Air action plan generation method and system combined with intent reasoning of opposite side

Abstract

The invention discloses an air action plan generation method and system combined with intention reasoning of an opposite party, and the method comprises the steps: receiving a real-time confrontation situation by a threat perception and plan display assembly during the execution of cluster confrontation, evaluating the threat degree of a cluster target, and sending a threat alarm to an intention reasoning model; the intention reasoning model recognizes the possible intention type of the opposite side, updates the knowledge graph, queries the entity relationship in the knowledge graph, and obtains an alternative action plan of our side; the effectiveness evaluation model evaluates the confrontation effectiveness between each possible intention of the opposite party and the alternative action plan of the our party in a template matching mode; the game solving model enables various confrontation conditions and confrontation efficiency to form a game matrix, the game matrix is converted into a target optimization problem, and the optimal action plan and execution probability of our party are calculated. According to the method, the intention of the opposite party is inferred by analyzing the real-time confrontation situation, the air action plan of our party is generated, and the method has the advantages of high dynamic, high reliability and good real-time performance.

Description

Technical Field

[0001] The present application relates to the field of distributed collaborative technology, and in particular to a method and system for generating an air action plan in combination with reasoning about the other party's intention. Background Art

[0002] In recent years, the scenarios and forms of swarm confrontations have become increasingly diverse, placing higher demands on the quality and efficiency of air action plan generation. Air action plans provide strategic guidance and resource allocation during swarm confrontations. High-quality air action plans can create an advantage in swarm confrontations, continuously improving our effectiveness and becoming a key factor in determining the outcome of swarm confrontations. However, existing air action plan generation systems often overlook issues such as the uncertainty and rapid changes in the enemy's intentions, resulting in a lack of dynamic, real-time, and reliable air action plans. Summary of the Invention

[0003] In view of this, the present application provides a method and system for generating an air action plan that combines reasoning about the other party's intention.

[0004] This application discloses a method for generating an air action plan in combination with opponent intention reasoning, which includes: When executing swarm confrontation, the threat perception and planning display component receives real-time confrontation status, assesses the threat level of the swarm target, and issues threat alerts to the intent reasoning model. The real-time confrontation status includes the platform's motion status, the number of remaining devices, the target type, and the target's trajectory. Swarm confrontation is the use of distributed and collaborative intelligent clusters to conduct adversarial offense and defense. The intention inference model identifies the other party's possible intention type, updates the knowledge graph, queries the entity relationships in the knowledge graph, and obtains our alternative action plan. The other party's possible intention includes the number of clusters, cluster movement trends, cluster action types, and cluster threat level. The knowledge graph includes entities and entity relationships. The effectiveness evaluation model uses template matching to evaluate the effectiveness of each possible opponent's intention against our alternative action plan. Our alternative action plan includes our current number of clusters, cluster types, cluster real-time locations, cluster perception range, cluster attack range, and cluster confrontation resources consumed. The game-solving model transforms various confrontation situations and confrontation effectiveness into a game matrix, converts it into a target optimization problem, and calculates our optimal action plan and execution probability.

[0005] Furthermore, the assessing of the threat level of the cluster target includes: Cluster Target Depend on Single target composition, Threat level of a single target for:

[0006] Among them, and are the weight values of the target air superiority and the target countermeasure ability respectively, , ; The threat level of the cluster target is: When

[0007] When , notify the intention inference model to start running, is the threat level warning threshold.

[0008] Furthermore, the evaluation of the threat level of the cluster target includes: The threat level of a single target is related to the target air superiority; the target air superiority is related to the confrontation angle, confrontation distance, and speed of both sides. The target air superiority is expressed as:

[0009] Among them, , , are the weight values of the confrontation angle, confrontation distance, and speed respectively; is the course angle between the target and our side; is the maximum detection distance of our side, is the distance between the detected target and our side; is the maximum speed of the other side, is the speed difference between the target and our side; , , have the same value range, all greater than 0 and less than 1.

[0010] Furthermore, the evaluation of the threat level of the cluster target includes: The threat level of a single target is related to the target countermeasure ability; the target countermeasure ability is related to the maneuverability, strike ability, and detection ability of both platforms. The target countermeasure ability is expressed as:

[0011] Among them, is the maneuverability gap between both platforms, is the strike ability gap between both sides, is the platform detection capability gap, and its value varies with the platform types of both parties; is the weight value of the difference in mobility between the two sides, is the weight value of the difference in attack capability between the two sides, Detect capability gaps between the two platforms.

