Game prediction method for air combat target intention based on nearest neighbor hierarchical clustering analysis

Abstract

The present invention provides an air combat target intention game prediction method based on nearest neighbor hierarchical clustering analysis. This method first addresses the uncertainty problem of drone air combat information and uses interval numbers to represent the uncertain air combat information collected by sensors. Secondly, a nearest neighbor hierarchical clustering algorithm is used to construct an intention game prediction model to obtain baseline eigenvectors for various air combat intention categories. Then, the interval number eigenvalues ​​of the target to be identified and the baseline eigenvalues ​​of the known intention are used as the two parties in a two-person zero-sum game to construct an intention game payoff matrix. Finally, the Nash equilibrium solution of the intention game payoff matrix is ​​solved to determine the air combat intention of the target to be identified. This method effectively solves the problem of target intention prediction under uncertain air combat information.

Description

Technical Field

[0001] The present invention belongs to the field of target intention prediction, and in particular relates to an air combat target intention game prediction method based on nearest neighbor hierarchical clustering analysis. Background Art

[0002] Target intention prediction technology is a crucial countermeasure in modern air combat and crucial for drone air combat decision-making. Target intention prediction analyzes the interconnectedness of air combat situational elements to predict enemy aircraft's combat intent, thereby assisting drone decision-making and shifting the air combat situation in a direction favorable to our drones.

[0003] Air combat target intention game prediction involves processing and analyzing the characteristic values ​​of known targets, combined with game theory knowledge, to determine the enemy's likely combat intention at the next moment in an air combat confrontation between enemy and friendly drones. Therefore, quickly and accurately predicting the enemy's combat intention will help our drone's air combat attack and defense game decision-making system formulate reasonable combat strategies and gain an advantageous position in the air combat. However, due to the increasing complexity of air combat environments, the variety of air combat targets, and the presence of decoys, battlefield interference, information transmission delays, and packet loss, the air combat information after data processing still contains uncertainty. Existing target intention prediction methods simply perform template comparison, and the template remains unchanged. In short, target intention prediction methods under uncertain information are relatively rare. Therefore, research on target intention game prediction in drone air combat under uncertain information has strong practical significance.

[0004] In engineering applications, data sometimes needs to be classified based on specific criteria (e.g., biological morphology, air pollution levels, etc.). However, in drone air combat, different types of intent lack clear boundaries. Therefore, to classify air combat intent under uncertain information, a nearest neighbor hierarchical clustering analysis method is employed. An algorithm for determining the degree of similarity between templates and targets, as well as rules for determining this similarity, is designed. These algorithms primarily include standardizing air combat data, establishing nearest neighbor boundaries for air combat targets, and performing cluster analysis of air combat targets.

[0005] In summary, in order to improve the accuracy and speed of UAV target intention prediction, we should continue to study the problem of UAV target intention prediction in complex air combat environments when the UAV flight characteristic data is uncertain. Summary of the Invention

[0006] Purpose of the Invention: To address the problem of target intention prediction in complex air combat environments, where target intention prediction results often suffer from long solution times and low prediction accuracy due to a series of issues, such as the influence of air combat information uncertainty and the specific quantification of each target's intention. This invention provides an air combat target intention game prediction method based on nearest neighbor hierarchical clustering analysis. First, to address the uncertainty of UAV air combat information and the need to quantify each target's intention, interval numbers are used to represent the uncertain air combat information collected by sensors. Second, a nearest neighbor hierarchical clustering algorithm is used to construct an intention game prediction model to obtain baseline eigenvectors for various air combat intention categories. Then, the interval number eigenvalues ​​of the target to be identified and the baseline eigenvalues ​​of the known intention are used as the two sides in a two-person zero-sum game. An intention game payoff matrix is ​​constructed, and the Nash equilibrium solution of the matrix is ​​solved to determine the air combat intention of the target to be identified. Finally, simulations verify the effectiveness of this method, demonstrating its promising prospects for air combat applications.

