Air maneuvering situation information estimation method and device based on multi-sample hierarchical clustering

Abstract

The invention relates to an air maneuvering situation information estimation method and device based on multi-sample hierarchical clustering. The method comprises the following steps of: clustering a sample sequence by adopting a multi-layer clustering method under the condition that a current hierarchical clustering value is given and the time of an instruction of the air maneuvering aircraft is used as a classification standard to obtain a plurality of state vectors; generating a corresponding situation condition node according to the state vector; judging the sample sequence type of the action instruction in the action instruction set, and stipulating action nodes under different situation condition nodes according to a judgment result; the occurrence probability of different action nodes under the situation condition nodes after reduction is calculated, and the similarity of state data in the game confrontation process is calculated by adopting a dynamic time warping algorithm of a time sequence; and selecting the probability of occurrence of different action nodes under the situation condition node with the minimum similarity as an air maneuvering situation information estimation result. By adopting the method, accurate air maneuvering situation information estimation can be realized.

Description

Technical Field

[0001] The present application relates to the technical field of air maneuver situation information estimation, and in particular to an air maneuver situation information estimation method and device based on multi-sample hierarchical clustering. Background Art

[0002] With the advent of the information and intelligent era, there are more and more types of aerial mobile aircraft. The situation information in the aerial mobile aircraft game confrontation environment presents problems such as real-time information, high complexity, one-sided information, and uncertainty. It is becoming more and more difficult to accurately grasp the trend of situation changes in the confrontation environment. Situation awareness is to understand the current situation and analyze the current intention by combining the obtained situation elements with expert knowledge. It is the basic support for the implementation of command decisions. Situation prediction is based on situation understanding, after a reasonable analysis and understanding of the current situation, to speculate on the development trend of the situation in the future period of time and the enemy's next action and intention.

[0003] In the game confrontation process, multiple aerial maneuvering vehicles generate corresponding strategies through decision-making models, then confront and interact with each other, and continuously improve their own decision-making models in the confrontation. In order to achieve sufficient intelligence, it is necessary to select different opponent schemes to train the own model, so that the own model can respond to the opponent's various schemes in different scenarios, which will generate diversified sample data. The key decision-making opportunities in the game confrontation process come from the diversified sample data generated in the game confrontation process. When faced with sample data of multiple schemes, how to effectively mine the situation condition nodes between the diversified sample data to achieve accurate situation perception and prediction is still relatively less studied. Summary of the invention

[0004] Based on this, it is necessary to provide an air maneuver situation information estimation method and device based on multi-sample hierarchical clustering that can achieve accurate air maneuver situation information estimation in response to the above technical problems.

[0005] A method for estimating air maneuver situation information based on multi-sample hierarchical clustering, the method comprising:

[0006] Obtain a sample sequence set of an aerial maneuvering aircraft in a confrontation game process; the sample sequence set includes multiple rounds of sample sequences; the sample sequence includes actions corresponding to each state of the aerial maneuvering aircraft in the confrontation game process; the state of the sample sequence includes multiple vectors;

[0007] A multi-layer clustering method is used to cluster the sample sequence with the time of the command of the aerial maneuvering aircraft as the classification standard under the given current level clustering value, until the distance from the vector in the sample sequence to each cluster center is less than a preset threshold, and multiple state vectors are obtained; the corresponding situation condition node is generated according to the state vector;

[0008] Obtaining a set of action instructions corresponding to the state under the situation condition node, judging the sample sequence category of the action instructions in the action instruction set, and regulating the action nodes under different situation condition nodes according to the judgment result;

[0009] The probability of occurrence of different action nodes under the situation condition node after the specification is calculated, and the dynamic time warping algorithm of time series is used to calculate the similarity between the process state data and the state data in the situation condition node during the game confrontation process; the probability of occurrence of different action nodes under the situation condition node with the smallest similarity is selected as the estimation result of the air maneuver situation information.

