A collaborative confrontation decision-making method, device and storage medium based on artificial intelligence and personnel

Abstract

The present invention discloses a collaborative confrontation decision-making method, device and storage medium based on artificial intelligence and personnel, and belongs to the field of artificial intelligence technology. The method includes: by setting environmental assumptions, preliminary planning settings and situation assumptions, and generating basic scenario assumptions; generating behavioral strategies according to the basic scenario assumptions; inputting the basic scenario assumptions and behavioral strategies into the intelligent confrontation deduction engine for simulation deduction, and obtaining simulation deduction results and simulation result scores; inputting the result scores, behavioral strategies and basic scenario assumptions into the artificial intelligence decision-making model, and the model is based on the simulation result scores, and the strategy selection factors are adjusted until the behavioral strategies under the preset target scores are output; the personnel exert their subjective initiative and make subjective judgments on the strategies. The present invention closely links the artificial intelligence strategy with the personnel strategy, so that the decision-making model can make quick and wise decisions in a complex and changeable environment, and the decision-making efficiency is significantly improved.

Description

Technical Field

[0001] The present invention relates to a collaborative confrontation decision-making method, device and storage medium based on artificial intelligence and personnel, and belongs to the technical field of artificial intelligence. Background Art

[0002] Military strategy simulation is an important means of military training and decision support, which can help military personnel better understand and deal with complex and changing battlefield environments. With the continuous advancement of technology, simulation methods will continue to be improved and optimized to provide more powerful and effective support for military decision-making.

[0003] Traditional military strategy simulation can only be simulated based on artificial strategies and assumptions, so it consumes a lot of manpower. With the increasing maturity of artificial intelligence, the current simulation devices are gradually transformed into related research that consumes a lot of manpower and is time-consuming to improve the comprehensive benefits of military intelligence research. However, although simulation can simulate battlefield environments and combat operations, it cannot fully simulate people's decision-making processes and psychological factors. On the real battlefield, commanders' decisions are often affected by many factors, such as emotions, experience, intuition, etc., which are difficult to accurately simulate in simulation.

[0004] Therefore, there is an urgent need to provide a collaborative confrontation decision-making method that combines artificial intelligence and personnel to solve the technical drawbacks faced in existing military strategy simulations. Summary of the invention

[0005] The purpose of the present invention is to provide a collaborative confrontation decision-making method, device and storage medium based on artificial intelligence and personnel, aiming to combine the objectivity and efficiency of artificial intelligence decision-making with the subjective initiative of personnel decision-making to improve the comprehensiveness and accuracy of decision-making.

[0006] In order to solve the above technical problems, the present invention is implemented by adopting the following technical solutions:

[0007] In one aspect, the present invention provides a collaborative confrontation decision-making method, comprising:

[0008] By setting environmental scenarios, preliminary planning and situation scenarios, and generating corresponding basic scenario scenarios;

[0009] Generate a behavior strategy based on the basic scenario, wherein the behavior strategy includes a first behavior strategy based on an artificial intelligence decision model and a second behavior strategy based on personnel;

[0010] Inputting the basic scenario assumptions and the behavior strategies into the intelligent confrontation deduction engine for simulation deduction to obtain simulation deduction results;

[0011] Scoring the simulation results to obtain simulation result scores based on the basic scenario assumptions and the behavior strategies;

[0012] The simulation result score, behavior strategy and basic scenario assumption are input into the artificial intelligence decision model, and the artificial intelligence decision model is based on the simulation result score, and the strategy selection factor is adjusted until the corresponding route, formation and wave under the preset target score are output;

[0013] Among them, the personnel exercise subjective initiative and make subjective judgment on the first behavioral strategy.

[0014] Big data learning in the prior art mostly requires a large number of artificial samples for learning. The present invention, through the self-production and sales of large sample data, feeds back the simulation decision results into the artificial intelligence decision model, thereby realizing the automatic generation of deep learning samples and the repeatable and automated application of result scoring; secondly, the original large sample learning still requires professional debugging and complex analysis, and the efficiency is relatively low. The present invention innovatively introduces the subjective judgment of personnel into the final determination link of decision-making, so that the intelligent body can automatically enter big data for sample learning under the dual guarantee of personnel verification, thereby realizing high efficiency and optimization of behavioral strategy generation, and having strong purposefulness.