[0007] The present invention comprises the following steps:

[0008] Step 1: Preprocess the air combat target database feature information and sensor detection information;

[0009] Step 2: Establish a nearest neighbor hierarchical clustering model for target intention prediction based on target data;

[0010] Step 3: Design a target to be identified and conduct a simulated game with a target with known intention to predict and judge the combat intention of the air combat target.

[0011] Step 1 includes:

[0012] Assume that the set of air combat targets with known eigenvalues ​​but unknown intentions is {k1, k2, …, k n}, where k i =[k i,1 ,k i,2 ,…,k i,m ] represents the i-th air combat target information, i=1,2,…,n,k i,m is the mth eigenvalue of the i-th air combat target, and the eigenvalue is a real number;

[0013] Setting air combat targets includes 5 characteristics: azimuth The distance R, speed V, entry angle θ, height H, and characteristic value normalization formula are shown in formula (1):

[0014]

[0015] Calculate the average value of the hth eigenvalue of all air combat targets by formula (2)

[0016]

[0017] The standard deviation σ of the hth eigenvalue of all air combat targets is calculated by formula (3): h :

[0018]

[0019] Among them, k i ' ,h is the hth eigenvalue of the ith air combat target after normalization, and h ranges from 1 to m.

[0020] Step 2 includes the following steps:

[0021] Step 2-1, construct the nearest boundary of the air combat target;

[0022] Step 2-2, hierarchical clustering of air combat targets;

[0023] Step 2-3, calculate the baseline feature vector of the intent.

[0024] Step 2-1 includes: defining the i-th air combat target x in a two-dimensional plane based on the five features i With the jth air combat target x j Similarity distance d ij for:

[0025]

[0026] where k j ' ,h represents the hth eigenvalue of the jth air combat target after normalization;

[0027] Then the i-th air combat target x i Nearest neighbor boundary NNB(x i )for:

[0028] NNB(x i )={(x,y)|x i -d ij ≤x≤x i +d ij ,y i -d ij ≤y≤y i +d ij} (5)

[0029] Among them, (x, y) is the area divided according to the boundary.

[0030] Step 2-2 includes: performing nearest neighbor search in the nearest neighbor boundary of each area according to the area divided in step 2-1, and finding paired mutual nearest neighbors in the nearest neighbor boundary of each area. The global mutual nearest neighbors indicate that two air combat targets are each other's nearest neighbors in the entire air combat target dataset.

[0031] Step 2-3 includes: setting the air combat target to be divided into p intention groups after the processing of step 2-1 to step 2-2, forming a set M = {M1, ..., M i ,…,M p}, i=1,2,…,p, where M p Represents the pth intention group, each intention group includes n i air combat targets, and each target is represented by an azimuth The five features of distance R, speed V, entry angle θ, and height H are specifically represented to obtain the reference feature matrix S0 of p intentions = (s i,h ) p×5 , where s i,h The baseline feature vector s of the i-th intention i h features, h represents the number of features, the baseline feature vector s of the i-th intention i The expression of is shown in formula (6):

[0032]

[0033] Among them, Φ i 、R_set i 、Θ i 、V_set i 、H_set i They respectively represent the azimuth set, distance set, entry angle set, speed set, and altitude set of the target group corresponding to the i-th combat intention.

[0034] Step 3 includes:

[0035] The target G to be identified j The characteristic vector of the i-th intention and the baseline characteristic vector s i The correlation function rw j,i As the payment function of the simulated game, it is shown in formula (7):

[0036]

[0037] in The target G that represents the intention to be identified j The characteristic vector of the i-th combat intention and the baseline characteristic vector s i The similarity degree on the kth feature, The target G to be identifiedj The interval number type of the h-th characteristic information, For the lower bound, is the upper bound, As shown in formula (8):

[0038]