[0010] In one embodiment, the distance from the vector in the sample sequence to each cluster center is

[0011]

[0012] Among them, ω ij is an indicator variable. i When it belongs to j, ω ij =1; otherwise ω ij =0,x i represents the i-th sample point, u j represents the jth cluster center point, i represents the i-th sample sequence, j represents the vector number, n represents the total number of sample sequence bureaus, and k represents the state number.

[0013] In one embodiment, the sample sequence category of the action instruction in the action instruction set is judged, and the action nodes under different situation nodes are regulated according to the judgment result, including:

[0014] Determine the sample category of the action instructions in the action instruction set. If the action instructions in the current action instruction set belong to the same sample sequence, specify the action nodes according to the task type. If the action instructions in the current action instruction set do not belong to the same sample sequence, determine the task styles of different action instructions. If the task styles are exactly the same, merge the two action instructions and perform action specification. If the task styles are inconsistent, specify the order of different action instructions. The task style includes the number of troops, strike target information and entity mounting information of the current action.

[0015] In one embodiment, the probability of occurrence of different action nodes under the situation node after the specification is calculated includes:

[0016] The probability of different action nodes occurring under the situation node after calculation is:

[0017]

[0018] in, Represents the situation node Sh Next action node a h The number of Represents the situation node S h Next action node a j The number of

[0019] In one embodiment, a dynamic time warping algorithm of a time series is used to calculate the similarity between the process state data and the state data in the situation condition node during the game confrontation process, including:

[0020] Construct a distance matrix between the process state data and the state data in the situation condition node in the game confrontation process at each time step; find a path with the minimum cumulative distance from the starting point to the end point in the distance matrix, and record the minimum cumulative distance as the DT distance of the time series; standardize the calculated DT distance and use it as the similarity between the process state data and the state data in the situation condition node in the game confrontation process.

[0021] In one embodiment, the distance matrix of the process state data and the state data in the situation condition node in the game confrontation process at each time step is constructed as follows:

[0022] {D ij =(X i -Y j ) 2 |i=1,2,…,t; j=1,2,…,t}

[0023] Among them, t represents the maximum time step of the sample sequence, X i Represents the process state data during the game confrontation process, Y j Represents the state data in the situation condition node, i represents the sample sequence of the i-th round, and j represents the vector sequence number.

[0024] In one embodiment, the minimum cumulative distance is recorded as the DT distance of the time series:

[0025] DT(X,Y)=min∑ i,j D(i,j)

[0026] Where D(i,j) represents the cumulative distance.

[0027] In one embodiment, the calculated DT distance is normalized and used as the similarity between the process state data in the game confrontation process and the state data in the situation condition node:

[0028]

[0029] Among them, DT(X,Y) represents the DT distance.

[0030] An air maneuver situation information estimation device based on multi-sample hierarchical clustering, the device comprising:

[0031] A data acquisition module is used to acquire a sample sequence set of an aerial maneuvering aircraft in a confrontation game process; the sample sequence set includes multiple rounds of sample sequences; the sample sequence includes actions corresponding to each state of the aerial maneuvering aircraft in the confrontation game process; the state of the sample sequence includes multiple vectors;

[0032] A multi-layer clustering module is used to cluster the sample sequence using the multi-layer clustering method under a given current level clustering value and the time when the aerial maneuvering aircraft issues an instruction as a classification standard until the distance from the vector in the sample sequence to each cluster center is less than a preset threshold, thereby obtaining multiple state vectors; and generating corresponding situation condition nodes according to the state vectors;

[0033] The situation condition node reduction module is used to obtain the action instruction set corresponding to the state under the situation condition node, judge the sample sequence category of the action instruction in the action instruction set, and reduce the action nodes under different situation condition nodes according to the judgment result;

[0034] The air maneuver situation information estimation module is used to calculate the probability of occurrence of different action nodes under the situation condition node after the regulation, and uses the dynamic time warping algorithm of time series to calculate the similarity between the process state data and the state data in the situation condition node during the game confrontation process; the probability of occurrence of different action nodes under the situation condition node with the smallest similarity is selected as the air maneuver situation information estimation result.