[0015] Optionally, the preliminary planning settings include equipment selection, load planning and action capability determination.

[0016] Optionally, the first behavior strategy generates a decision corresponding to the artificial intelligence model based on the basic scenario assumption and through deep learning of a large sample;

[0017] The second behavior strategy forms a formation for the equipment in the basic scenario, plans routes based on the situation and environment scenarios in the basic scenario, sets beams and waves corresponding to the time strategy, and generates corresponding decisions for the personnel.

[0018] Optionally, the simulation includes physical simulation and damage power simulation.

[0019] Optionally, the behavior strategy includes: resource pool, party relationship and formation pool.

[0020] The resource pool includes: the party, the major category and name of the entity unit, the total number of entity units and the number of allocated entity units;

[0021] The party relationships include: red and blue, red and unknown, and blue and unknown;

[0022] The formation pool includes: the master station method of the subordinate party, the tactics used by the subordinate party, the name of the subordinate party's target, the success or failure conditions of the subordinate party, the task type, the name of the task target, whether the task is a region, whether the task is main, the minimum height of the task area, the maximum height of the task area, the success or failure conditions of the task, the task area point set, the task receiving formation, the formation name, the formation controller, the formation intelligent algorithm, the member type of the formation, the wave time of the formation, the mounted unit of the formation, the center position of the formation, the task of the formation, the simulation name of the entity unit, the major category of the entity unit, the name of the entity unit, the type of the entity unit, the carrying formation of the entity unit, the components of the entity unit, the position of the entity unit, the posture of the entity unit, the behavior of the entity unit, the load of the entity unit, waypoint data, waypoint name, waypoint type, waypoint position, waypoint speed, waypoint standby time, waypoint attack mode, waypoint attack target, whether the waypoint radar is turned on, whether the waypoint jammer is turned on, the waypoint interference object and the waypoint formation offset data.

[0023] Optionally, the simulation results include: simulation results based on force statistics, and simulation results based on group statistics.

[0024] The simulation results by force statistics include: number of troops, lost troops, lost ammunition, loss ratio, simulation duration, tactics used by the party and simulation result score;

[0025] The simulation results by group statistics include: a row of formation data, formation name, force type, number of combat units under the formation, formation controller, survival status and mission objectives.

[0026] Optionally, the loss function of the artificial intelligence decision model is expressed as:

[0027]

[0028] In the formula, L(θ) represents the loss function, N represents the number of training samples, θ represents the parameters of the model, and x i represents the input vector, t i Represents the target output vector.

[0029] Optionally, the parameters of the artificial intelligence decision model are updated using a gradient descent method, expressed as:

[0030]

[0031] In the formula, η represents the learning rate, Represents the gradient of the loss function L(θ) with respect to θ.

[0032] Optionally, the simulation result score is expressed as:

[0033]

[0034] Where SC represents the simulation result score of the red team, a represents the survival ratio of the blue team, b represents the survival ratio of the red team, and p represents the engagement time consumption.

[0035] The survival ratio of the blue team and the survival ratio of the red team are respectively expressed as:

[0036] a=c / (c+d);

[0037] b = e / (e+f);

[0038] In the formula, c represents the number of survivors at the end of the blue side's deduction, d represents the number of deaths at the end of the blue side's deduction; e represents the number of survivors at the end of the red side's deduction, and f represents the number of deaths at the end of the red side's deduction. It is worth noting that the red side represents the strategic side, and the blue side represents the enemy side.

[0039] In a second aspect, the present invention provides a collaborative confrontation decision-making device, comprising:

[0040] The basic scenario module is configured to generate a basic scenario through setting an environment scenario, setting a preliminary planning setting, and setting a situation scenario;

[0041] A behavior strategy generation module is configured to generate a behavior strategy according to the basic scenario, wherein the behavior strategy includes a first behavior strategy based on an artificial intelligence decision model and a second behavior strategy based on personnel;

[0042] A simulation execution module is configured to input the basic scenario assumption and the behavior strategy into the intelligent confrontation simulation engine for simulation and obtain simulation results;

[0043] An execution result analysis module is configured to score the simulation result to obtain a simulation result score based on the basic scenario assumption and the behavior strategy;

[0044] And, a final decision determination module is configured to input the simulation result score, the behavior strategy and the basic scenario assumption into the artificial intelligence decision model, and the artificial intelligence decision model is based on the simulation result score, and adjusts the strategy selection factor until the corresponding route, formation and wave under the preset target score are output;

[0045] Among them, the personnel exercise subjective initiative and make subjective judgment on the first behavioral strategy.

