Combat decision-making method and device of combat entity, storage medium and electronic equipment

Abstract

The invention relates to the technical field of intelligent decision making, and provides a combat decision making method and device for a combat entity, a computer readable storage medium and electronic equipment, the surrounding space of the combat entity is divided into a plurality of defense areas, and N combat weapons are deployed on the combat entity, each combat weapon is deployed in at least one defense area on a combat entity, and the method comprises the following steps: obtaining feature parameters of M incoming attacking targets for the combat entity at the current moment, and obtaining overall combat effectiveness parameters of the combat entity for the M incoming attacking targets at the current moment; inputting the overall combat effectiveness parameters and the characteristic parameters into a trained combat decision model, and deciding a combat strategy for M incoming attack targets through the trained combat decision model; wherein the combat strategy is used for representing the attack probability of each combat weapon for each incoming attack target. The method can effectively cope with a multi-target and dynamically changing complex battlefield environment.

Description

Background Art

[0002] Under digital conditions, the battlefield of military operations is highly transparent, the entities involved in the combat are complex and exist in multi-dimensional spaces such as sea, land, air and space, and participate in combat operations in a coordinated manner, making the war process more complicated. This puts higher demands on the command and decision-making personnel's global information acquisition capabilities, global observation and operational capabilities, and the control and coordination capabilities of participating entities in multi-dimensional spaces.

[0003] In related technologies, it is generally necessary to rely on pre-set strategies and rules to deal with various incoming targets. This solution is effective when facing simple and foreseeable threats, but its limitations gradually become apparent when facing complex battlefield environments with multiple targets and dynamic changes.

[0004] In view of this, there is an urgent need in this field to develop a new combat decision-making method and device for combat entities.

[0005] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure. Summary of the invention

[0006] The purpose of the present disclosure is to provide a combat decision-making method for a combat entity, a combat decision-making device for a combat entity, a computer-readable storage medium and an electronic device, thereby overcoming, at least to a certain extent, the technical problem of being unable to cope with a complex battlefield environment with multiple targets and dynamic changes due to the limitations of related technologies.

[0007] Other features and advantages of the present disclosure will become apparent from the following detailed description, or may be learned in part by the practice of the present disclosure.

[0008] According to a first aspect of the present disclosure, a combat decision method for a combat entity is provided, wherein the surrounding space of the combat entity is divided into a plurality of defense areas, N combat weapons are deployed on the combat entity, and each of the combat weapons is deployed in at least one defense area on the combat entity, and the method comprises:

[0009] Acquire characteristic parameters of M incoming targets of the combat entity at the current moment, and acquire overall combat effectiveness parameters of the combat entity at the current moment for the M incoming targets;

[0010] Inputting the overall combat effectiveness parameter and the characteristic parameter into a trained combat decision model, and determining a combat strategy for the M incoming targets through the trained combat decision model;

[0011] The combat strategy is used to characterize the attack probability of each combat weapon against each incoming target, and M and N are both integers greater than 1.

[0012] In an exemplary embodiment of the present disclosure, the overall combat effectiveness parameter is composed of the combat effectiveness parameter of each of the combat weapons against each of the incoming targets at the current moment;

[0013] The combat effectiveness parameter of each of the combat weapons at the current moment for each of the incoming targets corresponds to L dimensions;

[0014] The overall combat effectiveness parameter corresponds to the N*M*L dimension;

[0015] Wherein, L is an integer greater than 1.

[0016] In an exemplary embodiment of the present disclosure, the combat effectiveness parameters of each of the combat weapons against each of the incoming targets at the current moment include:

[0017] Characterization parameters of the expected combat effect of each of the combat weapons against each of the incoming targets and the performance parameters of each of the combat weapons at the current moment.

[0018] In an exemplary embodiment of the present disclosure, the number N of the combat weapons is determined based on the number of types of the combat weapons, the number of each type of combat weapon, and the number of attack modes associated with each combat weapon.

[0019] In an exemplary embodiment of the present disclosure, the trained combat decision model includes a policy network, and the policy network includes a coding layer and a decision layer;

[0020] The step of inputting the overall combat effectiveness parameter and the characteristic parameter into a trained combat decision model, and determining a combat strategy for the M incoming targets through the trained combat decision model, comprises:

[0021] Inputting the overall combat effectiveness parameter into the coding layer, encoding the overall combat effectiveness parameter through the coding layer to obtain a coding matrix;

[0022] The decision layer determines the combat strategy for the M incoming targets based on the coding matrix and the characteristic parameters.

[0023] In an exemplary embodiment of the present disclosure, the encoding layer includes a preprocessing layer, a feature extraction layer, a first multi-head self-attention layer, and a first feedforward network;

[0024] The encoding process of the overall combat effectiveness parameter through the encoding layer to obtain the encoding matrix includes:

[0025] Preprocessing the overall combat effectiveness parameters through the preprocessing layer to obtain an initial matrix;

[0026] Performing feature extraction processing on the initial matrix through the feature extraction layer to obtain a word embedding matrix; the word embedding matrix is ​​composed of M row vectors;

[0027] Perform multi-head attention allocation on each of the row vectors through the first multi-head self-attention layer to obtain each attention output matrix;

[0028] The M attention output matrices are optimized through the first feedforward network to obtain the encoding matrix; the encoding matrix is ​​composed of M first matrices corresponding to the M attention output matrices.

[0029] In an exemplary embodiment of the present disclosure, the first multi-head self-attention layer includes a plurality of attention heads;

[0030] The performing multi-head attention allocation on each of the row vectors through the first multi-head self-attention layer to obtain each attention output matrix includes:

[0031] For each attention head, performing attention allocation processing on each of the row vectors according to the key weight matrix, the value weight matrix and the query weight matrix of each attention head to obtain an output matrix corresponding to each attention head;

[0032] Determine each attention output matrix according to the multiple output matrices corresponding to the multiple attention heads.

[0033] In an exemplary embodiment of the present disclosure, for each attention head, performing attention allocation processing on each row vector according to the key weight matrix, the value weight matrix and the query weight matrix of each attention head, and obtaining an output matrix corresponding to each attention head, includes:

[0034] For each attention head, multiply each row vector by the query weight matrix to obtain a query vector;

[0035] Multiply each of the row vectors by the key weight matrix to obtain a key vector;

[0036] Multiply each of the row vectors by the value weight matrix to obtain a value vector;

[0037] An output matrix corresponding to each of the attention heads is determined based on the query vector, the key vector, and the value vector.

[0038] In an exemplary embodiment of the present disclosure, determining the output matrix corresponding to each attention head based on the query vector, the key vector and the value vector includes:

[0039] Obtaining a transposed vector of the key vector, and performing a dot product operation on the query vector and the transposed vector of the key vector to obtain an operation result;

[0040] Normalizing the operation result to obtain an attention score;

[0041] The attention score is multiplied by the value vector to obtain an output matrix corresponding to each attention head.