[0012] Furthermore, the intention inference model infers the other party's possible intention based on the target information; the intention inference model consists of a neural network module and a knowledge inference module; the neural network module consists of a long short-term memory network (LSTM), a fully connected network, and a softmax layer. Each LSTM layer has multiple hidden layers, and each fully connected layer has multiple hidden layers. The ReLU excitation function is used, and the softmax layer logically classifies and outputs the intention type; the intention types include coordinated attack, circling, avoidance, and formation flight; The neural network module collects historical track information of cluster targets, identifies the current cluster target's intention type, and simultaneously updates the relevant information of the threat target to the knowledge reasoning module; the relevant information includes the component platform, type, and intention category; The knowledge reasoning module includes a knowledge graph and a knowledge rule set. The entities in the knowledge graph are divided into platforms, intent types, confrontation strategy types, clusters, action plans, and targets; the entity relationships in the knowledge graph are divided into plan execution, cluster management, attack targets, sensor loading, equipment, and command and control; the knowledge rule set includes multiple rules, each of which describes the conditions for the establishment of an entity relationship. If the entity relationship is established, the corresponding entity relationship is queried from the knowledge graph based on the entity relationship to obtain our alternative action plan.

[0013] Furthermore, the effectiveness evaluation model pre-emptively deduces the confrontation process between various possible intentions of the other party and our action plan, and uses three types of indicators, namely cost, benefit, and loss, as effectiveness evaluation results to form a template library; the template library includes multiple templates, each template including the relative situation of the two parties, our action plan, the intention of the other party, and the effectiveness evaluation result; The effectiveness evaluation model matches the template library with the current confrontation situation, combines the evaluation effectiveness value of the template with the highest similarity with the similarity weight, and calculates the evaluation effectiveness of the current confrontation situation; the current confrontation situation includes the relative situation of the two sides, our alternative action plan, and the possible intentions of the other party.

[0014] Furthermore, let the current relative situation between the two sides be , our alternative action plan , the other party may intend , then the template in the template library that is most similar to the current adversarial situation The following conditions are met:

[0015]

[0016] Among them, is the total number of templates in the template library, is the relative situation in the th template, is the action plan of our side in the th template, is the intention of the other party in the th template, is the current relative situation between the two sides and the relative situation in the th template, is the difference weight value between the alternative action plan of our side and the action plan of our side in the th template, is the possible intention of the other party and the intention of the other party in the th template, is the degree of difference between the th template and the current confrontation situation. The smaller the degree of difference, the more similar the template is to the current confrontation situation.

[0017] Furthermore, the effectiveness evaluation model evaluates the evaluation effectiveness of the current confrontation situation, including: According to the degree of difference of the template with the highest similarity to the current confrontation situation , a similarity weight is normalized. Then, the cost of our side , the benefit of our side , the loss of our side in the current confrontation situation, and the overall evaluation effectiveness value are respectively:

[0018]

[0019]

[0020] Among them, is the confrontation cost pre-calculated in the template , is 's calculation weight; is the confrontation benefit pre-calculated in the template , is the calculation weight of ; is the pre - calculated adversarial loss in the template is the calculation weight of is the normalization function.

[0021] Furthermore, the game - solving model forms a game matrix by combining various adversarial situations and evaluation effectiveness; the left - most column of the game matrix constructed by the game - solving model is the alternative action plans of our side, the top - most row is the possible intentions of the other side, and each matrix value is the adversarial effectiveness evaluation value of the alternative action plan of our side and the possible intentions of the other side .

[0022] Furthermore, the game - solving model transforms the game matrix into an objective optimization problem, and calculates the optimal action plan and execution probability of our side, including: The game - solving model maximizes the difference between the gains and losses of our side as the optimization goal, and constructs an optimization function , and the optimization variable is , where is the th alternative action plan of our side, is the execution probability of the th alternative action plan. This optimization function is solved by the particle swarm optimization (PSO) algorithm to obtain the optimal action plan and execution probability of our side; the core idea of the game - solving model is to find an optimal combination of the alternative action plans and execution probabilities of our side. Our side confronts the other side according to this optimal combination, and the sum of all confrontation results is optimal. The objective optimization solution form of the game - solving model is as follows:

[0023]

[0024]

[0025]

[0026] Among them, is the maximum value function, represents the confrontation gain of the th alternative action plan of our side and the set of possible intentions of the other side , represents the confrontation gain of the th alternative action plan of our side and the set of possible intentions of the other side Minimized loss Indicates that when adopting the th alternative action plan of our side against the th possible intention of the other side the benefit of our side in the current confrontation situation the weight value of, and the loss of our side in the current confrontation situation the weight value of Indicates the execution probability of the th alternative action plan against the th possible intention of the other side the operation result of Indicates the th alternative action plan of our side against the th possible intention of the other side the benefit value of confrontation Indicates the th alternative action plan of our side against the th possible intention of the other side the loss value of confrontation is the set of possible intentions of the other side the number of possible intentions of the other side in the set is the cost of our side in the current confrontation situation