[0046] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the collaborative confrontation decision-making method as described in the first aspect.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] (1) The present invention introduces subjective judgment of personnel into the determination link of behavioral decision-making, so that artificial intelligence and human decision-makers work together, are interrelated and influence each other, so that users can choose the best from the best to identify the best task plan strategy, saving a lot of time and labor costs, and having a strong purpose;

[0049] (2) At the same time, the integrated simulation and deduction platform provided by the present invention and the related engine can provide a large amount of real-time data and feedback for the artificial intelligence decision-making model by simulating the real battlefield environment and confrontation actions, thereby realizing the self-production and self-sale of large sample data, significantly improving the sample collection efficiency, and having strong practicality. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 The figure is a flow chart of the collaborative confrontation decision generation based on artificial intelligence and personnel in the present invention. DETAILED DESCRIPTION

[0051] The technical concept of the present invention is: first, by artificially assuming a basic assumption environment, the corresponding basic assumption file is produced to make preliminary preparations for the simulation; second, based on the generated basic assumption file, the behavior strategy is edited in two types of interfaces; further, based on the basic assumption file and the behavior strategy, the simulation is executed and the execution results are analyzed; finally, the result score obtained by the corresponding simulation is combined with the strategy file as a feedback sample of the artificial intelligence decision-making model for deep learning, and the strategy selection factors are adjusted sequentially to obtain the optimal simulation strategy under the preset confrontation target score.

[0052] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0053] Example 1

[0054] This embodiment introduces a collaborative confrontation decision-making method, such as Figure 1 As shown, it includes the following steps:

[0055] First, through the setting of environmental scenarios, preliminary planning settings and situational scenarios, basic scenario scenarios are generated;

[0056] Secondly, generating a behavior strategy according to the basic scenario assumption, wherein the behavior strategy includes a first behavior strategy based on an artificial intelligence decision model and a second behavior strategy based on personnel;

[0057] Furthermore, the basic scenario assumption and the behavior strategy are input into the intelligent confrontation deduction engine for simulation deduction to obtain simulation deduction results;

[0058] The behavioral strategy usually refers to a series of actions or decisions taken in a specific environment to achieve a predetermined goal. In the intelligent confrontation deduction engine, the behavioral strategy may involve the action mode, reaction method and response strategy of the opponent. The formulation of these strategies needs to consider many factors, including the strength comparison between the enemy and us, the battlefield environment and resource conditions.

[0059] Simulation technology provides important support for the formulation of behavioral strategies. Through simulation, the real battlefield environment can be simulated, and the possible results of different behavioral strategies can be predicted, thereby helping decision makers choose the optimal strategy. The intelligent confrontation deduction engine is one of the specific applications of simulation technology. It combines artificial intelligence and simulation technology, and can automatically conduct confrontation deductions to simulate the changes in battlefield situations under different strategies. In the decision-making process, artificial intelligence decision-making and human decision-making play a complementary role. Artificial intelligence decision-making uses technologies such as machine learning and data analysis to provide objective and efficient decision-making support; while human decision-making can give play to subjective initiative and make judgments based on experience and intuition. Through the combination of the two, decisions can be more comprehensive and accurate.

[0060] Further, the simulation results are scored to obtain simulation result scores based on the basic scenario assumptions and the behavioral strategies; the simulation result scores, behavioral strategies and basic environment assumptions are input into an artificial intelligence decision model, and the artificial intelligence decision model is based on the simulation result scores, and the strategy selection factors are adjusted until the corresponding route, formation and wave under the preset target score are output;

[0061] Among them, the strategy selection factors are the key factors that affect decision-making. These factors include the strength of both sides, battlefield environment, resource conditions, and time constraints. Through comprehensive analysis and evaluation of these factors, the priority and applicability of different strategies can be determined accordingly, so as to select the most appropriate strategy.