[0042] In an exemplary embodiment of the present disclosure, the decision layer includes a feature transformation layer, a second multi-head self-attention layer, a multi-head attention layer, and a second feedforward network;

[0043] Outputting the combat strategy for the M incoming targets through the decision layer according to the coding matrix and the characteristic parameters includes:

[0044] Performing feature transformation processing on the feature parameters through the feature transformation layer to obtain a transformation matrix;

[0045] Performing multi-head self-attention allocation on the transformation matrix through the second multi-head self-attention layer to obtain a second attention output matrix; the second multi-head self-attention layer includes M attention heads, and the second attention output matrix includes M second matrices corresponding to the M attention heads;

[0046] Performing multi-head attention allocation on the encoding matrix and the second attention output matrix through each attention head of the multi-head attention layer to obtain each third attention output matrix; the number of attention heads in the multi-head attention layer is equal to N;

[0047] Optimizing the N third attention output matrices through the second feedforward network to obtain a probability distribution matrix;

[0048] The probability distribution matrix is ​​used to characterize the combat strategy against the M incoming targets, the number of rows of the probability distribution matrix is ​​equal to the total number N of the combat weapons deployed on the combat entity, and the number of columns of the probability distribution matrix is ​​equal to the total number M of the at least one incoming target; the probability distribution matrix is ​​used to describe the attack probability of each of the combat weapons against each of the incoming targets.

[0049] In an exemplary embodiment of the present disclosure, the encoding matrix includes M first matrices, and the second attention output matrix includes M second matrices;

[0050] The performing multi-head attention allocation on the encoding matrix and the second attention output matrix through each attention head of the multi-head attention layer to obtain each third attention output matrix includes:

[0051] Multiply each of the first matrices by a preset key weight matrix to obtain each specified key vector;

[0052] Multiply each of the first matrices by a preset value weight matrix to obtain each specified value vector;

[0053] Multiply each of the second matrices by a preset query weight matrix to obtain each specified query vector;

[0054] Attention is allocated to each of the specified key vectors, each of the specified value vectors, and each of the specified query vectors by the multi-head attention layer to determine each specified fusion vector;

[0055] Determine each of the third attention output matrices based on the M specified fusion vectors.

[0056] According to a second aspect of the present disclosure, a combat decision-making device of a combat entity is provided, wherein the surrounding space of the combat entity is divided into a plurality of defense areas, N combat weapons are deployed on the combat entity, and each of the combat weapons is deployed in at least one defense area on the combat entity, and the device comprises:

[0057] A parameter acquisition module, used to acquire characteristic parameters of the M incoming targets of the combat entity at the current moment, and to acquire overall combat effectiveness parameters of the combat entity at the current moment for the M incoming targets;

[0058] A decision module, used for inputting the overall combat effectiveness parameter and the characteristic parameter into a trained combat decision model, and determining a combat strategy for the M incoming targets through the trained combat decision model;

[0059] The combat strategy is used to characterize the attack probability of each combat weapon against each incoming target, and M and N are both integers greater than 1.

[0060] According to a third aspect of the present disclosure, there is provided a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the combat decision-making method of the combat entity described in the first aspect above is implemented.

[0061] According to a fourth aspect of the present disclosure, an electronic device is provided, comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the combat decision-making method of the combat entity described in the first aspect above by executing the executable instructions.

[0062] It can be seen from the above technical solutions that the combat decision method of the combat entity, the combat decision device of the combat entity, the computer-readable storage medium and the electronic device in the exemplary embodiments of the present disclosure have at least the following advantages and positive effects:

[0063] In the technical solutions provided in some embodiments of the present disclosure, on the one hand, by training the combat decision model, it is possible to intelligently respond to multi-target, multi-dimensional complex environments through the model, and make decisions quickly, greatly shortening the decision-making time, improving the real-time response capability, and dynamically adapting to the complex battlefield environment. Furthermore, it is possible to use the deep learning model to conduct a comprehensive analysis of the battlefield situation, intelligently allocate weapon resources, avoid resource waste, and improve strike efficiency. In addition, various combat weapons can work together to maximize the overall defense effectiveness and achieve coordinated strikes with multiple weapons and multiple attack modes. On the other hand, by processing multi-dimensional and high-dimensional feature data through the model, it is possible to accurately identify and evaluate incoming targets, and improve the accuracy of decision-making.

[0064] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] The accompanying drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification are used to explain the principles of the present disclosure. Obviously, the accompanying drawings described below are only some embodiments of the present disclosure, and for ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without creative work.

[0066] Figure 1 A schematic diagram showing a flow chart of a combat decision-making method of a combat entity in an embodiment of the present disclosure;

[0067] Figure 2 A flowchart of inputting the overall combat effectiveness parameter and the characteristic parameter into the trained combat decision model and determining the combat strategy for M incoming targets through the trained combat decision model is shown in the embodiment of the present disclosure;

[0068] Figure 3 A schematic diagram showing how to encode the overall combat effectiveness parameters through the encoding layer to obtain the encoding matrix in the embodiment of the present disclosure;

[0069] Figure 4 A flow chart showing how to perform multi-head attention allocation on each row vector through the first multi-head self-attention layer to obtain each attention output matrix in an embodiment of the present disclosure;

[0070] Figure 5A flowchart showing how to perform attention allocation processing on each row vector according to the key weight matrix, value weight matrix and query weight matrix of each attention head to obtain an output matrix corresponding to each attention head in an embodiment of the present disclosure;

[0071] Figure 6 A schematic diagram of the network architecture of the coding layer in an embodiment of the present disclosure is shown;

[0072] Figure 7 A flowchart showing how to determine a combat strategy for M incoming targets based on a coding matrix and characteristic parameters through a decision layer in an embodiment of the present disclosure;

[0073] Figure 8 A flow chart showing how to perform multi-head attention allocation on the encoding matrix and the second attention output matrix through a multi-head attention layer in an embodiment of the present disclosure to obtain a third attention output matrix;

[0074] Fig. 9 A schematic diagram of the network architecture of the decision-making layer in an embodiment of the present disclosure is shown;

[0075] Fig.10 A schematic diagram showing the structure of a combat decision-making device of a combat entity in an exemplary embodiment of the present disclosure;

[0076] Fig.11 A schematic structural diagram of an electronic device in an exemplary embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0077] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as being limited to the examples set forth herein; on the contrary, these embodiments are provided so that the present disclosure will be more comprehensive and complete, and the concepts of the example embodiments are fully conveyed to those skilled in the art. The described features, structures, or characteristics may be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure may be practiced while omitting one or more of the specific details, or other methods, components, devices, steps, etc. may be adopted. In other cases, known technical solutions are not shown or described in detail to avoid obscuring various aspects of the present disclosure.

[0078] The terms "a", "an", "the" and "said" are used in this specification to indicate the presence of one or more elements / components / etc.; the terms "including" and "having" are used to express an open-ended inclusion and mean that additional elements / components / etc. may exist in addition to the listed elements / components / etc.; the terms "first" and "second" etc. are used only as labels and are not intended to limit the quantity of their objects.

[0079] In addition, the accompanying drawings are only schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings represent the same or similar parts, and their repeated descriptions will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities.

[0080] At present, the commonly used technical solutions in relevant combat decisions mainly include the following:

[0081] 1. Rule-based decision-making method: using pre-set rules and logic to determine how to use weapons to deal with incoming targets. For example, deciding which type of weapon to use depends on the distance and speed of the target.

[0082] The advantages of this solution are simple implementation, easy deployment and maintenance. The disadvantages are that it cannot dynamically adapt to changes in the battlefield environment and has poor decision-making flexibility.

[0083] (ii) Expert System: Expert System (ES) refers to the use of computer models of human expert reasoning to deal with complex problems in the real world that require expert explanations and reach the same conclusions as experts. Through expert knowledge bases and reasoning mechanisms, the decision-making process of human experts is simulated to deal with threats.

[0084] The advantage of this solution is that it can handle more complex decision-making scenarios and has a certain degree of intelligence. The disadvantage is that it relies on the completeness and accuracy of expert knowledge, has difficulty in dealing with unknown threats, and is difficult to update and maintain the knowledge base.