[0027] This application also discloses an air action plan generation system combined with the inference of the other side's intention, which realizes the air action plan generation method combined with the inference of the other side's intention as described above. It includes a threat perception and plan display component, an intention inference model, an effectiveness evaluation model, and a game solution model; When performing cluster confrontation, the threat perception and plan display component is used to receive the real-time confrontation situation and evaluate the threat degree of the cluster target; the real-time confrontation situation includes the movement state of our platform, the remaining number of devices, the target type, and the target track; The intention inference model is used to identify the possible intention types of the other side and initially screen out the alternative action plans of our side using knowledge inference; the possible intentions of the other side include the number of the other side's clusters, the cluster movement trend, the cluster action type, and the cluster threat degree; The effectiveness evaluation model is used to evaluate the confrontation effectiveness between each possible intention of the other side and the alternative action plans of our side in a template matching manner; the alternative action plans of our side include the current number of our clusters, the cluster types, the real-time positions of the clusters, the cluster perception ranges, the cluster attack ranges, and the resources consumed by cluster confrontation; The game-solving model is used to form a game matrix from various confrontation situations and confrontation effectiveness, transform it into an objective optimization problem, and calculate the optimal action plan and execution probability for our side.

[0028] Due to the adoption of the above technical solution, the present application has the following advantages: 1. The present application combines the dynamic game method with the generation of air action plans, constructs an air action plan generation system, and focuses our limited air resources on the opponent's core objectives by analyzing the threat level of the opponent's cluster; secondly, the system fully considers the confrontation effectiveness between various possible intentions of the opponent and our action plan options to ensure the optimal confrontation effectiveness for our side. The system has high dynamicity, real-time performance, and reliability.

[0029] 2. The target confrontation ability of the present application is related to the maneuverability, strike ability, and detection ability of both sides' platforms. When analyzing the threat level of a single target, the threat perception and plan display component reads the various ability values and weights of this type of platform from the database and normalizes them into the various abilities of the platform. The maneuverability of each type of platform in the database consists of ability factors such as the maximum speed, overload, turning angle, and range of the platform; the strike ability includes ability factors such as the maximum attack distance, damage range, strike accuracy, and flight speed of the equipment; the detection ability is composed of ability factors such as the detection range, false alarm rate, and frequency band range of the sensor.

[0030] 3. The intention inference model of the present application outputs the types of the opponent's intentions through a neural network module completed by offline training. When the neural network module is offline trained, its training data includes three types of features: the target historical track, the target platform type, and the target threat level. The target historical track consists of 30 historical moment waypoints sampled at intervals of 0.5 seconds, and each waypoint includes the longitude, latitude, altitude, heading angle, and speed of the target. This data set has a total of 8000 samples, which are divided into a training set and a test set in a ratio of 8:2. The time window of the long short-term memory network is set to 10s, and the number of training times is set to 200 times.

[0031] 4. The knowledge inference module of the intention inference model of the present application obtains multiple alternative action plans for our side for the target intention based on the real-time situation and expert knowledge, and multiple knowledge rule sets (Semantic Web Rule Language). In the knowledge inference module, expert knowledge and the real-time situation constitute a real-time updated knowledge graph, and the attributes and relationships of each entity in the knowledge graph are refreshed by parsing each frame of real-time situation data. The entities in this knowledge graph are divided into 6 categories: platforms, intention types, confrontation strategy types, clusters, action plans, and targets; the entity relationships can be divided into 7 categories: plan execution, strategy adoption, cluster management, strike object, sensor loading, equipment allocation, and command and control. The knowledge rule set (Semantic Web Rule Language) infers the alternative action plans for our side by querying the entity status in the knowledge graph.

[0032] 5. The effectiveness evaluation model of this application evaluates the confrontation effectiveness between the possible intentions of the other party and our alternative action plans in the way of template matching. A set of template libraries is constructed inside the effectiveness evaluation model, and various confrontation templates are stored in the template libraries. When generating each template offline, the confrontation conditions such as the initial confrontation situation, our action plan, and the intentions of the other party are randomized and the deduction is started. After the deduction ends, its confrontation effectiveness is calculated, and each confrontation condition and the corresponding confrontation effectiveness are stored in the template library as a template.

[0033] 6. After the game matrix of this application is transformed into an objective optimization problem, the optimal result is solved by the PSO optimization algorithm. When solving this problem with the PSO optimization algorithm, its optimization parameters are set as follows: the population size is 40, the number of iterations is 200, the particle search space is , and the range of particle velocity is limited to , the space dimension is 2, the particle inertia weight is 0.9, the self-learning factor is 0.5, and the group learning factor is 0.8.