[0062] It is worth noting that the personnel exercise their subjective initiative and make subjective judgments on the first behavior strategy. The close connection and interaction between artificial intelligence strategy and personnel strategy enables the artificial intelligence model to make quick and wise decisions in a complex and changing environment to achieve the predetermined goals in a targeted manner.

[0063] Example 2

[0064] Based on Example 1, this embodiment also has the following design:

[0065] This embodiment divides the overall intelligent simulation deduction process into a basic assumption setting stage, a behavior strategy generation stage, a deduction execution stage, and an execution result analysis stage. The artificial intelligence decision model continuously adjusts the strategy selection factor, and the personnel simultaneously make subjective judgments on the behavior decision, so that the simulation result score continuously approaches the preset target score, and then determines the optimal deduction simulation strategy under the preset target score. The specific implementation process is as follows:

[0066] The first stage: setting the basic scenario, through artificial assumption of the production of basic assumption environment, which includes three module functions: environment assumption setting (Environment, algorithm abbreviation EV), preliminary planning setting (Resource, algorithm abbreviation RS), and situation assumption setting (Mission, algorithm abbreviation MS). It is used to make preliminary preparations for the simulation, mainly setting the simulation's time and space environment, the capabilities of all parties and the situation distribution, and finally generating a basic scene file SceneFile (EV+RS+MS).

[0067] The military capabilities used in the first phase need to be set by the user, and the set parameters must comply with the rules and descriptions of the simulation equipment database parameters in Table 1, as follows:

[0068] Table 1 Simulation equipment database parameters

[0069]

[0070]

[0071] In this stage, the environment scenario setting (Environment, algorithm abbreviation EV), preliminary planning setting (Resource, algorithm abbreviation RS), and situation scenario setting (Mission, algorithm abbreviation MS) can be assumed as follows:

[0072] Environmental scenario setting EV: wind force is level 4, 4 am, location is 121 degrees east longitude, 24 degrees north latitude;

[0073] Situation setting (mission) MS: The Red side is the drone attacker, and the mission goal is to attack the Blue side's airport tower;

[0074] RS was set in the early planning: the Red side has 10 suicide drones with a speed of 50 kilometers per hour, which can take off 30 kilometers away from the target; the Blue side's scene layout is 1 radar, 1 tower, and 2 directed energy weapons. The radar can detect drones with a coverage radius of 20 kilometers. The tower is 10 kilometers away from the radar, and 2 directed energy weapons are near the tower. The interval between directed energy attacks is 5 minutes each time.

[0075] The SceneFile file saves the above information as a basic scene scenario file, and the specific parameter information of the equipment contained in it can use the default value.

[0076] The second stage: Generate the behavior strategy BehaviorLogic, and edit the behavior strategy in two class interfaces based on the basic scene file SceneFile generated in the first stage;

[0077] The first category is the human interactive operation HumanBehaviorLogic, which forms a formation (Form, algorithm abbreviation FM) of the equipment in the basic scenario, and plans the route (WayPoint, algorithm abbreviation WP) based on the situation and environment assumptions in the basic scenario, and sets the beam and wave (Wave, algorithm abbreviation WV) corresponding to the time strategy to generate corresponding decisions for the personnel.

[0078] Although artificial intelligence can provide a large amount of data and simulation results, human decision makers can make comprehensive considerations and trade-offs based on their own experience, intuition and judgment, combined with the information provided by artificial intelligence. Therefore, the flexibility and innovation of human decision-making enable the comprehensive and determined behavioral strategies to cope with complex and changing situations and take into account some factors that are difficult to quantify.

[0079] The second category is based on the basic scene SceneFile assumptions and generates AIBehaviorLogic corresponding to the artificial intelligence model through expert experience learning and large-sample deep learning. Through the analysis and learning of large amounts of data, artificial intelligence can simulate and predict adversarial actions in different scenarios. It automatically makes decisions and deductions based on preset rules and algorithms, thereby providing decision support and assistance. Therefore, artificial intelligence can greatly improve the speed and accuracy of deductions, allowing personnel decision makers to have a more comprehensive understanding of battlefield situations and potential threats.