[0039] in is the minimum operator, is the maximum operator, ρ is the distance characteristic coefficient, dw j,i (h) is the target G to be identified on the kth feature j The characteristic vector of the i-th combat intention and the baseline characteristic vector s i The distance between them is calculated as shown in formula (9):

[0040]

[0041] The air combat game is a two-player zero and non-cooperative game problem. The matrix calculated according to formulas (7), (8), and (9) is the intention correlation game payoff matrix R1 = (rw j,i ) p×m ;

[0042] Set the intention mixing strategy X = (x1, x2, ..., x m ), where x m is the mth intention to be identified, and the Nash equilibrium v ​​of the intention target to be identified is shown in the following formula (10):

[0043]

[0044] Formula (10) is solved to determine the intention of the target to be identified.

[0045] The beneficial effects of the present invention are as follows:

[0046] 1. Using interval numbers to represent the air combat targets to be identified can well represent the uncertainty of air combat information.

[0047] 2. The nearest neighbor hierarchical clustering analysis is used to construct the nearest neighbor hierarchical clustering model for target intention prediction, which can well distinguish the various air combat intention categories and solve the problem of fuzzy boundaries of intention categories. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments, and the above and / or other advantages of the present invention will become more apparent.

[0049] Figure 1 Game prediction process for air combat target intentions.

[0050] Figure 2 This is a flowchart for air combat target intention game prediction based on nearest neighbor hierarchical clustering analysis.

[0051] Figure 3 This is a dynamic classification diagram of 15 combat groups. DETAILED DESCRIPTION

[0052] The present invention provides an air combat target intention game prediction method based on nearest neighbor hierarchical clustering analysis, comprising the following steps:

[0053] Step 1: Preprocess the feature information in the air combat target database and the drone information detected by the sensor

[0054] Assume that the set of air combat targets with known eigenvalues ​​but unknown intentions is {k1,k2,…,k n}, where k i =[k i,1 ,k i,2 ,…,k i,m ](i=1,2,…,n) represents the i-th air combat target, k i,m represents the mth eigenvalue of the i-th air combat target, and the eigenvalue is a real number, i = 1, ..., n. Assume that the air combat target is composed of 5 features - azimuth Distance R (m), speed V (m / s), entry angle The height H (m) is specifically expressed.

[0055] Since the dimensions and magnitudes of the eigenvalues ​​of air combat targets are not completely consistent, in order to facilitate subsequent matrix calculation and intent classification, the eigenvalues ​​of the air combat targets in the set are standardized so that all eigenvalues ​​converge within the pre-set data range. The specific eigenvalue standardization formula is shown in formula (1):

[0056]

[0057] Formula (2) is used to calculate the average value of the hth eigenvalue of all air combat targets:

[0058]

[0059] Formula (3) is the formula for calculating the standard deviation of the hth eigenvalue of all air combat targets:

[0060]

[0061] where k' i,h is the hth eigenvalue of the i-th air combat target after normalization; is the average value of the hth eigenvalue of all air combat targets; σh is the standard deviation of the hth eigenvalue of all air combat targets.

[0062] Step 2: Establish a nearest neighbor hierarchical clustering model for target intent prediction based on target data in the target database whose feature values ​​are known but intent is unknown

[0063] 1) Construct the nearest boundary of the air combat target

[0064] In the two-dimensional plane, the air combat target x i With air combat target x j Based on the five features, the similarity distance is defined as:

[0065]

[0066] Then its nearest neighbor boundary NNB(x i )for:

[0067] NNB(x i )={(x,y)|x i -d ij ≤x≤x i +d ij ,y i -d ij ≤y≤y i +d ij} (5)

[0068] Among them, (x, y) is the area divided according to the boundary.

[0069] 2) Hierarchical clustering of air combat targets

[0070] According to the region boundaries divided in step 2-1, the nearest neighbor search is performed in the nearest neighbor boundaries of each region. Pairs of mutual nearest neighbors can be found in the nearest neighbor boundaries of each region. The global mutual nearest neighbors indicate that two air combat targets are the nearest neighbors of each other in the entire air combat target dataset.