[0034] 7. The air action plan generation system constructed using this application has the characteristics of high dynamicity, high real-time performance, and high reliability, and has a wide range of applications, and is highly applicable to task scenarios such as distributed collaborative tasks and cluster game confrontations.

[0035] 8. This application is applicable to fields such as distributed collaborative tasks, cluster game confrontations, and situation awareness, especially the field of distributed collaborative tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions in the embodiments of this application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments recorded in the embodiments of this application, and those of ordinary skill in the art can also obtain other drawings based on these drawings.

[0037] Figure 1 is the architecture diagram of the air action plan generation system combining opponent intention reasoning in the embodiments of this application; Figure 2 is the threat assessment schematic diagram of the threat perception and plan display component in the embodiments of this application; Figure 3 is the functional implementation schematic diagram of the intention reasoning model in the embodiments of this application; Figure 4 is the confrontation effectiveness evaluation schematic diagram of the effectiveness evaluation model in the embodiments of this application; Figure 5 is the optimization problem solving flow chart of the game solving model in the embodiments of this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0038] The present application will be further described in conjunction with the accompanying drawings and embodiments. The described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art shall fall within the scope of protection of the embodiments of the present application.

[0039] See Figure 1 , an embodiment of a method for generating an air operation plan by combining the inference of the other party's intention is provided in the present application, which includes: During the execution of cluster confrontation, the threat perception and plan display component receives the real-time confrontation situation, evaluates the threat level of the cluster target, and issues a threat warning to the intention inference model; the real-time confrontation situation includes the movement state of our platform, the remaining number of devices, the target type, and the target track; cluster confrontation is an act of using a distributed and collaborative intelligent agent cluster for adversarial offense and defense. The intention inference model identifies the possible intention types of the other party, updates the knowledge graph, queries the entity relationships in the knowledge graph, and obtains our alternative operation plans; the possible intentions of the other party include the number of the other party's clusters, the cluster movement trend, the cluster action type, and the cluster threat level; the knowledge graph includes entities and entity relationships. The effectiveness evaluation model evaluates the confrontation effectiveness between each possible intention of the other party and our alternative operation plans in a template matching manner; our alternative operation plans include our current number of clusters, cluster types, cluster real-time positions, cluster perception ranges, cluster attack ranges, and cluster confrontation resource consumption. The game-solving model forms a game matrix with various confrontation situations and confrontation effectiveness, transforms it into an objective optimization problem, and calculates our optimal operation plan and execution probability.

[0040] Optionally, the evaluation of the threat level of the cluster target includes: The cluster target is composed of single targets. The threat level of the th single target is:

[0041] Wherein, , are the weight values of the target air superiority and the target confrontation ability respectively, , ; The threat level of the cluster target is: is:

[0042] When the notification intent inference model starts running, is the threat level warning threshold.

[0043] Optionally, the threat level of the evaluated cluster target includes: The threat level of a single target is related to the target's air superiority; the target's air superiority is related to the confrontation angle, confrontation distance, and speed of both sides. The target's air superiority is expressed as:

[0044] Among them, , , are the weight values of the confrontation angle, confrontation distance, and speed respectively; is the course angle between the target and our side; is our maximum detection distance, is the distance between the detected target and our side; is the maximum speed of the other side, is the speed difference between the target and our side; , , have the same value range, all greater than 0 and less than 1.

[0045] Optionally, the threat level of the evaluated cluster target includes: The threat level of a single target is related to the target's confrontation ability; the target's confrontation ability is related to the maneuverability, strike ability, and detection ability of both platforms. The target's confrontation ability is expressed as:

[0046] Among them, is the maneuverability gap between both platforms, is the strike ability gap between both sides, is the platform detection ability gap, and its value varies with the types of both platforms; among them is the weight value of the maneuverability gap between both sides, is the weight value of the strike ability gap between both sides, is the platform detection ability gap between both sides. , , have the same value range, all greater than 0 and less than 1.

[0047] Optionally, the intention reasoning model infers the possible intentions of the other party based on the target information; the intention reasoning model consists of a neural network module and a knowledge reasoning module; the neural network module consists of a long short-term memory network (LSTM), a fully connected network, and a softmax layer. Each LSTM layer has multiple hidden layers, and each fully connected layer has multiple hidden layers. The ReLU activation function is used, and finally, the intention type is output through logical classification by the softmax layer. The intention types include coordinated attack, hovering, avoidance, and formation flight. The neural network module collects the historical track information of the cluster target, identifies the intention type of the current cluster target, and simultaneously updates the relevant information of the threat target to the knowledge reasoning module; the relevant information includes the constituent platform, type, and intention category. The knowledge reasoning module includes a knowledge graph and a knowledge rule set. The entities in the knowledge graph are divided into platforms, intention types, countermeasure types, clusters, action plans, and targets. The entity relationships in the knowledge graph are divided into plan execution, cluster management, strike object, equipped with sensors, equipped with devices, command and control. The knowledge rule set includes multiple rules, and each rule describes the establishment conditions of the entity relationship. If the entity relationship is established, the corresponding entity relationship is queried from the knowledge graph according to the entity relationship to obtain the alternative action plans for our side.