[0080] The construction and training process of the artificial intelligence decision-making model in this stage is as follows:

[0081] Construct a neural network model a function f, which maps the input vector x to the output vector y. Specifically, the input vector x contains environment (EV), resource (RS) and mission (MS) information, and the output vector y contains formation information (FM), route information (WP) and wave information (WV).

[0082] y=f(x;θ);

[0083] Here, θ represents the parameters (weights and biases) of the model, which will be optimized during the training process.

[0084] Furthermore, a loss function L(θ) is constructed to measure the difference between the model prediction and the true value. In supervised learning, we usually have a target output, and we hope that the model prediction is as close to this target as possible. The choice of loss function depends on the nature of the problem. For example, mean square error (MSE) is often used for regression problems.

[0085] The loss function is expressed as:

[0086]

[0087] In the formula, L(θ) represents the loss function, N represents the number of training samples, θ represents the parameters of the model, and x i represents the input vector, t i Represents the target output vector.

[0088] The final behavior strategy produced through these two types of interfaces conforms to the unified script specification, and in the second stage, a behavior strategy script BehaviorLogicFile (FM+WP+WV) designed based on the basic scenario file SceneFile in the first stage is generated.

[0089] The behavior strategy script parameters include: resource pool, party relationship and formation pool. The rules and descriptions of specific behavior strategy script parameters are shown in Table 2:

[0090] Table 2 Behavior strategy script parameters

[0091]

[0092]

[0093] At this stage, the corresponding formation information (FM), route information (WP) and wave information (WV) are as follows:

[0094] Formation Information FM: The 10 Red Army drones are divided into three groups, three, three, and four sorties respectively, and the formation used is a straight column;

[0095] Route information WP: Close to the sea, attack the target from the northwest, north, and northeast at economic speed;

[0096] Wave information WV: Set to arrive at the target point at the same time.

[0097] The set strategy is saved in BehaviorLogicFile, and the specific formation and waypoint information contained therein are not repeated here.

[0098] The third stage: simulation execution, based on the basic scene file SceneFile generated in the first stage and the behavior strategy script BehaviorLogicFile generated in the second stage, the simulation is performed. During the simulation process, the environmental data is called to participate in the simulation, and simulation modules such as physical simulation and damage power simulation are performed, and finally the simulation results ResultFile are collected.

[0099] At this stage, the corresponding deduction results are: the Red side achieves its goal and the remaining two drones return.

[0100] The fourth stage: parse the execution results. The simulation result ResultFile obtained in the third stage needs to be scored by the evaluation algorithm, and finally the deduction result score (Score, algorithm abbreviation SC) based on the current assumption and current strategy is obtained. The result score combined with the strategy file BehaviorLogicFile will be used as a feedback sample for artificial intelligence decision-making for deep learning.

[0101] The simulation result parameters include: a total table of statistical results by power and a sub-table of statistical results by group. The specific simulation result parameter description is shown in Table 3:

[0102] Table 3 Simulation results parameters

[0103]

[0104] The simulation result score is expressed as:

[0105]

[0106] Where SC represents the simulation result score of the red team, a represents the survival ratio of the blue team, b represents the survival ratio of the red team, and p represents the engagement time consumption.

[0107] The survival ratio of the blue team and the survival ratio of the red team are respectively expressed as:

[0108] a=c / (c+d);

[0109] b = e / (e+f);

[0110] Where c represents the number of survivors at the end of the blue team's simulation, and d represents the number of deaths at the end of the blue team's simulation; e represents the number of survivors at the end of the red team's simulation, and f represents the number of deaths at the end of the red team's simulation.

[0111] Among them, the red side represents the strategic side, and the blue side represents the enemy side. According to the calculations in the first three stages, the red side's score in this stage is 68 points.

[0112] The obtained score result SC, the SceneFile generated in the first stage, and the BehaviorLogic generated in the second stage are input into the artificial intelligence decision-making model to train and adjust the intelligent agent until the generated BehaviorLogic data no longer changes, and the optimal deployment strategy is obtained.