[0085] (III) Resource allocation algorithm based on optimization algorithm: Calculate the optimal resource allocation plan of the weapon in the time domain, space domain, and frequency domain through optimization algorithms (such as linear programming, dynamic programming, etc.) to ensure the maximum efficiency of the ship's defense weapon system.

[0086] The advantage of this solution is that it can optimize resource allocation to a certain extent and improve the overall efficiency of the ship's defense weapon system. The disadvantage is that the calculation complexity is high and it is difficult to respond to the dynamically changing battlefield environment in real time.

[0087] Overall, the above methods are effective when facing simple and foreseeable threats, but their limitations gradually become apparent when dealing with complex battlefield environments with multiple targets and dynamic changes, as shown in the following:

[0088] Slow response: Pre-set strategies cannot adapt to changes in the battlefield environment in a timely manner, resulting in delayed decision-making;

[0089] Unreasonable resource allocation: Failure to dynamically adjust weapon resources according to actual conditions may result in certain threats being ignored or resources being wasted;

[0090] Lack of coordinated combat capability: There is a lack of effective coordination mechanism between the various weapons, and the overall defense capability cannot be maximized.

[0091] In the embodiments of the present disclosure, a combat decision-making method for a combat entity is first provided, which at least to a certain extent overcomes the defect of the related technology that it cannot cope with a complex battlefield environment with multiple targets and dynamic changes.

[0092] Figure 1 A flow chart of a combat decision-making method of a combat entity in an embodiment of the present disclosure is shown. The execution subject of the combat decision-making method of the combat entity may be a server that makes combat decisions.

[0093] refer to Figure 1 According to an embodiment of the present disclosure, a combat decision-making method of a combat entity includes the following steps:

[0094] Step S110, obtaining characteristic parameters of M incoming targets of the combat entity at the current moment, and obtaining overall combat effectiveness parameters of the combat entity at the current moment for the M incoming targets;

[0095] Step S120, inputting the overall combat effectiveness parameter and the characteristic parameter into the trained combat decision model, and determining the combat strategy for the M incoming targets through the trained combat decision model;

[0096] Among them, the combat strategy is used to characterize the attack probability of each combat weapon against each incoming target, and M and N are both integers greater than 1.

[0097] exist Figure 1 In the technical solution provided by the illustrated embodiment, on the one hand, by training the combat decision model, it is possible to intelligently respond to multi-target, multi-dimensional complex environments through the model, and make decisions quickly, greatly shortening the decision-making time, improving the real-time response capability, and dynamically adapting to the complex battlefield environment. Furthermore, it is possible to use the deep learning model to comprehensively analyze the battlefield situation, intelligently allocate weapon resources, avoid waste of resources, and improve strike efficiency. In addition, various combat weapons can work together to maximize the overall defense effectiveness and achieve coordinated strikes with multiple weapons and multiple attack modes. On the other hand, by processing multi-dimensional and high-dimensional feature data through the model, it is possible to accurately identify and evaluate incoming targets, and improve the accuracy of decision-making.

[0098] The following Figure 1 The specific implementation process of each step is described in detail:

[0099] Before step S110, it should be noted that the combat entity in the present disclosure refers to a unit or system that can perform combat missions. For example, it can be a ship, a fighter, a bomber, a tank, an infantry fighting vehicle, etc., which can be set according to actual conditions. The present disclosure does not make any special limitations on this. The following embodiments are described using the combat entity as a ship as an example.

[0100] In an exemplary embodiment of the present disclosure, the space surrounding a ship can be divided into multiple defense areas (also referred to as defense sectors), and the multiple defense sectors can completely cover the entire airspace outside the ship, and there is a certain overlapping angle between two adjacent defense sectors.

[0101] The following is a description of how to define each defense sector in the present disclosure:

[0102] First, establish the global NED coordinate system (North East Down, also known as the navigation coordinate system): the center of the ship is the origin O(0,0,0), the N axis is forward along the ship's heading, the E axis is perpendicular to the N axis and points to the right, and the D axis points to the center of the earth. Let the length and width of the ship be L and W respectively.

[0103] Second, establish a local polar coordinate system for each corner:

[0104] Assume that the four corners of the ship are set to A, B, C and D respectively, then the coordinates of point A are (-L / 2, -W / 2), the coordinates of point B are (L / 2, -W / 2), the coordinates of point C are (L / 2, W / 2), and the coordinates of point D are (-L / 2, W / 2).

[0105] Third, establish a local polar coordinate system for each corner:

[0106] A local polar coordinate system is established at each corner of A, B, C, and D, with each corner as the origin. A point in each local polar coordinate system is represented by (r, φ), where r is the distance to the origin (a fixed point at the center of the coordinate system, i.e., any of ABCD above), and φ is the polar angle (measured from the polar axis to the line from the pole to the point).

[0107] Fourth, define the defense sector:

[0108] The defense sectors at each corner of A, B, C, and D have a central angle θ (that is, the angle of each sector's defense), and there is an overlap angle Δθ between adjacent sectors. Assume that the defense sector at point A has an angle range of: The minimum radius of each defense sector is denoted as r min , the maximum radius is represented by r max.

[0109] After defining the above multiple defense sectors, assume that the position of an incoming target is expressed as (N t ,E t ), then the defense sector in which the incoming target is located can be determined based on the following steps:

[0110] First, transform the global NED coordinates into the local rectangular coordinate system:

[0111] For the target position (N t ,E t ), it is necessary to convert it from the global NED coordinate system to the local rectangular coordinate system of each corner. Take point A as an example, the local rectangular coordinates of point A are:

[0112] Second, calculate the coordinates in the local rectangular coordinate system:

[0113] Assume that the coordinates of the target position in the local rectangular coordinate system of point A are (x A ,y A ), the conversion formula is as follows:

[0114] Third, calculate the polar angle φ t :

[0115] Calculate the polar angle φ in the local rectangular coordinate system t :

[0116] Fourth, determine whether the target location is within a certain defense sector:

[0117] Check the polar angle φ of the target position t Is it within the angle range of the defense sector? Also check if the target location is within the radius of the defense sector: If r min ≤r t ≤r max , the target position is within the defense sector.

[0118] It should be noted that the present invention can deploy combat weapons in each defense sector. Exemplarily, the combat weapons can include four types. Exemplarily, the types of combat weapons can include missiles, cannons, decoys, and electronic jamming (ew), among which missiles and cannons are hard weapons, and decoys and electronic jamming are soft weapons.

[0119] More specifically, the missile can include two attack modes (also called two guidance modes), which can be radar mode and infrared mode; the machine gun can include two attack modes (also called two auxiliary aiming modes), which can be radar mode and infrared mode; the decoy bomb can include three attack modes (also called three interference modes), which can be active infrared decoy bomb, outboard active decoy and passive infrared jammer; the electronic jamming can include two attack modes (also called two interference modes), which can be passive chaff interference and active noise interference.

[0120] Exemplarily, the above four types of combat weapons can be deployed in each defense sector. Given that each type of combat weapon can include multiple attack modes, when determining the specific number of combat weapons deployed on the ship, it is necessary to jointly determine the number of types of combat weapons, the number of combat weapons, and the number of attack modes associated with each combat weapon. For example, assuming that two types of combat weapons (missiles and machine guns) are deployed in a certain defense sector, one of each type of combat weapon is deployed, and the missile includes two attack modes, and the machine gun also includes two attack modes, then the number of combat weapons deployed in the defense sector can be 1*2+1*2=4. After calculating the number of combat weapons deployed in each defense sector one by one, the numbers of combat weapons corresponding to the above multiple defense sectors can be accumulated to determine the specific number N of combat weapons deployed on the ship.