[0048] Optionally, the effectiveness evaluation model pre-offline deduces the confrontation processes of multiple possible intentions of the other party and our action plans in advance, takes the three types of indicators of cost, benefit, and loss as the effectiveness evaluation results, and forms a template library; the template library includes multiple templates, and each template includes the relative situation between the two sides, our action plan, the intention of the other party, and the effectiveness evaluation result. The effectiveness evaluation model matches the template library with the current confrontation situation, combines the evaluation effectiveness value of the template with the highest similarity and the similarity weight to calculate the evaluation effectiveness of the current confrontation situation; the current confrontation situation includes the relative situation between the two sides, our alternative action plans, and the possible intentions of the other party.

[0049] Optionally, let the current relative situation between the two sides be , our alternative action plan , the possible intention of the other party , then the template in the template library with the highest similarity to the current confrontation situation satisfies the following conditions:

[0050]

[0051] Among them, is the total number of templates in the template library, is the relative situation in the th template, is the Our action plan for a template For the opponent's intention for a template Is the current relative situation between the two parties And the relative situation in the nth template Is the difference weight value between our alternative action plan And the our action plan in the nth template Is the possible intention of the opponent And the opponent's intention in the nth template For the difference degree between the nth template and the current confrontation situation. The smaller the difference degree, the more similar the template is to the current confrontation situation.

[0052] Optionally, the effectiveness evaluation model evaluates the evaluation effectiveness of the current confrontation situation, including:[[]] According to the template with the highest similarity to the current confrontation situation The difference degree it has , normalized to generate a similarity weight , then the cost of our side in the current confrontation situation , the benefit of our side in the current confrontation situation , the loss of our side in the current confrontation situation , and the overall evaluation effectiveness value Are respectively:[[]]

[0053]

[0054]

[0055] Among them, Is the confrontation cost pre-calculated in the template , Is The calculation weight of; Is the confrontation benefit pre-calculated in the template , Is The calculation weight of; Is the confrontation loss pre-calculated in the template , Is The calculation weight of; Is the normalization function.

[0056] Optionally, the game-solving model forms a game matrix with various confrontation situations and evaluation effectiveness; the leftmost column of the game matrix constructed by the game-solving model is the alternative action plans of our side, the top row is the possible intentions of the other side, and each matrix value is the alternative action plan of our side and the possible intentions of the other side for the confrontation effectiveness evaluation value .

[0057] Optionally, the game-solving model transforms the game matrix into an objective optimization problem, and calculates the optimal action plan and execution probability of our side, including: The game-solving model maximizes the difference between the gains and losses of our side as the optimization objective, and constructs an optimization function , and the optimization variable is , where is the th alternative action plan of our side, is the execution probability of the th alternative action plan, and this optimization function is solved by the particle swarm PSO optimization algorithm to obtain the optimal action plan and execution probability of our side; the core idea of the game-solving model is to find an optimal combination of the alternative action plans and execution probabilities of our side. Our side confronts the other side according to this optimal combination, and the sum of the results of all confrontations is the best. The objective optimization solution form of the game-solving model is as follows:

[0058]

[0059]

[0060]

[0061] Among them, is the maximum value function, represents the th alternative action plan of our side and the confrontation gain with the set of possible intentions of the other side , represents the minimization of losses of the th alternative action plan of our side and the set of possible intentions of the other side , represents the gain of our side in the current confrontation situation when using the th alternative action plan of our side against the th possible intention of the other side ​ The weight value and the losses of our side in the current confrontation situation The weight value, Indicates the execution probability of the th alternative action plan And the execution probability of the th possible intention of the other party The operation result of (for example, ), Indicates the th alternative action plan of our side And the th possible intention of the other party The benefit value of the confrontation, Indicates the th alternative action plan of our side And the th possible intention of the other party The loss value of the confrontation, Is the set of possible intentions of the other party The number of possible intentions of the other party in the set of possible intentions of the other party, Is the cost of our side in the current confrontation situation. Among them, the set of possible intentions of the other party .