[0113] Optimization algorithms are used to update the parameters θ of the AI ​​decision model to minimize the loss function L(θ). In deep learning, commonly used optimization algorithms include gradient descent, Adam, RMSprop, etc. These algorithms work by calculating the gradient of the loss function with respect to the model parameters and updating the parameters in the opposite direction of the gradient.

[0114] Taking gradient descent as an example, the parameter update rule is as follows:

[0115]

[0116] Among them, η represents the learning rate, which is used to control the step size of parameter update; represents the gradient of the loss function with respect to θ.

[0117] The judgment function is used to calculate a score SC based on the output y of the model. This score is used to evaluate the performance of the model on a specific task. The specific form of the judgment function depends on the requirements and evaluation criteria of the task.

[0118] SC = g(y);

[0119] During training, our goal is to find a set of parameters θ so that for a given input, the model output can obtain the highest possible score SC through the judgment function.

[0120] In the specific application of this embodiment, first, the parameters θ of the model are randomly initialized; secondly, for each training sample, the model output is calculated. Among them, the loss function L is used to calculate the difference between the model output and the true value, and the gradient of the loss function with respect to the model parameters is calculated; at the same time, the model parameters θ are updated using the optimization algorithm until the stopping condition is met (such as reaching the maximum number of iterations or loss convergence); in addition, the judgment function is used to calculate the score SC of the model on the validation set or test set. The decision-making agent of this embodiment learns in the process of continuous interaction with the simulation training environment, replacing the decision-making role of people in the confrontation between the red and blue parties. The deep neural network will be used to vectorize the environment, the red and blue states, and the task objectives. By continuously updating the strategy selection factor, the optimal deployment strategy is found in the exploration and utilization based on the deep reinforcement learning algorithm, so as to support the intelligent operation of the deduction system.

[0121] Example 3

[0122] This embodiment provides a collaborative confrontation decision-making device, which includes:

[0123] The basic scenario module is configured to generate a basic scenario through setting an environment scenario, setting a preliminary planning setting, and setting a situation scenario;

[0124] A behavior strategy generation module is configured to generate a behavior strategy according to the basic scenario, wherein the behavior strategy includes a first behavior strategy based on an artificial intelligence decision model and a second behavior strategy based on personnel;

[0125] A simulation execution module is configured to input the basic scenario assumption and the behavior strategy into the intelligent confrontation simulation engine for simulation and obtain simulation results;

[0126] An execution result analysis module is configured to score the simulation result to obtain a simulation result score based on the basic scenario assumption and the behavior strategy;

[0127] And, a final decision determination module is configured to input the simulation result score, behavior strategy and basic scenario setting into the artificial intelligence decision model, and the artificial intelligence decision model adjusts the strategy selection factor based on the simulation result score until the corresponding route, formation and wave under the preset target score are output;

[0128] Among them, the personnel exercise subjective initiative and make subjective judgment on the first behavioral strategy.

[0129] The device in this embodiment provides intuitive and realistic scene environment editing through a complete tool chain, which can simulate possible future objects and tasks under a certain task background. Accordingly, users can be familiar with the task objects and task areas, test and familiarize themselves with the task plans, and conduct demonstration and evaluation of the task plans and improve and inspire new plans; at the same time, they can also carry out deductions according to the established task plans and rules to improve the user's decision-making level and ability.

[0130] In summary, through repeated simulations and the use of advanced technologies such as reinforcement learning, users can select the best from the best, identify the best task solutions and save a lot of time and labor costs.

[0131] Example 4

[0132] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the collaborative confrontation decision-making method described in Example 1 or 2 are implemented.