[0071] 3) Calculation of the baseline feature vector of intent

[0072] Assume that the air combat targets can be divided into p intention groups M = {M1,…,M i ,…,M p}(i=1,2,…,p), where M p The pth intention group, each intention group consists of n i air combat targets, and each target is composed of azimuth Distance R (m), speed V (m / s), entry angle The five features of height H(m) are specifically represented to obtain the benchmark feature matrix S0=(si,h ) p×5 , where s i,h The baseline feature vector s of the intention i h features, h represents the number of features, the baseline feature vector s of the i-th intention i The expression of is shown in formula (6):

[0073]

[0074] Among them, Φ i 、R_set i 、Θ i 、V_set i 、H_set i Represents the azimuth set, distance set, entry angle set, speed set, and altitude set of the target group corresponding to the i-th combat intention.

[0075] Step 3: Design a target to be identified and conduct a simulated game with a known target to solve the Nash equilibrium solution of the intention, and predict and judge the combat intention of the air combat target.

[0076] According to the interval number grey relational formula, the target G to be identified is j The characteristic vector of the i-th intention and the baseline characteristic vector s i The correlation function rw j,i As the payment function of the simulated game, it is shown in formula (7):

[0077]

[0078] in The target G that represents the intention to be recognized j The characteristic vector of the i-th combat intention and the baseline characteristic vector s i Similarity on the kth feature The target G to be identified j The interval number type of the h-th characteristic information, For the lower bound, is the upper bound, As shown in formula (8):

[0079]

[0080] in is the minimum operator, is the maximum operator, ρ is the distance feature coefficient, and the target G to be identified on the kth feature j The characteristic vector of the i-th combat intention and the baseline characteristic vector s i The distance calculation method between is shown in formula (9):

[0081]

[0082] The air combat game is a two-player zero and non-cooperative game problem. The characteristic vector of the target to be identified and the baseline characteristic vector of the known combat intention are used as the two parties of the intention game. The matrix calculated according to formulas (7), (8), and (9) is the intention correlation game payoff matrix R1 = (rw j,i ) p×m Assume that the intention mixed strategy of the target to be identified is x=(x1,x2,…,x m ), where x m For the mth intention to be identified, the intention Nash equilibrium v ​​of the intention target to be identified is obtained according to the intention association game payoff matrix R1 as shown in the following formula (10).

[0083]

[0084] According to the Nash equilibrium solution of intention obtained by formula (10), the intention of the target to be identified is judged, x i The larger the value of , the higher the possibility that the target to be identified is the i-th intention.

[0085] The process of the air combat target intention game prediction method based on nearest neighbor hierarchical clustering analysis is as follows:

[0086] Step 1 performs data preprocessing on the real-number data set of air combat information in the air combat target information database, whose features are known but intentions are unknown.

[0087] Step 2 is based on the data preprocessing of Step 1, according to the air combat target x i With air combat target x j The nearest boundary of the air combat target is established based on the comprehensive similarity of the five features.

[0088] Step 3 selects an appropriate confidence level λ∈[0,1] and searches for the global nearest neighbors to obtain the intent classification.

[0089] Step 4: Calculate the baseline feature vector s for each intent based on the intent classification results. i expression.

[0090] Step 5: The target air combat information to be identified collected by the sensor is represented by interval numbers.

[0091] Step 6: Calculate the target G to be identified based on the calculation results of Step 4. j The characteristic vector of the i-th intention and the baseline characteristic vector s i The correlation function rw j,i .

[0092] Step 7: The characteristic vector of the target to be identified and the baseline characteristic vector of the known combat intention are used as the two parties in the intention game. The matrix calculated according to formulas (7), (8), and (9) is the intention correlation game payoff matrix R1 = (rw j,i ) p×m .