[0062] This application also provides an embodiment of an air action plan generation system that combines the inference of the other party's intentions, implementing the air action plan generation method that combines the inference of the other party's intentions described in the above embodiment. It includes a threat perception and plan display component, an intention inference model, an effectiveness evaluation model, and a game solution model; When performing cluster confrontation, the threat perception and plan display component is used to receive the real-time confrontation situation and evaluate the threat level of the cluster target; the real-time confrontation situation includes the movement state of our platform, the remaining number of devices, the target type, and the target track; The intention inference model is used to identify the possible intention types of the other party and initially screen out the alternative action plans of our side using knowledge inference; the possible intentions of the other party include the number of the other party's clusters, the cluster movement trend, the cluster action type, and the cluster threat level; The effectiveness evaluation model is used to evaluate the confrontation effectiveness between each possible intention of the other party and the alternative action plans of our side in a template matching manner; the alternative action plans of our side include the current number of our clusters, the cluster types, the real-time positions of the clusters, the cluster perception range, the cluster attack range, and the resources consumed by the cluster confrontation; The game solution model is used to form a game matrix with various confrontation situations and confrontation effectiveness, transform it into an objective optimization problem, and calculate the optimal action plan and execution probability of our side.

[0063] For the convenience of understanding, this application gives a more specific embodiment: Refer toFigure 1 , a system for generating an air operation plan by inferring the intentions of the other party according to the present application consists of a threat perception and plan display component, an intention inference model, an effectiveness evaluation model, and a game solution model. The threat perception and plan display component calculates the cluster threat level of the other party through the real-time confrontation situation and the inherent ability attributes of the platform, and issues a threat warning; the intention inference model identifies the possible intention types of the other party through an LSTM neural network, and updates the knowledge graph composed of expert knowledge in real time. The entity status in the knowledge graph is queried by a knowledge rule set (Semantic Web Rule Language) to obtain the alternative operation plans of our side; the effectiveness evaluation model matches the offline constructed template library with the current confrontation situation, combines the evaluation effectiveness value of the template with the highest similarity and the similarity weight, and calculates the evaluation effectiveness of the current confrontation situation; the game solution model constructs an objective optimization function with maximizing benefits and minimizing losses as the optimization goal, and uses the PSO (Particle Swarm Optimization) algorithm to solve the optimal operation plan and execution probability of our side.

[0064] Refer to Figure 2 , the operation process of the threat perception and plan display component according to the present application is as follows: when the threat perception and plan display component receives the real-time confrontation situation, it calculates the threat level of each single target from bottom to top. The air superiority of each can be calculated through the relative situation between each single target and our side; the target confrontation ability can be calculated according to the platform type and inherent ability attributes of the single target; the threat level of each single target is obtained through weighted calculation, and finally the average value of the threat levels of all platforms in a cluster is taken as the threat level of the entire cluster. When the threat level of the cluster is greater than the set threat threshold, this component issues a threat warning to notify other models to start processing.

[0065] Refer to Figure 3 , the function implementation process of the intention inference model according to the present application is as follows: when the intention inference model receives the threat target information, it collects the historical track information of the cluster target in the recent 10 seconds. The trained LSTM neural network identifies the intention category of the current cluster target, and at the same time synchronously updates the data such as the composition platform, type, and intention category of the threat target to the knowledge graph. The status of the corresponding entity in the knowledge graph is queried by the written knowledge rule set to obtain the alternative operation plans of our side.

[0066] Refer to Figure 4 , the confrontation effectiveness evaluation process of the effectiveness evaluation model according to the present application is as follows: when the effectiveness evaluation model receives the real-time situation, the alternative operation plans of our side, and the possible intentions of the other party, it combines the alternative operation plans of our side and the possible intentions of the other party respectively, and forms multiple confrontation situations in combination with the real-time situation. For each confrontation situation, the template with the highest similarity is found by traversing the offline constructed template library, the similarity degree between this template and the current confrontation situation is normalized to the similarity weight, and the effectiveness evaluation value of the current confrontation situation is calculated according to the effectiveness evaluation value of this template and the similarity weight.

[0067] See Figure 5 Figure 5 , the solution process for the optimization problem of the game-solving model of this application is as follows: When the game-solving model receives various confrontation situations and their effectiveness evaluation values, a game matrix is constructed and transformed into an objective optimization problem. The optimization objective is determined to be the maximum benefit and the minimum loss, and the optimization variables are the alternative action plans and execution probabilities of our side; When solving this problem, parameters such as the population size, dimension, number of iterations, weight, learning factor, etc. are initialized. In each iteration, the positions of the particle individuals and the population are updated, and the optimal fitness values of the individuals and the population are calculated. When the maximum number of iterations is reached, or the change in the fitness value between two times is less than the set threshold, the optimal action plan and action probability of our side are output.

[0068] Finally, it should be noted that: The above embodiments are only used to illustrate the technical solutions of this application and not to limit them. Although this application has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: It is still possible to modify the specific implementation manners of this application or make equivalent replacements. Any modification or equivalent replacement that does not depart from the spirit and scope of this application shall be covered by the protection scope of the claims of this application.