[0121] In view that combat weapons may be located in different defense sectors and contain different attack modes, different combat weapons can be distinguished according to the identification of the defense sector and the identification of the attack mode. Specifically, the weapon configuration on the ship can refer to Table 1:

[0122] Table 1

[0123]

[0124]

[0125] The following describes the specific implementation of how to intelligently make combat decisions in the present disclosure in combination with steps S110 to S120:

[0126] In step S110, characteristic parameters of the M incoming targets of the combat entity at the current moment are obtained, and the overall combat effectiveness parameters of the combat entity at the current moment against the M incoming targets are obtained.

[0127] In this step, characteristic parameters of M (M is an integer greater than 1) incoming targets of the combat entity at the current moment can be obtained, as well as the overall combat effectiveness parameters of the combat entity at the current moment against the M incoming targets.

[0128] Among them, the characteristic parameters of each incoming target can be composed of a set of attributes describing its current state (i.e., characteristic variables of multiple dimensions). These attributes are combined into a vector, which includes the distance, azimuth, pitch, speed, speed (moving direction relative to a certain reference direction), x, y, z (x, y, z correspond to the N-axis coordinates of the NED coordinate system, the E-axis coordinates of the NED coordinate system, and the D-axis coordinates of the NED coordinate system), Vx, Vy, Vz (Vx, Vy, Vz correspond to the N-axis speed of the NED coordinate system, the E-axis speed of the NED coordinate system, and the D-axis speed of the NED coordinate system), absolute heading, absolute speed, etc., as well as electronic reconnaissance parameters such as threat level, credibility, etc. These dimensions include sufficient information about the incoming target itself to support comprehensive evaluation and decision-making of the incoming target.

[0129] Exemplarily, the dimension of the above characteristic parameter may be G. Referring to Table 2, Table 2 shows the specific contents of the dimension included in the above characteristic parameter:

[0130] Table 2

[0131]

[0132]

[0133] The overall combat effectiveness parameter of the combat entity against M incoming targets at the current moment can be called the overall combat effectiveness parameter. The overall combat effectiveness parameter is composed of the combat effectiveness parameter of each combat weapon against each incoming target at the current moment, and the combat effectiveness parameter of each combat weapon against each incoming target at the current moment can include parameters from two aspects, namely, the expected combat effect characterization parameter of each combat weapon against each incoming target and the performance parameter of each combat weapon at the current moment.

[0134] Specifically, referring to Table 3, the combat effectiveness parameters of each combat weapon against each incoming target at the current moment may include the following L (L is an integer greater than 1) dimensions:

[0135] Table 3

[0136]

[0137]

[0138]

[0139] It should be noted that in the above Table 3, Corresponding to the performance parameters of each of the above-mentioned combat weapons at the current moment, the rest correspond to the parameters characterizing the expected combat effects of each of the above-mentioned combat weapons against each incoming target.

[0140] Given that the combat effectiveness parameters of each combat weapon against each incoming target at the current moment correspond to L dimensions, while the total number of combat weapons is N, and the total number of incoming targets is M, the above-mentioned overall combat effectiveness parameters correspond to N*M*L dimensions, that is, the above-mentioned overall combat effectiveness parameters are composed of M N*L first matrices.

[0141] In step S120, the overall combat effectiveness parameters and characteristic parameters are input into the trained combat decision model, and the combat strategy for the M incoming targets is determined by the trained combat decision model.

[0142] In this step, after obtaining the above-mentioned overall combat effectiveness parameters and characteristic parameters, the above-mentioned overall combat effectiveness parameters and characteristic parameters can be input into the trained combat decision model, and the combat strategy for M incoming targets can be decided by the trained combat decision model.

[0143] Specifically, the trained combat decision model may include a strategy network, which may include a coding layer and a decision layer. Exemplarily, the coding layer and the decision layer may adopt the ReactFormer architecture. ReactFormer is an innovative decision network designed for weapon decision systems, combining the latest deep learning technology and reinforcement learning framework.

[0144] You can refer to Figure 2 , Figure 2 A flowchart of inputting the overall combat effectiveness parameter and the characteristic parameter into the trained combat decision model in the disclosed embodiment is shown, and the combat strategy for M incoming targets is determined by the trained combat decision model, including steps S201-S202:

[0145] In step S201, the overall combat effectiveness parameter is input into the coding layer, and the overall combat effectiveness parameter is coded by the coding layer to obtain a coding matrix.

[0146] In this step, the encoding layer may include a preprocessing layer, a feature extraction layer, a first multi-head self-attention layer, and a first feedforward network. Figure 3 , Figure 3 A flow chart showing how to encode the overall combat effectiveness parameters through the encoding layer to obtain the encoding matrix in the embodiment of the present disclosure includes steps S301 to S304:

[0147] In step S301, the overall combat effectiveness parameters are preprocessed through a preprocessing layer to obtain an initial matrix.

[0148] In this step, the purpose of the preprocessing layer is to reorganize the data structure to make it more suitable for the subsequent processing flow. Referring to the relevant explanation of the above step S110, it can be seen that the above overall combat effectiveness parameter corresponds to N*M*L dimensions, so the preprocessing layer can stack the above N*M*L dimensional overall combat effectiveness parameters according to M dimensions, thereby obtaining M N*L feature graphs, that is, for each incoming target, there is an N*L dimensional feature graph associated with it, so that the feature graphs of each incoming target are independent of each other during the processing process, which is convenient for individual analysis and decision-making.

[0149] In step S302, feature extraction is performed on the initial matrix through a feature extraction layer to obtain a word embedding matrix.

[0150] In this step, the feature extraction layer can be a bitwise convolutional network, so that the initial matrix can be subjected to feature extraction processing by the bitwise convolutional network to obtain a word embedding matrix (the word embedding matrix consists of M row vectors). The "bitwise convolution" here refers to applying the same set of convolution processing to the feature map of each incoming target (i.e., the N*L matrix at each position), and these target feature maps are independent of each other during processing. This method ensures independent evaluation of each target, while allowing the network to capture the key features of the relationship between the weapon and the target and extract more advanced abstract features.

[0151] Exemplarily, the above bitwise convolutional network can be configured as follows:

[0152] Convolutional layer configuration:

[0153] Convolution kernel size: 3x3, usually a smaller convolution kernel can capture local features;

[0154] Stride: 1, retain details without loss;

[0155] Padding: 1 (if using a 3x3 convolution kernel), to ensure that the output feature map size remains unchanged;

[0156] Activation function: ReLU, providing nonlinear processing capabilities;

[0157] Number of convolutional layers: 2-3 layers, enough to extract key features;

[0158] Number of output channels: Starting from L, it can be doubled layer by layer and return to 1 at the last layer, for example, 2 in the first layer, 4 in the second layer, and 1 in the third layer.

[0159] Pooling layer (optional):

[0160] Type: max pooling or average pooling;

[0161] Size: 2x2, halving the feature map size;

[0162] Location: Pooling is usually applied after several consecutive convolutional layers to reduce the number of parameters and increase the level of abstraction.

[0163] Therefore, referring to the relevant explanations of the above steps, the initial matrix is ​​the N*L feature map corresponding to each of the M incoming targets. After being input into the above-mentioned bitwise convolutional network, each N*L feature map can be convolved separately. In this way, the network can generate a feature response map for each incoming target, with the aim of extracting the key features of the relationship between the combat weapon and the incoming target. This can be achieved through multiple convolutional layers and activation layers to gradually extract more advanced abstract features.