[0093] Step8: According to the intention correlation game payoff matrix R1=(rw j,i ) p×m Find the Nash equilibrium solution of the target to be identified.

[0094] In summary, the flowchart of air combat target intention game prediction based on nearest neighbor hierarchical clustering analysis is as follows: Figure 2 shown.

[0095] To demonstrate the effectiveness of the method proposed in this invention, 15 batches of air combat target data with known eigenvalues ​​but unknown intentions are shown in Table 1, and the eigenvalues ​​of three air combat targets with intentions to be identified are shown in Table 2. The method proposed in this invention is used to predict the air combat intentions of the three targets.

[0096] The simulation results are as follows Figure 3 , as shown in Table 3. Figure 3 is a dynamic classification diagram of combat groups, composed of Figure 3 It can be seen that the features refer to the 15 known air combat targets divided into 6 combat groups with the following intentions: {k1, k3, k5, k6} (reconnaissance), {k8, k 10}(penetration), {k 12 ,k 13 ,k 15}(electronic jamming), {k2,k4,k7}(attack), {k9}(surveillance), {k 11 ,k 14}(defense), the baseline feature vectors of each air combat intention are shown in Table 3. According to formula (6) and formula (7), the target to be identified G is calculated. i The characteristic vector of (i=1,2,3) and the characteristic vector s of combat intention j The association degree between (j=1,2,3,4,5,6) is obtained to obtain the intention association matrix R1, which is also the intention association game payoff matrix, R1 is shown in formula (11), and the mixed strategy Nash equilibrium of the payoff matrix is ​​solved.

[0097]

[0098] The Nash equilibrium result of the target intention game prediction is [0.1009, 0, 0.6948, 0, 0, 0.2043]. According to the analysis of the Nash equilibrium result, the combat intention of target G1 is penetration, the combat intention of target G2 is electronic jamming, and the combat intention of target G3 is electronic jamming.

[0099] Table 1

[0100]

[0101]

[0102] Table 2

[0103]

[0104] Table 3

[0105]

[0106] In a specific implementation, the present application provides a computer storage medium and a corresponding data processing unit, wherein the computer storage medium is capable of storing a computer program that, when executed by the data processing unit, can execute the invention content of the method for predicting air combat target intentions based on nearest neighbor hierarchical clustering analysis provided by the present invention and some or all of the steps in each embodiment. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0107] Those skilled in the art can clearly understand that the technical solutions in the embodiments of the present invention can be implemented by means of computer programs and their corresponding general hardware platforms. Based on this understanding, the technical solutions in the embodiments of the present invention, in essence or in other words, the part that contributes to the prior art, can be embodied in the form of a computer program, i.e., a software product. The computer program software product can be stored in a storage medium and includes several instructions for enabling a device including a data processing unit (which can be a personal computer, a server, a single-chip microcomputer, a MUU, or a network device, etc.) to execute the methods described in various embodiments of the present invention or certain parts of the embodiments.

[0108] The present invention provides a method for predicting target intentions in air combat based on nearest neighbor hierarchical clustering analysis. There are many methods and approaches for implementing this technical solution. The above is merely a preferred embodiment of the present invention. It should be noted that those skilled in the art may make various improvements and modifications without departing from the principles of the present invention, and such improvements and modifications are also within the scope of protection of the present invention. Any components not specified in this embodiment may be implemented using existing technologies.