Claims

1. A method for generating an air operation plan by combining the inference of the other party's intention, characterized in that Including: During cluster confrontation execution, the threat perception and plan display component receives real-time confrontation situation, evaluates the threat level of cluster targets, and issues threat warnings to the intention reasoning model; the real-time confrontation situation includes the movement state of our platform, the remaining number of devices, target types, and target tracks; Cluster confrontation is an act of using a distributed and collaborative intelligent agent cluster for adversarial offense and defense; The intention reasoning model identifies the possible intention types of the other party, updates the knowledge graph, queries the entity relationships in the knowledge graph, and obtains our alternative action plans; the possible intentions of the other party include the number of the other party's clusters, the cluster movement trend, the cluster action type, and the cluster threat level; the knowledge graph includes entities and entity relationships; The effectiveness evaluation model evaluates the confrontation effectiveness between each possible intention of the other party and our alternative action plans in a template matching manner; our alternative action plans include our current cluster number, cluster types, cluster real-time positions, cluster perception ranges, cluster attack ranges, and resources consumed in cluster confrontation; The game-solving model forms a game matrix from various confrontation situations and confrontation effectiveness, transforms it into an objective optimization problem, and calculates our optimal action plan and execution probability.

2. The method for generating an air operation plan by inferring the intention of the other party according to claim 1, characterized in that The evaluation of the threat level of cluster targets includes: Cluster target Consisting of single targets, the threat level of the th single target is as follows: Among them, , are respectively the weight values of the target air superiority , the target countermeasure ability , , ; Cluster target Threat level is as follows: When the notification intention inference model is notified to start running, is the threat level warning threshold.

3. The method for generating an air operation plan by inferring the intention of the other party according to claim 2, wherein, The evaluation of the threat level of cluster targets includes: The threat level of a single target is related to the target's air superiority; the target's air superiority is related to the confrontation angle, confrontation distance, and speed of both sides, and the target's air superiority is expressed as: Among them, , , are the weight values of the confrontation angle, confrontation distance, and speed respectively; is the course angle between the target and us; is the maximum detection distance of us, is the distance between the detected target and us; is the maximum speed of the other party, is the speed difference between the target and us; , , have the same value range, all greater than 0 and less than 1.

4. The method for generating an air operation plan by reasoning based on the intentions of the other party according to claim 2, characterized in that The evaluation of the threat level of cluster targets includes: The threat level of a single target is related to the target's countermeasure ability; the target's countermeasure ability is related to the maneuverability, strike ability, and detection ability of both platforms, and the target's countermeasure ability is expressed as: Among them, is the mobility ability gap between the two platforms, is the strike ability gap between the two sides, is the platform detection ability gap, and its value varies with the types of the two platforms; among them is the weight value of the mobility ability gap between the two sides, is the weight value of the strike ability gap between the two sides, is the platform detection ability gap between the two sides.

5. The method for generating an air operation plan by inferring the intention of the other party according to claim 1, characterized in that The intention reasoning model infers the possible intentions of the other party based on target information; the intention reasoning model consists of a neural network module and a knowledge reasoning module; the neural network module consists of a long short-term memory network (LSTM), a fully connected network, and a softmax layer. Each LSTM has multiple hidden layers, and each fully connected layer has multiple hidden layers. The ReLU activation function is used, and finally, the intention type is output through logical classification by the softmax layer; The intention types include coordinated attack, hovering, evasion, and formation flight; The neural network module collects the historical track information of cluster targets, identifies the intention type of the current cluster targets, and simultaneously synchronously updates the relevant information of the threat targets to the knowledge reasoning module; the relevant information includes the constituent platforms, types, and intention categories; The knowledge reasoning module includes a knowledge graph and a knowledge rule set. The entities in the knowledge graph are divided into platforms, intention types, adversarial strategy types, clusters, action plans, and targets; the entity relationships in the knowledge graph are divided into plan execution, cluster management, strike object, equipped with sensors, equipped with devices, command and control; the knowledge rule set includes multiple rules, and each rule describes the establishment conditions of entity relationships. If the entity relationship is established, the corresponding entity relationship is queried from the knowledge graph according to the entity relationship to obtain our alternative action plans.

6. The method for generating an air operation plan by inferring the intention of the other party according to claim 1, wherein The effectiveness evaluation model pre-offline deduces the confrontation processes between various possible intentions of the other party and our action plans in advance, takes three types of indicators, namely cost, benefit, and loss, as the effectiveness evaluation results, and forms a template library; the template library includes multiple templates, and each template includes the relative situation between both sides, our action plan, the intention of the other party, and the effectiveness evaluation result. The effectiveness evaluation model matches the template library with the current confrontation situation, combines the evaluation effectiveness value of the template with the highest similarity with the similarity weight, and calculates the evaluation effectiveness of the current confrontation situation; the current confrontation situation includes the relative situation of both sides, our alternative action plans, and the possible intentions of the other side.