[0164] Afterwards, each feature map obtained after the convolution operation can be flattened into a one-dimensional vector, so that M row vectors can be obtained.

[0165] In step S303, multi-head attention is allocated to each row vector through the first multi-head self-attention layer to obtain each attention output matrix.

[0166] In this step, the first multi-head attention layer is used to further analyze and process the features from the bitwise convolutional layer. The main purpose of the self-attention mechanism is to analyze the mutual influence and dependency between each target feature vector and other vectors in the sequence (i.e. other targets), which helps the model understand the possible connections between different incoming targets, as well as the priority and adaptability of each combat weapon to these incoming targets.

[0167] Specifically, the first multi-head self-attention layer may include multiple attention heads, each of which focuses on capturing different aspects of target features. For example, one head may focus on the distance information of the incoming target, while another head may focus on the type or motion characteristics of the target. Exemplarily, the multiple attention heads may be set to 8 heads, which can be set according to actual conditions, and the present disclosure does not make any special limitation on this.

[0168] refer to Figure 4 , Figure 4 A flow chart showing how to perform multi-head attention allocation on each row vector through the first multi-head self-attention layer in the embodiment of the present disclosure to obtain each attention output matrix includes steps S401-S402:

[0169] In step S401, for each attention head, attention allocation processing is performed on each row vector according to the key weight matrix, value weight matrix and query weight matrix of each attention head to obtain the output matrix corresponding to each attention head.

[0170] In this step, refer to Figure 5 , Figure 5A flowchart showing how to perform attention allocation processing on each row vector according to the key weight matrix, value weight matrix and query weight matrix of each attention head to obtain the output matrix corresponding to each attention head in the embodiment of the present disclosure includes steps S501 to S504:

[0171] In step S501, for each attention head, each row vector is multiplied by the query weight matrix to obtain a query vector.

[0172] In this step, for each attention head, each row vector can be compared with the query weight matrix W Q Perform multiplication to obtain the query vector Q.

[0173] In step S502, each row vector is multiplied by the key weight matrix to obtain a key vector.

[0174] In this step, each row vector can be combined with the key weight matrix W K Perform multiplication to obtain the key vector K.

[0175] In step S503, each row vector is multiplied by the value weight matrix to obtain a value vector.

[0176] In this step, each row vector can be compared with the value weight matrix W V Perform multiplication to obtain the value vector V.

[0177] In step S504, based on the query vector, the key vector and the value vector, the output matrix corresponding to each attention head is determined.

[0178] In this step, the transposed vector K of the key vector can be obtained T , and perform a dot product operation on the query vector and the transposed vector of the key vector to obtain the operation result (i.e. QK T ); Normalize the operation result to obtain the attention score (which can be expressed as softmax(QK T )); Multiply the attention score by the value vector V to obtain the output matrix corresponding to each attention head.

[0179] Next reference Figure 4 In step S402, each attention output matrix is ​​determined according to the multiple output matrices corresponding to the multiple attention heads.

[0180] In this step, the multiple output matrices corresponding to the above-mentioned multiple attention heads can be concatenated to determine each attention output matrix.

[0181] Next reference Figure 3In step S304, the M attention output matrices are optimized through the first feedforward network to obtain a coding matrix.

[0182] In this step, the M attention output matrices mentioned above can be optimized through a first feedforward network to obtain a coding matrix, which is composed of M first matrices corresponding to the M attention output matrices mentioned above.

[0183] Among them, the first feedforward network is used to further refine features and enhance the stability of the model through residual connection and layer normalization.

[0184] Specifically, the first feedforward network may include multiple linear layers with activation functions such as ReLU in the middle to enhance the nonlinear processing capability of the model.

[0185] Exemplarily, the first feedforward network may refer to the following specific configuration:

[0186] Layer: Two layers of linear transformation, with ReLU activation in the middle;

[0187] Internal dimension: usually 4 times the input dimension, for example, if the input dimension is 64, the internal dimension is 256;

[0188] Output dimension: The same as the input dimension, maintaining the consistency of feature information.

[0189] After optimizing the above M attention output matrices through the first feedforward network, M first matrices can be obtained.

[0190] refer to Figure 6 , Figure 6 A schematic diagram of the network architecture of the coding layer in the embodiment of the present disclosure is shown as follows: Figure 6 As shown:

[0191] The input of the encoding layer is the overall combat effectiveness parameter of N*M*L dimension. The preprocessing layer can perform position encoding on the above overall combat effectiveness parameter to obtain an initial matrix. The feature extraction layer can perform feature extraction on the above initial matrix to obtain a word embedding matrix (composed of M row vectors). After that, the first multi-head attention mechanism can perform multi-head attention allocation on each of the above row vectors to obtain each attention output matrix. After that, each attention output matrix can be residually connected and normalized. After the first feedforward network, each attention output matrix can be optimized. After another residual connection and normalization process, the final output encoding matrix can be obtained.

[0192] Next reference Figure 2 In step S202, the decision layer decides on the combat strategy for the M incoming targets based on the coding matrix and the feature parameters.

[0193] In this step, the decision layer expands the multi-head attention mechanism of the traditional Transformer, and each "head" corresponds to a combat weapon. It directly processes the sequence of all dynamic incoming targets in the current scene at the current moment of interaction with the environment. This enables ReactFormer to respond to environmental changes in real time, optimize tactical decisions, and improve defense efficiency and the overall response speed of the system. In the decision-making at each moment of interaction with the environment, the decision layer generates a priority attack probability vector for each weapon system. This vector describes the attack priority of the combat weapon to all incoming targets in the current scene. The vectors of all combat weapons constitute a priority attack probability matrix, which reflects the attack priority of each combat weapon to each incoming target.

[0194] Specifically, the above-mentioned decision layer may include a feature transformation layer, a second multi-head self-attention layer, a multi-head attention layer, a second feedforward network and an output layer.

[0195] refer to Figure 7 , Figure 7 A flowchart showing how to determine a combat strategy for M incoming targets based on a coding matrix and characteristic parameters through a decision layer in an embodiment of the present disclosure includes steps S701 to S704:

[0196] In step S701, feature transformation processing is performed on feature parameters through a feature transformation layer to obtain a transformation matrix.

[0197] In this step, similar to Transformer, position encoding is added to enable the model to understand the order of the input sequence. Since the input is based on real-time target data, position encoding helps the model capture the relative position of the target in the sequence.

[0198] Therefore, the feature transformation layer can use the same sine and cosine function transformation as the Transformer to process the above feature parameters (including G dimensions) to obtain a transformation matrix.

[0199] In step S702, multi-head self-attention allocation is performed on the transformation matrix through the second multi-head self-attention layer to obtain a second attention output matrix.

[0200] In this step, the role of the second multi-head self-attention layer is to process the features of each incoming target so that the representation of each target can integrate the information of other targets.

[0201] Specifically, the second multi-head self-attention layer can refer to the following configuration:

[0202] Number of targets: depends on the total number of incoming targets M, for example, set to 30;

[0203] Model dimension: consistent with feature dimension G;

[0204] Key / Query / Value vector dimension: usually G / number of heads, i.e. G / 30;

[0205] Dropout rate: Usually set to 0.1 to prevent overfitting.