[0035] The above-mentioned method and device for estimating air maneuver situation information based on multi-sample hierarchical clustering, this application uses multi-sample aggregation technology, hierarchical clustering to extract condition nodes based on the original state sequence composed of multiple samples, mines situation condition nodes in different scheme confrontations, regulates action nodes under different situation nodes, calculates the probability of situation occurrence corresponding to key nodes, realizes accurate prediction of situation information, and effectively solves the problem of predicting the trend of situation changes in confrontation environments. This application can better realize situation estimation in confrontation environments and can be used in game confrontation decision-making scenarios such as intelligent decision-making of air maneuver aircraft. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 A schematic diagram of a flow chart of an air maneuver situation information estimation method based on multi-sample hierarchical clustering in one embodiment;

[0037] Figure 2 A schematic diagram of a conditional node generation process based on hierarchical clustering in one embodiment;

[0038] Figure 3It is a structural block diagram of an air maneuver situation information estimation device based on multi-sample hierarchical clustering in one embodiment;

[0039] Figure 4 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0041] In one embodiment, Figure 1 As shown, a method for estimating air maneuver situation information based on multi-sample hierarchical clustering is provided, comprising the following steps:

[0042] Step 102, obtaining a sample sequence set of the aerial maneuvering aircraft in the process of the confrontation game; the sample sequence set includes multiple sample sequences; the sample sequence includes the actions corresponding to each state of the aerial maneuvering aircraft in the process of the confrontation game; the state of the sample sequence includes multiple vectors.

[0043] Obtain multiple samples of previous aerial maneuvering aircraft in the confrontation game process, extract the state sequence in the sample. Record each state of the aerial maneuvering aircraft in the confrontation process The corresponding action Recorded as the sample sequence generated by the aerial maneuvering vehicle during the current confrontation process Where i represents the i-th sample, and k represents the k-th state of the current sample. The state sequence in multiple samples is recorded as S = {S1, S2, ... S m}, where m represents the number of samples, and the state S of the i-th sample i Expressed as That is, the sample state of each game is composed of j vectors, where j represents the number of states with action nodes generated by the model under the current sample.

[0044] Step 104, using a multi-layer clustering method to cluster the sample sequence with the time when the aerial maneuverable vehicle issues instructions as the classification standard under a given current level clustering value, until the distance from the vector in the sample sequence to each cluster center is less than a preset threshold, and multiple state vectors are obtained; corresponding situation condition nodes are generated according to the state vectors.

[0045] Hierarchical clustering is used to extract key nodes of different situations. For the state sequence S, given the k value that needs to be clustered at the current level, the time when the aerial maneuvering aircraft gives instructions is used as the classification standard at the beginning. Assuming that it is divided into k categories, find the center point of each category at present, recorded as {μ1,μ2,…,μ k}, calculate the distance from the vector to each cluster center as:

[0046]

[0047] Among them, ω ij is an indicator variable. i When it belongs to j, ω ij =1; otherwise ω ij = 0. Given the minimum threshold δ from the vector to the cluster center, that is, if the distance J from the vector to each cluster center is less than the threshold δ, clustering is stopped. The k-means clustering algorithm is used to perform hierarchical clustering on the state sequence by setting different k values ​​and thresholds δ, and the vectors of different classes are classified into different states. Then, corresponding conditional nodes are generated according to different state vectors to mine the key conditional nodes in the situation, that is, the situation conditional nodes. Figure 2 shown.

[0048] Step 106, obtain the action instruction set corresponding to the state under the situation condition node, judge the sample sequence category of the action instruction in the action instruction set, and standardize the action nodes under different situation condition nodes according to the judgment result, so as to avoid the redundancy of the same action information under different situation condition nodes and disrupt the accuracy of situation information prediction.