[0133] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0134] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (apparatus), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0135] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0136] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0137] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A collaborative confrontation decision-making method, characterized in that: include: Generate basic scenario scenarios through setting environmental scenarios, preliminary planning settings and situation scenarios; Generate a behavior strategy based on the basic scenario, wherein the behavior strategy includes a first behavior strategy based on an artificial intelligence decision model and a second behavior strategy based on personnel; Inputting the basic scenario assumptions and the behavior strategies into a human-in-the-loop intelligent confrontation deduction engine for simulation deduction to obtain simulation deduction results; Scoring the simulation results to obtain simulation result scores based on the basic scenario assumptions and the behavior strategies; The simulation result score, behavior strategy and basic scenario setting are input into the artificial intelligence decision model, and the artificial intelligence decision model is based on the simulation result score, and adjusts the strategy selection factor until the corresponding route, formation and wave under the preset target score are output; The personnel exert their subjective initiative and make a subjective judgment on the first behavior strategy, including: the personnel select the first behavior strategy of the artificial intelligence decision model according to the simulation result score to obtain the first behavior strategy after the selection; based on the first behavior strategy after the selection, the simulation result score is continuously approached to the preset target score, thereby determining the optimal collaborative confrontation decision; The loss function of the artificial intelligence decision model is expressed as: ； In the formula, represents the loss function, represents the number of training samples, represents the parameters of the model, represents the input vector, represents the target output vector; The parameters of the artificial intelligence decision model are updated using the gradient descent method, which is expressed as: ； In the formula, represents the learning rate, Represents the loss function about gradient.

2. The collaborative confrontation decision-making method according to claim 1, characterized in that: The preliminary planning settings include equipment selection, load planning and action capability determination.

3. The collaborative confrontation decision-making method according to claim 2, characterized in that: The first behavior strategy generates a decision corresponding to the artificial intelligence model based on the basic scenario assumption and through deep learning of a large sample; The second behavior strategy is to form a formation for the equipment in the basic scenario, plan routes based on the situation and environment scenarios in the basic scenario, set waves corresponding to the time strategy, and generate corresponding decisions for the personnel.

4. The collaborative confrontation decision-making method according to claim 3, characterized in that: The simulation includes physical simulation and damage power simulation.

5. The collaborative confrontation decision-making method according to claim 4, characterized in that: The behavior strategy includes: resource pool, party relationship and formation pool; The simulation results include: simulation results based on force statistics and simulation results based on group statistics.

6. The collaborative confrontation decision-making method according to claim 1 or 5, characterized in that: The simulation result score is expressed as: ； In the formula, represents the simulation result score of the red team, Represents the survival ratio of the blue side, represents the survival ratio of the red side, Indicates the time consumed in the battle; The survival ratio of the blue team and the survival ratio of the red team are respectively expressed as: ； ； In the formula, Indicates the number of survivors at the end of the blue team's deduction. Indicates the number of deaths at the end of the Blue team's game; Indicates the number of survivors at the end of the Red Team’s game. Indicates the number of deaths at the end of the Red Team game.

7. A collaborative confrontation decision-making device, comprising: The basic scenario module is configured to generate a basic scenario through setting an environment scenario, setting a preliminary planning setting, and setting a situation scenario; A behavior strategy generation module is configured to generate a behavior strategy according to the basic scenario, wherein the behavior strategy includes a first behavior strategy based on an artificial intelligence decision model and a second behavior strategy based on personnel; A simulation execution module is configured to input the basic scenario assumption and the behavior strategy into the intelligent confrontation simulation engine for simulation and obtain simulation results; An execution result analysis module is configured to score the simulation result to obtain a simulation result score based on the basic scenario assumption and the behavior strategy; And, a final decision determination module is configured to input the simulation result score, behavior strategy and basic scenario assumption into an artificial intelligence decision model, and the artificial intelligence decision model adjusts the strategy selection factor based on the simulation result score until the corresponding route, formation and wave under the preset target score are output; The personnel exert their subjective initiative and make a subjective judgment on the first behavior strategy, including: the personnel select the first behavior strategy of the artificial intelligence decision model according to the simulation result score to obtain the first behavior strategy after the selection; based on the first behavior strategy after the selection, the simulation result score is continuously approached to the preset target score, thereby determining the optimal collaborative confrontation decision; The loss function of the artificial intelligence decision model is expressed as: ； In the formula, represents the loss function, represents the number of training samples, represents the parameters of the model, represents the input vector, represents the target output vector; The parameters of the artificial intelligence decision model are updated using the gradient descent method, which is expressed as: ； In the formula, represents the learning rate, Represents the loss function about gradient.

8. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps of the collaborative confrontation decision-making method as described in any one of claims 1 to 6 are performed.

Images