7. The method for generating an air operation plan by inferring the intention of the other party according to claim 6, characterized in that Assume the current relative situation between the two sides is , our alternative action plan , the possible intentions of the other party , then the template in the template library with the highest similarity to the current confrontation situation satisfies the following conditions: Among them, is the total number of templates in the template library, is the relative situation in the th template, is the action plan of our side for the th template, is the intention of the other party for the th template, is the current relative situation between the two sides and the relative situation in the th template is the difference weight value, is the alternative action plan of our side and the action plan of our side in the th template is the difference weight value, is the possible intention of the other party and the intention of the other party in the th template is the difference weight value, For the degree of difference between the template and the current adversarial situation, the smaller the degree of difference, the more similar the template is to the current adversarial situation.

8. The method for generating an air operation plan by inferring the intention of the other party according to claim 7, characterized in that, The effectiveness evaluation model evaluates the evaluation effectiveness of the current confrontation situation, including: According to the template with the highest similarity to the current confrontation situation The degree of difference it has , normalize to generate the similarity weight , then the cost of our side in the current confrontation situation , the benefit of our side in the current confrontation situation , the loss of our side in the current confrontation situation , and the overall evaluation efficiency value are respectively: Among them, is the adversarial cost pre-computed in the template, is the calculation weight of is the adversarial benefit pre-computed in the template, is the calculation weight of is the adversarial loss pre-computed in the template, is the calculation weight of is the normalization function.

9. The method for generating an air operation plan by reasoning based on the intention of the other party according to claim 8, characterized in that The game-solving model forms a game matrix by combining various confrontation situations with evaluation effectiveness; the leftmost column of the game matrix constructed by the game-solving model is the alternative action plans of our side, the top row is the possible intentions of the other side, and each matrix value is the alternative action plan of our side and the possible intentions of the other side evaluation value of confrontation effectiveness .

10. The method for generating an air operation plan by inferring the intention of the other party according to claim 8 or 9, characterized in that The game-solving model transforms the game matrix into an objective optimization problem, and calculates our optimal action plan and execution probability, including: The game-solving model maximizes the difference between our gains and losses as the optimization objective and constructs an optimization function , where the optimization variables are . Among them is our th alternative action plan, is the execution probability of the th alternative action plan. This optimization function is solved by the particle swarm optimization (PSO) algorithm to obtain our optimal action plan and execution probability. The core idea of the game-solving model is to find an optimal combination of our alternative action plans and execution probabilities. We confront the opponent according to this optimal combination, and the sum of all confrontation results is optimal. The objective optimization solution form of the game-solving model is as follows: Among them, is the maximum value function, represents the th alternative action plan of our side and the confrontation benefit with the set of possible intentions of the other side , represents the th alternative action plan of our side and the minimized loss with the set of possible intentions of the other side , represents the weight value of the benefit of our side in the current confrontation situation when adopting the th alternative action plan of our side against the th possible intention of the other side , as well as the weight value of the loss of our side in the current confrontation situation , , represents the th execution probability of the alternative action plan and the th execution probability of the possible intention of the other side of the operation result, represents the benefit value of the confrontation between the th alternative action plan of our side and the th possible intention of the other side , represents the loss value of the confrontation between the th alternative action plan of our side and the th possible intention of the other side , is the number of possible intentions of the other side in the set of possible intentions of the other side , is the cost of our side in the current confrontation situation.

11. An air operation plan generation system that combines reasoning of the other party's intentions, implementing the method for generating an air operation plan that combines reasoning of the other party's intentions according to any one of claims 1-10, characterized in that, It includes a threat perception and plan display component, an intention reasoning model, an effectiveness evaluation model, and a game-solving model; When executing cluster confrontation, the threat perception and plan display component is used to receive the real-time confrontation situation and evaluate the threat level of the cluster target; the real-time confrontation situation includes our platform's movement state, the remaining number of devices, the target type, and the target track; The intention reasoning model is used to identify the possible intention types of the other side and initially screen out our alternative action plans using knowledge reasoning; the possible intentions of the other side include the number of the other side's clusters, the cluster movement trend, the cluster action type, and the cluster threat level; The effectiveness evaluation model is used to evaluate the confrontation effectiveness between each possible intention of the other side and our alternative action plans in a template matching manner; our alternative action plans include our current cluster number, cluster type, cluster real-time position, cluster perception range, cluster attack range, and cluster confrontation resource consumption; The game-solving model is used to form a game matrix with various confrontation situations and confrontation effectiveness, transform it into an objective optimization problem, and calculate our optimal action plan and execution probability.

Images