[0049] To reduce the action nodes under different situation condition nodes, first obtain the corresponding state S under the key condition node generated in step 104 h A collection of action instructions Where f represents the current state S h The number of corresponding action instructions;

[0050] Judgment set A h The sample category of the action instruction in . If the current set A h The action instructions in the same sample are reduced according to the task type. If the current set A h The action instructions in the examples do not belong to the same sample. The task styles of different action instructions are judged (the task style includes the number of troops, strike targets, entity mounts and other information of the current action). If the task styles are exactly the same, the two action instructions are merged and action specifications are performed. If the task styles are inconsistent, the order of different action instructions is specified.

[0051] Step 108, calculate the probability of occurrence of different action nodes under the situation condition node after the regulation, and use the dynamic time warping algorithm of time series to calculate the similarity between the process state data and the state data in the situation condition node during the game confrontation process; select the probability of occurrence of different action nodes under the situation condition node with the smallest similarity as the estimation result of the air maneuver situation information.

[0052] First, count each situation condition node S h The number of different action nodes Where m represents the number of action nodes, and action node a under the current condition node h The probability of occurrence is:

[0053]

[0054] Repeat the steps of reducing the action nodes under different situation nodes - calculating the probabilities of different actions under the situation nodes, and complete the calculation of the action node probabilities corresponding to the action instructions under all conditional node states in the steps of reducing the action nodes under different situation nodes.

[0055] Then match the state information of the game confrontation in the new deduction, calculate the similarity, match the corresponding situation conditions, and select the corresponding probability to complete the situation information estimation. The specific process includes:

[0056] First, the process state data obtained in the game confrontation and the state data in the conditional node include the state of the aerial maneuvering aircraft, spatial coordinates, the state of the other party and other information, which are multidimensional time series. The dynamic time warping algorithm of the time series is used to calculate the similarity of the two data. First, the Euclidean distance matrix D of the sample data X and Y at each time step is constructed:

[0057] {D ij =(X i -Y j ) 2 |i=1,2,…,t; j=1,2,…,t}

[0058] Among them, t is the maximum time step of the sample sequence. Then find a path with the minimum cumulative distance D(i,j) from the starting point to the end point in the matrix D. This minimum cumulative distance is recorded as the DT distance of the time series:

[0059]

[0060] Finally, the calculated DT distance is normalized as the similarity measure between the deduced state data X and the state data Y in the conditional node:

[0061]

[0062] Select the condition node with the smallest similarity value Sim, and select the probability value in the corresponding situation condition node.

[0063] In the above-mentioned method for estimating the situation information of air maneuver based on multi-sample hierarchical clustering, this application uses multi-sample aggregation technology to extract condition nodes based on the original state sequence composed of multiple samples by hierarchical clustering, mines the situation condition nodes in the confrontation of different schemes, regulates the action nodes under different situation nodes, calculates the probability of the situation corresponding to the key nodes, realizes the accurate prediction of situation information, and effectively solves the problem of predicting the trend of situation changes in the confrontation environment. This application can better realize situation estimation in the confrontation environment, and can be used in game confrontation decision-making scenarios such as intelligent decision-making of air maneuver aircraft.

[0064] In one embodiment, the distance from the vector in the sample sequence to each cluster center is

[0065]

[0066] Among them, ω ij is an indicator variable. i When it belongs to j, ω ij =1; otherwise ω ij =0,x i represents the i-th sample point, u j represents the jth cluster center point, i represents the i-th sample sequence, j represents the vector number, n represents the total number of sample sequence bureaus, and k represents the state number.

[0067] In one embodiment, the sample sequence category of the action instruction in the action instruction set is judged, and the action nodes under different situation nodes are regulated according to the judgment result, including:

[0068] Determine the sample category of the action instructions in the action instruction set. If the action instructions in the current action instruction set belong to the same sample sequence, specify the action nodes according to the task type. If the action instructions in the current action instruction set do not belong to the same sample sequence, determine the task styles of different action instructions. If the task styles are exactly the same, merge the two action instructions and perform action specification. If the task styles are inconsistent, specify the order of different action instructions. The task style includes the number of troops, strike target information and entity mounting information of the current action.