[0206] Specifically, you can refer to the relevant explanation of the above step S303, and perform multi-head self-attention allocation on the transformation matrix through the second multi-head self-attention layer to obtain the second attention output matrix. It will not be repeated here. It should be noted that the second attention output matrix contains M second matrices corresponding to the M attention heads.

[0207] In step S703, multi-head attention allocation is performed on the encoding matrix and the second attention output matrix by each attention head of the multi-head attention layer to obtain each third attention output matrix.

[0208] In this step, multi-head attention allocation can be performed on the above-mentioned encoding matrix and the above-mentioned second attention output matrix through a multi-head attention layer to obtain a third attention output matrix.

[0209] It should be noted that the multi-head attention layer can refer to the following configuration:

[0210] Number of heads: Keep it consistent with the total number of combat weapons N (i.e. set N weapon heads), for example, set it to 20;

[0211] Key / Query / Value vector dimensions: set to G / number of heads;

[0212] Dropout rate: Also set to 0.1.

[0213] Specifically, given that the encoding matrix includes the first matrices corresponding to the M incoming targets, and the second attention output matrix includes the second matrices corresponding to the M incoming targets, the first matrix and the second matrix corresponding to each incoming target can be regarded as a set of data, and then, reference can be made to Figure 8 , Figure 8 A flowchart showing how to perform multi-head attention allocation on the encoding matrix and the second attention output matrix through each attention head of the multi-head attention layer in the embodiment of the present disclosure to obtain each third attention output matrix includes steps S801 to S805:

[0214] In step S801, each first matrix is ​​multiplied by a preset key weight matrix to obtain each designated key vector.

[0215] In this step, the first matrix corresponding to each incoming target is multiplied by the preset key weight matrix (for example: Wk1 ), obtain each specified key vector, thereby, M specified key vectors (for example: K1) can be obtained.

[0216] It should be noted that the preset key weight matrices corresponding to each attention head in the present disclosure are independent and unrelated to each other.

[0217] In step S802, each first matrix is ​​multiplied by a preset value weight matrix to obtain each specified value vector.

[0218] In this step, the first matrix corresponding to each incoming target can be multiplied by a preset weight matrix (for example: W V1 ), obtain each specified value vector (for example: V1), thereby, M specified value vectors can be obtained.

[0219] It should be noted that the preset value weight matrix corresponding to each attention head in the present disclosure is independent and unrelated. In step S803, each second matrix is ​​multiplied by the preset query weight matrix to obtain each specified query vector.

[0220] In this step, the second matrix of each incoming target can be multiplied by a preset query weight matrix (for example: W D1 ), obtain each specified query vector (for example: Q1), thereby, M specified query vectors can be obtained.

[0221] It should be noted that the preset query weight matrices corresponding to each attention head in the present disclosure are independent and unrelated to each other.

[0222] In step S804, attention is allocated to each designated key vector, each designated value vector, and each designated query vector through a multi-head attention layer to determine each designated fusion vector.

[0223] In this step, the multi-head attention layer can be used to allocate attention to the designated key vector, designated value vector, and designated query vector corresponding to each incoming target to determine the designated fusion vector corresponding to each incoming target.

[0224] Exemplarily, the transposed vector K1 of the specified key vector can be obtained T , and perform a dot product operation on the query vector and the transposed vector of the key vector to obtain the operation result (i.e. Q1K1 T ); Normalize the operation result to obtain the attention score (which can be expressed as softmax(Q1K1 T )); Multiply the attention score and the value to V1 to obtain the specified fusion vector corresponding to each incoming target, which is a row vector of 1*G.

[0225] In step S805, each third attention output matrix is ​​determined according to the M specified fusion vectors.

[0226] In this step, the third attention output matrix is ​​a matrix containing the above-mentioned M specified fusion vectors (1*G).

[0227] It should be noted that after steps S801 to S805, what is obtained is the processing result corresponding to one attention head in the multi-head attention layer. One attention head corresponds to a third attention matrix composed of M specified fusion vectors. The length of each specified fusion vector is G. Therefore, each head corresponds to an M×G-dimensional third attention matrix.

[0228] Given that the total number of combat weapons is N, N attention heads are set, so that after traversing the N attention heads, N M×G dimensional matrices can be obtained.

[0229] Next reference Figure 7 In step S704, the N third attention output matrices are optimized through the second feedforward network to obtain a probability distribution matrix.

[0230] In this step, the second feedforward network is used to process the M×G dimensional matrix corresponding to each attention head (a total of N weapon heads) passing through the multi-head attention layer. The second feedforward network can refer to the following configuration:

[0231] First layer dimension: expanded to 4G, if G = 10, then 40;

[0232] Activation function: ReLU;

[0233] The second dimension: shrinks back to G;

[0234] Inter-layer Dropout rate: 0.1;

[0235] Residual connections and layer normalization: applied after each sub-layer.

[0236] Through the above-mentioned second feedforward network, N weapon heads can output N groups of M×G-dimensional initial probability distribution matrices accordingly.

[0237] Among them, the above-mentioned probability distribution matrix is ​​used to characterize the combat strategy against the M incoming targets, the number of rows of the probability distribution matrix is ​​equal to the total number N of the combat weapons deployed on the combat entity, and the number of columns of the probability distribution matrix is ​​equal to the total number M of the at least one incoming target; the probability distribution matrix is ​​used to describe the attack probability of each of the combat weapons against each of the incoming targets.

[0238] refer to Fig. 9 , Fig. 9A schematic diagram of the network architecture of the decision-making layer in the embodiment of the present disclosure is shown as follows: Fig. 9 As shown:

[0239] The feature parameters of the M incoming targets can first be transformed through the feature transformation layer to obtain a transformation matrix, and then the second multi-head attention layer can be used to perform multi-head self-attention allocation on the transformation matrix to obtain a second attention output matrix, and then the second attention output matrix can be residually connected and normalized, and each second matrix contained in the second attention output matrix can be multiplied by a preset query weight matrix to obtain a specified query vector;

[0240] For the M first matrices output by the coding layer, each first matrix may be multiplied by a preset key weight matrix to obtain each designated key vector, and each first matrix may be multiplied by a preset value weight matrix to obtain each designated value vector;

[0241] Afterwards, the designated query vector, designated key vector, and designated value vector corresponding to each incoming target can be input into the multi-head attention layer, and the attention can be distributed through the multi-head attention layer to obtain each of the M designated G×1-dimensional designated fusion vectors under the N weapon heads, so as to finally obtain N groups of M G×1-dimensional designated fusion vectors;

[0242] Furthermore, N groups of M G×1 dimensional specified fusion vectors can be residually connected and normalized, and optimized through the second feedforward network to decide on combat strategies for M incoming targets (N groups of M×G dimensional initial probability distribution matrices).

[0243] Based on the above technical solutions, the present disclosure has at least the following technical effects:

[0244] First, it can improve real-time response capabilities to dynamically adapt to the battlefield environment. Specifically, by training the combat decision-making model, it can effectively respond to multi-target, multi-dimensional complex threats through the model, and make decisions quickly, greatly shortening the decision-making time and improving the system's response speed;

[0245] Second, it can intelligently allocate weapon resources. Specifically, this technical solution uses a deep learning model to conduct a comprehensive analysis of the battlefield situation, intelligently allocate weapon resources, avoid resource waste, and improve strike efficiency. In addition, based on the ReactFormer model, each defense unit can work together to maximize the overall defense effectiveness and achieve coordinated strikes with multiple weapons and multiple attack modes.

[0246] Third, by processing multi-dimensional and high-dimensional feature data through deep learning technology, it is possible to accurately identify and evaluate incoming targets and improve the accuracy of decision-making.

