advantage of artificial intelligence in decision aspect. And finally, sending the optimized result to a scheme generation layer through a scheme generation engine.

Specifically, the scheme generation module comprises a model construction unit, a data processing unit and a scheme generation unit;

the model construction unit is used for constructing the multi-objective optimization model and the reinforcement learning model;

the data processing unit is used for carrying out target combination sequencing on the target combat data in the knowledge graph to obtain a sequencing result; the scheme generating unit is used for carrying out multi-objective overall optimal processing on the sequencing results based on the multi-objective optimization model, and generating the initial combat scheme according to the optimal processing results.

In modern war, the number of first-chosen hit targets is correspondingly increased due to the numerous troops participating in army of military arms, analyzing the comprehensive value of the multi-objective combination under the current situation, and further predicting that the optimal target combination of the fire striking becomes important and difficult work in the combat planning process. Analyzing the comprehensive value of the targets under the current situation, realizing the value comparison and sorting among the targets based on the comprehensive value, and selecting the targets with the top ranking as fire striking objects. In the fire fight, the problems of damage effect to the target, ammunition consumption and guarantee difficulty of the my and the like are comprehensively considered, so that a multi-target optimization model is established to optimize the overall scheme of the fight, as shown in fig. 3.

Because the first hit targets are numerous in modern war and the targets have different values, the hit targets need to be ranked first, and a high-value target/target combination is selected from the targets to serve as the first hit target, so that a better hit effect is achieved. From a different perspective, the targets may be classified. Objects are classified by their maneuver characteristics, and may be classified into fixed objects and movable objects. Objects can be classified by object protection characteristics into soft objects, light objects, heavy objects, and firm objects. The targets can be classified into point targets, line targets and plane targets according to the target distribution characteristics.

The process of target combination ordering can be described as: after the fight scheme staff obtains the enemy situation, the fight scheme staff carries out target nomination according to the fight plan or battlefield situation, and then comprehensively considers the task correlation degree, situation influence degree, firepower threat degree, target defense degree and attack difficulty degree to provide a target hit initial list, and carries out the prior capability assessment by combining factors such as weather, topography, prior weapon, reaction time, attack limit and the like to determine the target hit list.

Specifically, the data processing unit comprises a target classifying unit, a target analyzing unit and a target sorting unit; the target classifying unit is used for classifying the hit targets in the target combat data after summarizing to obtain a classifying result; the target analysis unit is used for determining target combinations after single-target value analysis and multi-target value analysis are carried out according to the classification result; the target ranking unit is used for ranking targets according to a combat plan or battlefield situation, comprehensively considering the task correlation degree, situation influence degree, fire threat degree, target defense degree and attack difficulty degree to provide a target hitting initial list, and carrying out existing capability assessment on the target combination by combining weather, topography, existing weapons, reaction time and attack limiting factors to determine the ranking result of target hitting.

In constructing the multi-objective optimization model, the hit objective is first determined based on the above analysisIs marked as

In the above, the common objectAnd each.

The threat level of the target to the my is affected by factors such as distance, target type and the like, so the threat level is different from target to target. Giving a set of threat levels：

There is a mapping relationship between threat values and hit targets in the equation.

In addition to target sets, there are also My fire arrays in battlefield environments：

The above shows a fire striking device.

The devices in the my air defense array have different striking effects on the target by different devices due to the influence of different distances from the target, different types of the devices and other factors. Probability matrices can be usedRepresenting the probability of damage to the target by different devices:

in the middle ofIndicate use +.>Device pair->Probability of damage to the target being hit.

By usingIndicate use +.>Device pair->State of target striking:

then the matrixThe hitting condition of the current scene my device on the target can be represented:

the my devices can strike, i.e. matrix, at most one target at a timeThe value of each row summed by row satisfies the following condition:

arbitrary objectIs strapped by a plurality of devices>The survival probability after striking is:

the fire distribution of the whole threat target to the my air defense array is carried out through the fire striking of the my air defense array, so that the threat degree of the threat target of the enemy is minimum, namely:

setting each time of fireAmmunition consumed by strikingThe method comprises the following steps:

in the middle ofRepresents->The amount of ammunition consumed in the secondary hit.

The difficulty level of the guarantee is an index which is difficult to quantify, and the fuzzy number is used for the convenience of calculationIt is expressed as:

in summary, the multi-objective optimization function for the combat plan can be written as:

further, the main objective of the scheme optimizing layer of the system is to optimize and adjust the initial combat scheme according to the real-time battlefield situation to obtain the target combat scheme, and the scheme optimizing layer comprises a scheme optimizing unit and a scheme evaluating unit. The scheme optimizing unit is used for performing multi-round optimization of the multi-objective function based on the optimal processing result generated by the scheme generating unit of the reinforcement learning model scheme generating module, and analyzing the optimized result into a combat scheme of a battle layer and a tactical layer through the scheme generating engine. The scheme evaluation unit is used for comprehensively evaluating the scheme quality of the combat schemes of the battle layer and the tactical layer based on the analytic hierarchy process to obtain a target combat scheme. It should be noted that the evaluation of the protocol here is multi-tiered. If the scheme evaluation does not pass, returning the scheme to the scheme optimization layer for optimization; if the scheme evaluation is passed, the scheme is issued to a combat unit for execution, and meanwhile, the scheme is stored in a knowledge graph layer and used as a new knowledge resource to assist the intelligent generation of the follow-up scheme.