[0069] In one embodiment, the probability of occurrence of different action nodes under the situation node after the specification is calculated includes:

[0070] The probability of different action nodes occurring under the situation node after calculation is:

[0071]

[0072] in, Represents the situation node S h Next action node a h The number of Represents the situation node S h Next action node a j The number of

[0073] In one embodiment, a dynamic time warping algorithm of a time series is used to calculate the similarity between the process state data and the state data in the situation condition node during the game confrontation process, including:

[0074] Construct a distance matrix between the process state data and the state data in the situation condition node in the game confrontation process at each time step; find a path with the minimum cumulative distance from the starting point to the end point in the distance matrix, and record the minimum cumulative distance as the DT distance of the time series; standardize the calculated DT distance and use it as the similarity between the process state data and the state data in the situation condition node in the game confrontation process.

[0075] In one embodiment, the distance matrix of the process state data and the state data in the situation condition node in the game confrontation process at each time step is constructed as follows:

[0076] {D ij =(X i -Y j ) 2 |i=1,2,…,t; j=1,2,…,t}

[0077] Among them, t represents the maximum time step of the sample sequence, X i Represents the process state data during the game confrontation process, Y j Represents the state data in the situation condition node, i represents the sample sequence of the i-th round, and j represents the vector sequence number.

[0078] In one embodiment, the minimum cumulative distance is recorded as the DT distance of the time series:

[0079] DT(X,Y)=min∑ i,j D(i,j)

[0080] Among them, D(i,j) represents the cumulative distance.

[0081] In one embodiment, the calculated DT distance is normalized and used as the similarity between the process state data in the game confrontation process and the state data in the situation condition node:

[0082]

[0083] Among them, DT(X,Y) represents the DT distance.

[0084] In a specific embodiment, based on three samples generated by one's own intelligent model and two different opponents in a certain game confrontation scenario,

[0085] 1. Record the action samples generated in each round of confrontation Record the state vector S = {S1, S2, S3} in the 3 rounds of samples, where the state S1 of the first round of samples is expressed as That is, the sample state of the first round is composed of j vectors; the state S2 of the sample of the second round is expressed as That is, the sample state of the second round consists of l vectors, and the state S3 of the sample of the third round is expressed as That is, the sample state of the third round consists of m vectors.

[0086] 2. For the state sequence S, given the k value k = 2 that needs to be clustered at the current level, and given the minimum threshold δ from the vector to the cluster center, the k-means clustering algorithm is used to cluster the state sequence For clustering, given the k value that needs to be clustered at the current level, the time when the agent gives instructions is used as the classification standard at the beginning, and the center point of each class is found, recorded as {μ1,μ2}, and the distance between each data point and its cluster center is calculated as:

[0087]

[0088] Among them, ω ij is an indicator variable. i When it belongs to j, ω ij =1; otherwise ω ij =0, and then generate corresponding 2 conditional nodes according to different state vectors.

[0089] 3. Specify the action nodes under different situation nodes. First, obtain the set of action instructions corresponding to the state S1 set under condition node 1 generated in step 2. The set of action instructions corresponding to the state S2 set under condition node 2

[0090] 4. Calculate the probability of different actions under the situation node.

[0091] First, count each situation condition node S h The number of different action nodes Where m represents the number of action nodes, and action node a under the current condition node h The probability of occurrence is:

[0092]

[0093] Repeat steps 2 to 3 to complete the calculation of the action node probabilities corresponding to the action instructions under all conditional node states in step 3.