Claims

1. An air combat target intention game prediction method based on nearest neighbor hierarchical clustering analysis is characterized by: The following steps are involved: Step 1: Preprocess the air combat target database feature information and sensor detection information; Step 2: Establish a nearest neighbor hierarchical clustering model for target intention prediction based on target data; Step 3: Design a target to be identified and conduct a simulated game with a known target to predict and judge the combat intention of the air combat target; Step 3 includes: The target G to be identified j The characteristic vector of the i-th intention and the baseline characteristic vector s i The correlation function rw j,i As the payment function of the simulated game, it is shown in formula (7): in The target G that represents the intention to be identified j The characteristic vector of the i-th combat intention and the baseline characteristic vector s i The similarity degree on the kth feature, The target G to be identified j The interval number type of the h-th characteristic information, For the lower bound, is the upper bound, As shown in formula (8): in is the minimum operator, is the maximum operator, ρ is the distance characteristic coefficient, dw j,i (h) is the target G to be identified on the kth feature j The characteristic vector of the i-th combat intention and the baseline characteristic vector s i The distance between them is calculated as shown in formula (9): where s i,h is the baseline feature vector s of the i-th intention i The hth feature of , where h represents the number of features. The air combat game is a two-player zero and non-cooperative game problem. The matrix calculated according to formulas (7), (8), and (9) is the intention correlation game payoff matrix R1 = (rw j,i ) p×m ; p represents the serial number of the intended group; Set the intention mixing strategy X = (x1, x2, ..., x m ), where x m is the mth intention to be identified, and the Nash equilibrium v ​​of the intention target to be identified is shown in the following formula (10): Formula (10) is solved to determine the intention of the target to be identified.

2. The method according to claim 1, characterized in that Step 1 includes: Assume that the set of air combat targets with known eigenvalues ​​but unknown intentions is {k1, k2, …, k n }, where k i =[k i,1 ,k i,2 ,…,k i,m' ] represents the i-th air combat target information, i=1,2,…,n,k i,m' is the m'th eigenvalue of the i-th air combat target, and the eigenvalue is a real number; Setting air combat targets includes 5 characteristics: azimuth The distance R, speed V, entry angle θ, height H, and characteristic value normalization formula are shown in formula (1): Calculate the average value of the hth eigenvalue of all air combat targets by formula (2) The standard deviation σ of the hth eigenvalue of all air combat targets is calculated by formula (3): h : Among them, k' i,h is the hth eigenvalue of the ith air combat target after normalization, and the value of h ranges from 1 to m'.

3. The method according to claim 2, characterized in that Step 2 includes the following steps: Step 2-1, construct the nearest boundary of the air combat target; Step 2-2, hierarchical clustering of air combat targets; Step 2-3, calculate the baseline feature vector of the intent.

4. The method according to claim 3, characterized in that Step 2-1 includes: defining the i-th air combat target x in a two-dimensional plane based on the five features i With the jth air combat target x j Similarity distance d ij for: where k' j,h represents the hth eigenvalue of the jth air combat target after normalization; Then the i-th air combat target x i Nearest neighbor boundary NNB(x i )for: NNB(x i )={(x,y)∣x i -d ij ≤x≤x i +d ij ,y i -d ij ≤y≤y i +d ij } (5) Among them, (x, y) is the area divided according to the boundary.

5. The method according to claim 4, characterized in that Step 2-2 includes: performing nearest neighbor search in the nearest neighbor boundary of each area according to the area divided in step 2-1, and finding paired mutual nearest neighbors in the nearest neighbor boundary of each area. The global mutual nearest neighbors indicate that two air combat targets are each other's nearest neighbors in the entire air combat target dataset.

6. The method according to claim 5, characterized in that Step 2-3 includes: setting the air combat target to be divided into p intention groups after the processing of step 2-1 to step 2-2, forming a set M = {M1, ..., M i ,…,M p }, i=1,2,…,p, where M p Represents the pth intention group, each intention group includes n i air combat targets, and each target is represented by an azimuth The five features of distance R, speed V, entry angle θ, and height H are specifically represented to obtain the reference feature matrix S0 of p intentions = (s i,h ) p×5 , where the baseline feature vector s of the i-th intention i The expression of is shown in formula (6): Among them, Φ i 、R_set i 、Θ i 、V_set i 、H_set i They respectively represent the azimuth set, distance set, entry angle set, speed set, and altitude set of the target group corresponding to the i-th combat intention.

Images