Optimization of the battle plan is not a one-shot process, but rather requires comprehensive consideration of multiple factors and multiple iterations to complete. Considering that in the actual combat planning, there is no fixed template for reference, each step of the plan can only depend on the calculation under the current situation, so a scheme optimization method based on reinforcement learning is proposed, as shown in fig. 4.

The method can be specifically divided into the steps of target clustering, firepower distribution, bullet mesh matching and the like. The purpose of object clustering is to group objects that are close to each other into a group, which is a key and fundamental step in simplifying the task allocation problem. The purpose of fire distribution is to distribute M clusters to M fire alliances, and the optimal strategy of target task distribution of the stage is obtained through a reinforcement learning algorithm, namely, each fire is distributed with one task cluster. In each fire power alliance, the fire power points in the fire power alliance are interacted with the target tasks through a reinforcement learning algorithm, and the optimal strategy that the fire power alliance completes the target tasks is obtained.

Meanwhile, the scheme optimizing unit further comprises a target clustering unit, a firepower distributing unit and a bullet matching unit. The target clustering unit is used for classifying the combat targets and simplifying task allocation. The fire distribution unit is used for distributing the task clusters to fire alliances, and in each fire alliance, the fire points in the fire alliances are interacted with the target tasks through a reinforcement learning algorithm to obtain the optimal strategy of target task distribution. And the bullet matching unit is used for matching the bullets of the combat target according to the optimal strategy.

The core idea of the strengthening algorithm is-network: an actor network and a reviewer network. The network calculates actions to be performed based on the acquired states, and the network evaluates the actions calculated by the behavioural network to improve the performance of the behavioural network. The experience replay buffer area is used for storing a certain amount of training experiences, and the network is read randomly when updating the network, so that the correlation in training data is broken, and the training is more stable. In the training stage, the network only acquires observation information from the network, and the network acquires information such as actions and observations of other subjects. In the execution phase, no network is involved, only one network is needed, which means that the execution is decentralized. The pseudo code of this algorithm is shown in table 1.

A reasonable and efficient combat action sequence is generated through reinforcement learning technology and is executed by a combat layer and a tactical layer. And combining battlefield situations and reasonable resource allocation requirements, and generating a combat plan intelligently in a layering manner. The tactical layer is used for weapon cooperative firepower striking, and mainly combines various levels of information sources, data and the like and commander intention and combat plan to form a single synthetic diagram. The battlefield target state is mastered in real time through relay detection and tracking of multiple sensors, cooperative network aiming and the like, and the cooperative firepower striking of multiple platforms is supported. The battle layer is used for battle command in battle direction, and based on battle layer situation, the combined battle field target data and environment data are added to form a common battle diagram by combining the intention of commander, battle plan and situation of the friend and neighbor battle area. The battlefield situation change is mastered in real time/quasi-real time through situation synchronization, multi-sensor combined detection, multi-sensor fusion target identification and the like, and formation combat command is supported.

In order to better serve the quality of the battle scheme, the intelligently generated battle scheme needs to be evaluated, and the quality of the battle scheme is determined according to the evaluation result. The operational scenario evaluation requires a number of factors to be considered, as shown in fig. 5. As can be seen from fig. 5, the factors involved in the evaluation of the combat plan are numerous, and the expression modes are different, such as numerical value type, character type, etc.; the numerical model has a score of larger, better, smaller, better, etc., so that a unified evaluation framework needs to be constructed to handle various criteria. For this, a battle plan evaluation method based on analytic hierarchy process was constructed as shown in FIG. 6. In the method, the related index cleaning, normalization processing, index weight extraction, parameter matrix judgment, scheme ordering and other processes can be completed by only summarizing the quantities of the related indexes into a program, and finally the ordering result of the scheme is obtained.

After the combat scheme evaluation result is obtained, if the result is not ideal, returning the result to a scheme optimization layer, and carrying out scheme optimization again; if the result is ideal, the scheme can be executed and stored in the knowledge graph, so that the subsequent calling is convenient.

The knowledge map layer, the scheme optimizing layer and the scheme generating layer of the system have layering property, interactivity and relevance, and jointly form a framework for the association and intelligent generation of the combat scheme. In the actual scheme generation process, the related technology is continuously optimized, and the efficiency and the effect of scheme management and intelligent generation are continuously improved.

The foregoing is merely a preferred embodiment of the present application, but the scope of the present application is not limited thereto, and any changes or substitutions easily conceivable by those skilled in the art within the technical scope of the present application should be covered in the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.