[0094] 5. Similarity calculation: First, construct the distance matrix D between the process state data X obtained in the game confrontation and the state data Y in the conditional node at each time step:

[0095] {D ij =(X i -Y j ) 2 |i=1,2,…,t; j=1,2,…,t}

[0096] Among them, t is the maximum time step of the sample sequence. Then find a path with the minimum cumulative distance D(i,j) from the starting point to the end point in the matrix D. This minimum cumulative distance is recorded as the DT distance of the time series:

[0097] DT(X,Y)=min∑ i,j D(i,j)

[0098] Finally, the calculated DT distance is normalized as the similarity measure between the deduced state data X and the state data Y in the conditional node:

[0099]

[0100] Select the conditional node with the smallest similarity value Sim, and select the probability value in the corresponding conditional node.

[0101] It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0102] In one embodiment, Figure 3 As shown, an air maneuver situation information estimation device based on multi-sample hierarchical clustering is provided, comprising: a data acquisition module 302, a multi-layer clustering module 304, a situation condition node reduction module 306 and an air maneuver situation information estimation module 308, wherein:

[0103] The data acquisition module 302 is used to acquire a sample sequence set of the aerial maneuvering aircraft in the confrontation game process; the sample sequence set includes multiple rounds of sample sequences; the sample sequence includes actions corresponding to each state of the aerial maneuvering aircraft in the confrontation game process; the state of the sample sequence includes multiple vectors;

[0104] The multi-layer clustering module 304 is used to cluster the sample sequence using the multi-layer clustering method under the given current level clustering value and the time when the aerial maneuvering aircraft issues the command as the classification standard until the distance from the vector in the sample sequence to each cluster center is less than a preset threshold, thereby obtaining multiple state vectors; and generating corresponding situation condition nodes according to the state vectors;

[0105] The situation condition node reduction module 306 is used to obtain the action instruction set corresponding to the state under the situation condition node, judge the sample sequence category of the action instruction in the action instruction set, and reduce the action nodes under different situation condition nodes according to the judgment result;

[0106] The air maneuver situation information estimation module 308 is used to calculate the probability of occurrence of different action nodes under the situation condition node after the regulation, and use the dynamic time warping algorithm of the time series to calculate the similarity between the process state data and the state data in the situation condition node during the game confrontation process; the probability of occurrence of different action nodes under the situation condition node with the smallest similarity is selected as the air maneuver situation information estimation result.

[0107] For the specific definition of the air maneuver situation information estimation device based on multi-sample hierarchical clustering, please refer to the definition of the air maneuver situation information estimation method based on multi-sample hierarchical clustering in the above text, which will not be repeated here. Each module in the above-mentioned air maneuver situation information estimation device based on multi-sample hierarchical clustering can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0108] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 4As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for estimating air maneuver situation information based on multi-sample hierarchical clustering is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse, etc.

[0109] Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0110] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0111] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0112] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

[0113] In summary, the above are only preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for estimating air maneuver situation information based on multi-sample hierarchical clustering, characterized in that: The method comprises: Acquire a sample sequence set of an aerial maneuvering aircraft in a confrontation game process; the sample sequence set includes multiple rounds of sample sequences; the sample sequence includes actions corresponding to each state of the aerial maneuvering aircraft in the confrontation game process; the state of the sample sequence includes multiple vectors; A multi-layer clustering method is used to cluster the sample sequence with the time of the aerial maneuvering aircraft issuing instructions as a classification standard under a given current level clustering value, until the distance from the vector in the sample sequence to each cluster center is less than a preset threshold, thereby obtaining a plurality of state vectors; and a corresponding situation condition node is generated according to the state vector; Obtaining a set of action instructions corresponding to the state under the situation condition node, judging the sample sequence category of the action instructions in the action instruction set, and regulating the action nodes under different situation condition nodes according to the judgment result; The probability of occurrence of different action nodes under the situation condition node after the specification is calculated, and the dynamic time warping algorithm of time series is used to calculate the similarity between the process state data and the state data in the situation condition node during the game confrontation process; the probability of occurrence of different action nodes under the situation condition node with the smallest similarity is selected as the estimation result of the air maneuver situation information.