[0247] The present disclosure also provides a combat decision-making device for a combat entity, wherein the surrounding space of the combat entity is divided into a plurality of defense areas, and N combat weapons are deployed on the combat entity, each of the combat weapons being deployed in at least one defense area on the combat entity. Fig.10 A schematic diagram showing the structure of a combat decision-making device of a combat entity in an exemplary embodiment of the present disclosure; Fig.10 As shown, the combat decision device 1000 of the combat entity may include a parameter acquisition module 1010 and a combat decision module 1020. Among them:

[0248] The parameter acquisition module 1010 is used to acquire characteristic parameters of the M incoming targets of the combat entity at the current moment, and to acquire the overall combat effectiveness parameters of the combat entity at the current moment for the M incoming targets;

[0249] A decision module 1020 is used to input the overall combat effectiveness parameter and the characteristic parameter into a trained combat decision model, and determine a combat strategy for the M incoming targets through the trained combat decision model;

[0250] The combat strategy is used to characterize the attack probability of each combat weapon against each incoming target, and M and N are both integers greater than 1.

[0251] In an exemplary embodiment of the present disclosure, the overall combat effectiveness parameter is composed of the combat effectiveness parameter of each of the combat weapons against each of the incoming targets at the current moment;

[0252] The combat effectiveness parameter of each of the combat weapons at the current moment for each of the incoming targets corresponds to L dimensions;

[0253] The overall combat effectiveness parameter corresponds to the N*M*L dimension;

[0254] Wherein, L is an integer greater than 1.

[0255] In an exemplary embodiment of the present disclosure, the combat effectiveness parameters of each of the combat weapons against each of the incoming targets at the current moment include:

[0256] Characterization parameters of the expected combat effect of each of the combat weapons against each of the incoming targets and the performance parameters of each of the combat weapons at the current moment.

[0257] In an exemplary embodiment of the present disclosure, the number N of the combat weapons is determined based on the number of types of the combat weapons, the number of each type of combat weapon, and the number of attack modes associated with each combat weapon.

[0258] In an exemplary embodiment of the present disclosure, the trained combat decision model includes a policy network, and the policy network includes a coding layer and a decision layer;

[0259] The combat decision module 1020 inputs the overall combat effectiveness parameter and the characteristic parameter into the trained combat decision model, and determines the combat strategy for the M incoming targets through the trained combat decision model, including:

[0260] Inputting the overall combat effectiveness parameter into the coding layer, encoding the overall combat effectiveness parameter through the coding layer to obtain a coding matrix;

[0261] The decision layer determines the combat strategy for the M incoming targets based on the coding matrix and the characteristic parameters.

[0262] In an exemplary embodiment of the present disclosure, the encoding layer includes a preprocessing layer, a feature extraction layer, a first multi-head self-attention layer, and a first feedforward network;

[0263] The combat decision module 1020 performs encoding processing on the overall combat effectiveness parameter through the encoding layer to obtain an encoding matrix, including:

[0264] Preprocessing the overall combat effectiveness parameters through the preprocessing layer to obtain an initial matrix;

[0265] Performing feature extraction processing on the initial matrix through the feature extraction layer to obtain a word embedding matrix; the word embedding matrix is ​​composed of M row vectors;

[0266] Perform multi-head attention allocation on each of the row vectors through the first multi-head self-attention layer to obtain each attention output matrix;

[0267] The M attention output matrices are optimized through the first feedforward network to obtain the encoding matrix; the encoding matrix is ​​composed of M first matrices corresponding to the M attention output matrices.

[0268] In an exemplary embodiment of the present disclosure, the first multi-head self-attention layer includes a plurality of attention heads;

[0269] The combat decision module 1020 performs multi-head attention allocation on each of the row vectors through the first multi-head self-attention layer to obtain each attention output matrix, including:

[0270] For each attention head, performing attention allocation processing on each of the row vectors according to the key weight matrix, the value weight matrix and the query weight matrix of each attention head to obtain an output matrix corresponding to each attention head;

[0271] Determine each attention output matrix according to the multiple output matrices corresponding to the multiple attention heads.

[0272] In an exemplary embodiment of the present disclosure, the combat decision module 1020 performs attention allocation processing on each row vector according to the key weight matrix, value weight matrix and query weight matrix of each attention head for each attention head, and obtains an output matrix corresponding to each attention head, including:

[0273] For each attention head, multiply each row vector by the query weight matrix to obtain a query vector;

[0274] Multiply each of the row vectors by the key weight matrix to obtain a key vector;

[0275] Multiply each of the row vectors by the value weight matrix to obtain a value vector;

[0276] An output matrix corresponding to each of the attention heads is determined based on the query vector, the key vector, and the value vector.

[0277] In an exemplary embodiment of the present disclosure, the combat decision module 1020 determines the output matrix corresponding to each attention head based on the query vector, the key vector and the value vector, including:

[0278] Obtaining a transposed vector of the key vector, and performing a dot product operation on the query vector and the transposed vector of the key vector to obtain an operation result;

[0279] Normalizing the operation result to obtain an attention score;

[0280] The attention score is multiplied by the value vector to obtain an output matrix corresponding to each attention head.

[0281] In an exemplary embodiment of the present disclosure, the decision layer includes a feature transformation layer, a second multi-head self-attention layer, a multi-head attention layer, and a second feedforward network;

[0282] The combat decision module 1020 outputs a combat strategy for the M incoming targets through the decision layer according to the coding matrix and the characteristic parameters, including:

[0283] Performing feature transformation processing on the feature parameters through the feature transformation layer to obtain a transformation matrix;

[0284] Performing multi-head self-attention allocation on the transformation matrix through the second multi-head self-attention layer to obtain a second attention output matrix; the second multi-head self-attention layer includes M attention heads, and the second attention output matrix includes M second matrices corresponding to the M attention heads;

[0285] Performing multi-head attention allocation on the encoding matrix and the second attention output matrix through each attention head of the multi-head attention layer to obtain each third attention output matrix; the number of attention heads in the multi-head attention layer is equal to N;

[0286] Optimizing the N third attention output matrices through the second feedforward network to obtain a probability distribution matrix;

[0287] The probability distribution matrix is ​​used to characterize the combat strategy against the M incoming targets, the number of rows of the probability distribution matrix is ​​equal to the total number N of the combat weapons deployed on the combat entity, and the number of columns of the probability distribution matrix is ​​equal to the total number M of the at least one incoming target; the probability distribution matrix is ​​used to describe the attack probability of each of the combat weapons against each of the incoming targets.

[0288] In an exemplary embodiment of the present disclosure, the encoding matrix includes M first matrices, and the second attention output matrix includes M second matrices;

[0289] The combat decision module 1020 performs multi-head attention allocation on the encoding matrix and the second attention output matrix through each attention head of the multi-head attention layer to obtain each third attention output matrix, including:

[0290] Multiply each of the first matrices by a preset key weight matrix to obtain each specified key vector;

[0291] Multiply each of the first matrices by a preset value weight matrix to obtain each specified value vector;

[0292] Multiply each of the second matrices by a preset query weight matrix to obtain each specified query vector;

[0293] Attention is allocated to each of the specified key vectors, each of the specified value vectors, and each of the specified query vectors by the multi-head attention layer to determine each specified fusion vector;

[0294] Determine each of the third attention output matrices based on the M specified fusion vectors.

[0295] The specific details of each module in the combat decision-making device of the above-mentioned combat entity have been described in detail in the combat decision-making method of the corresponding combat entity, so they will not be repeated here.

[0296] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into multiple modules or units to be embodied.

[0297] In addition, although the steps of the method in the present disclosure are described in a specific order in the drawings, this does not require or imply that the steps must be performed in this specific order, or that all the steps shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps, etc.

[0298] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, a mobile terminal, or a network device, etc.) to execute the method according to the implementation of the present disclosure.

[0299] The present disclosure also provides a computer-readable storage medium, which may be included in the electronic device described in the above embodiments; or may exist independently without being assembled into the electronic device.

[0300] The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0301] Computer-readable storage media can send, propagate or transmit programs for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable storage medium can be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0302] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by an electronic device, the electronic device implements the method described in the above embodiments.

[0303] In addition, an electronic device capable of implementing the above method is also provided in an embodiment of the present disclosure.

[0304] Those skilled in the art will appreciate that various aspects of the present disclosure may be implemented as systems, methods or program products. Therefore, various aspects of the present disclosure may be specifically implemented in the following forms, namely: complete hardware implementation, complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software, which may be collectively referred to herein as "circuits", "modules" or "systems".

[0305] Refer to the following Fig.11 1100 according to this embodiment of the present disclosure is described. Fig.11 The electronic device 1100 shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0306] like Fig.11 As shown, the electronic device 1100 is in the form of a general computing device. The components of the electronic device 1100 may include, but are not limited to: at least one processor 1110, at least one memory 1120, a bus 1130 connecting different system components (including the memory 1120 and the processor 1110), and a display 1140.

[0307] The memory stores program codes, which can be executed by the processor 1110, so that the processor 1110 executes the steps according to various exemplary embodiments of the present disclosure described in the above “Exemplary Method” section of this specification. For example, the processor 1110 can execute the following steps: Figure 1As shown in: Step S110, obtaining characteristic parameters of the M incoming targets of the combat entity at the current moment, and obtaining the overall combat effectiveness parameters of the combat entity at the current moment for the M incoming targets; Step S120, inputting the overall combat effectiveness parameters and the characteristic parameters into the trained combat decision model, and deciding the combat strategy for the M incoming targets through the trained combat decision model; wherein the combat strategy is used to characterize the attack probability of each combat weapon for each incoming target, and M and N are both integers greater than 1.

[0308] The memory 1120 may include a readable medium in the form of a volatile storage, such as a random access memory (RAM) 11201 and / or a cache memory 11202 , and may further include a read-only memory (ROM) 11203 .

[0309] Memory 1120 may also include a program / utility 11204 having a set (at least one) of program modules 11205, such program modules 11205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0310] The bus 1130 may represent one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using any of a variety of bus architectures.

[0311] The electronic device 1100 may also communicate with one or more external devices 1200 (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 1100, and / or communicate with any device that enables the electronic device 1100 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed via an input / output (I / O) interface 1150. Furthermore, the electronic device 1100 may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via a network adapter 1160. As shown, the network adapter 1160 communicates with other modules of the electronic device 1100 via a bus 1130. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 1100, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0312] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. The present disclosure is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary technical means in the art that are not disclosed in the present disclosure. The description and examples are to be considered exemplary only, and the true scope and spirit of the present disclosure are indicated by the claims.

Claims

1. A combat decision-making method for a combat entity, characterized in that: The space around the combat entity is divided into a plurality of defense areas, N combat weapons are deployed on the combat entity, and each combat weapon is deployed in at least one defense area on the combat entity. The method includes: Acquire characteristic parameters of M incoming targets of the combat entity at the current moment, and acquire overall combat effectiveness parameters of the combat entity at the current moment for the M incoming targets; Inputting the overall combat effectiveness parameter and the characteristic parameter into a trained combat decision model, and determining a combat strategy for the M incoming targets through the trained combat decision model; The combat strategy is used to characterize the attack probability of each combat weapon against each incoming target, and M and N are both integers greater than 1.

2. The method according to claim 1, characterized in that: The overall combat effectiveness parameter is composed of the combat effectiveness parameter of each combat weapon against each incoming target at the current moment; The combat effectiveness parameter of each of the combat weapons at the current moment for each of the incoming targets corresponds to L dimensions; The overall combat effectiveness parameter corresponds to the N*M*L dimension; Wherein, L is an integer greater than 1.

3. The method according to claim 2, characterized in that The combat effectiveness parameters of each combat weapon against each incoming target at the current moment include: Characterization parameters of the expected combat effect of each of the combat weapons against each of the incoming targets and the performance parameters of each of the combat weapons at the current moment.

4. The method according to claim 1, characterized in that The number N of the combat weapons is determined based on the number of types of the combat weapons, the number of each type of combat weapon, and the number of attack modes associated with each combat weapon.

5. The method according to claim 1, characterized in that The trained combat decision model includes a strategy network, and the strategy network includes a coding layer and a decision layer; The step of inputting the overall combat effectiveness parameter and the characteristic parameter into a trained combat decision model, and determining a combat strategy for the M incoming targets through the trained combat decision model, comprises: Inputting the overall combat effectiveness parameter into the coding layer, encoding the overall combat effectiveness parameter through the coding layer to obtain a coding matrix; The decision layer determines the combat strategy for the M incoming targets based on the coding matrix and the characteristic parameters.

6. The method according to claim 5, characterized in that The encoding layer includes a preprocessing layer, a feature extraction layer, a first multi-head self-attention layer and a first feedforward network; The encoding process of the overall combat effectiveness parameter through the encoding layer to obtain the encoding matrix includes: Preprocessing the overall combat effectiveness parameters through the preprocessing layer to obtain an initial matrix; Performing feature extraction processing on the initial matrix through the feature extraction layer to obtain a word embedding matrix; the word embedding matrix is ​​composed of M row vectors; Perform multi-head attention allocation on each of the row vectors through the first multi-head self-attention layer to obtain each attention output matrix; The M attention output matrices are optimized through the first feedforward network to obtain the encoding matrix; the encoding matrix is ​​composed of M first matrices corresponding to the M attention output matrices.

7. The method according to claim 6, characterized in that The first multi-head self-attention layer includes multiple attention heads; The performing multi-head attention allocation on each of the row vectors through the first multi-head self-attention layer to obtain each attention output matrix includes: For each attention head, performing attention allocation processing on each of the row vectors according to the key weight matrix, the value weight matrix and the query weight matrix of each attention head to obtain an output matrix corresponding to each attention head; Determine each attention output matrix according to the multiple output matrices corresponding to the multiple attention heads.

8. A combat decision-making device of a combat entity, characterized in that: The space around the combat entity is divided into a plurality of defense areas. N combat weapons are deployed on the combat entity. Each combat weapon is deployed in at least one defense area on the combat entity. The device includes: A parameter acquisition module, used to acquire characteristic parameters of the M incoming targets of the combat entity at the current moment, and to acquire overall combat effectiveness parameters of the combat entity at the current moment for the M incoming targets; A decision module, used for inputting the overall combat effectiveness parameter and the characteristic parameter into a trained combat decision model, and determining a combat strategy for the M incoming targets through the trained combat decision model; The combat strategy is used to characterize the attack probability of each combat weapon against each incoming target, and M and N are both integers greater than 1.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the combat decision-making method of the combat entity described in any one of claims 1 to 7 is implemented.

10. An electronic device, characterized in that: include: processor; as well as A memory, configured to store executable instructions of the processor; Wherein, the processor is configured to execute the combat decision-making method of the combat entity described in any one of claims 1 to 7 by executing the executable instructions.

Images