2. The method according to claim 1, characterized in that The distance from the vector in the sample sequence to each cluster center is Among them, ω ij is an indicator variable. i When it belongs to j, ω ij =1; otherwise ω ij =0,x i represents the i-th sample point, u j represents the jth cluster center point, i represents the i-th sample sequence, j represents the vector number, n represents the total number of sample sequence bureaus, and k represents the state number.

3. The method according to claim 1, characterized in that The sample sequence category of the action instruction in the action instruction set is judged, and the action nodes under different situation nodes are regulated according to the judgment result, including: Determine the sample category of the action instructions in the action instruction set. If the action instructions in the current action instruction set belong to the same sample sequence, specify the action nodes according to the task type. If the action instructions in the current action instruction set do not belong to the same sample sequence, determine the task styles of different action instructions. If the task styles are completely consistent, merge the two action instructions and perform action specification. If the task styles are inconsistent, specify the order of different action instructions. The task style includes the number of troops, strike target information and entity mounting information of the current action.

4. The method according to any one of claims 1 to 3, characterized in that: Calculate the probability of different action nodes occurring under the situation node after the specification, including The probability of different action nodes occurring under the situation node after calculation is: in, Represents the situation node S h Next action node a h The number of Represents the situation node S h Next action node a j The number of 5. The method according to claim 1, characterized in that The dynamic time warping algorithm of time series is used to calculate the similarity between the process state data and the state data in the situation condition node during the game confrontation process, including: Construct a distance matrix between the process state data and the state data in the situation condition node in the game confrontation process at each time step; find a path with the minimum cumulative distance from the starting point to the end point in the distance matrix, and record the minimum cumulative distance as the DT distance of the time series; standardize the calculated DT distance and use it as the similarity between the process state data and the state data in the situation condition node in the game confrontation process.

6. The method according to claim 5, characterized in that The distance matrix between the process state data and the state data in the situation condition node in the game confrontation process at each time step is constructed as {D ij =(X i -Y j ) 2 |i=1,2,…,t; j=1,2,…,t} Among them, t represents the maximum time step of the sample sequence, X i Represents the process state data during the game confrontation process, Y j Represents the situation condition node S j The state data in , i represents the sample sequence of the i-th round, and j represents the vector sequence number.

7. The method according to claim 5, characterized in that The minimum cumulative distance is recorded as the DT distance of the time series: DT(X,Y)=min∑ i,j D(i,j) Among them, D(i,j) represents the cumulative distance.

8. The method according to claim 5, characterized in that The calculated DT distance is standardized and used as the similarity between the process state data in the game confrontation process and the state data in the situation condition node: Among them, DT(X,Y) represents the DT distance.

9. An air maneuver situation information estimation device based on multi-sample hierarchical clustering, characterized in that: The device comprises: A data acquisition module is used to acquire a sample sequence set of an aerial maneuvering aircraft in a confrontation game process; the sample sequence set includes multiple rounds of sample sequences; the sample sequence includes actions corresponding to each state of the aerial maneuvering aircraft in the confrontation game process; the state of the sample sequence includes multiple vectors; A multi-layer clustering module is used to cluster the sample sequence using a multi-layer clustering method under a given current level clustering value and taking the time when the aerial maneuvering aircraft issues an instruction as a classification standard until the distance from the vector in the sample sequence to each cluster center is less than a preset threshold, thereby obtaining a plurality of state vectors; and generating corresponding situation condition nodes according to the state vectors; A situation condition node reduction module is used to obtain a set of action instructions corresponding to the state under the situation condition node, judge the sample sequence category of the action instructions in the action instruction set, and reduce the action nodes under different situation condition nodes according to the judgment result; The air maneuver situation information estimation module is used to calculate the probability of occurrence of different action nodes under the situation condition node after the regulation, and uses the dynamic time warping algorithm of time series to calculate the similarity between the process state data and the state data in the situation condition node during the game confrontation process; the probability of occurrence of different action nodes under the situation condition node with the smallest similarity is selected as the air maneuver situation information estimation result.

Images