Large-scale target threat assessment method in air defense combat

Abstract

The invention discloses a large-scale target threat assessment method in air defense combat, and relates to the technical field of regional air defense, and the main points of the technical scheme are as follows: for a large-scale complex combat scene, a threat assessment method selected by an improved skyline method is proposed, and firstly, a clustering method is designed; a parameter adaptive density grid clustering method based on a fuzzy theory is provided, and large-scale targets are processed, so that threat assessment can be quickly carried out subsequently; secondly, a multi-level skyline selection method is provided, the complex processes of weight setting and the like of a traditional threat assessment method are avoided, and numerous interference factors such as personal preference and the like in the weight setting process are eliminated; and finally, a multi-index representation method of the radar map is established, and a basis is provided for a commander to make a targeted decision. Simulation comparison experiment results show that the threat assessment method designed by the invention is used for subsequent target distribution and firepower interception, and the finally achieved combat effectiveness is better.

Description

Technical Field

[0001] The present invention relates to the field of regional air defense technology, and more specifically, to a method for assessing the threat of large-scale targets in air defense operations. Background Art

[0002] Target threat assessment is a key link in combat command and control, and provides an important basis for the commander to make decisions in the next step. Especially in the process of air defense and anti-missile operations, there is a high degree of uncertainty in the number of enemy air attack weapons, the direction of attack, and air attack tactics and methods. In response to the enemy's reverse gradient attack, the defender is very likely to be in a passive position. How to maximize the use of limited air defense resources to achieve combat objectives requires the assessment of the threat level of different targets and the maximum use of defense resources. At the same time, in large-scale regional joint operations, the amount of information data is large and extremely complex, which places extremely high demands on the capabilities of the command and control system. Therefore, in large-scale regional joint air defense operations, on the one hand, it is necessary to solve the problem of how to use limited command and control resources to process a large amount of target data, and on the other hand, how to optimize the existing threat assessment method, and on the premise of appropriately abstracting and simplifying military issues without covering up the game nature of war issues, so as to accurately assess the threat of targets and give comprehensive assessment results, rather than a simple threat ranking list, so that commanders can make adaptive decisions based on the assessment results for different battlefield environments, and provide a basis for commanders to choose tactics and methods.

[0003] At present, when there are fewer incoming weapons, air defense weapons generally adopt the "first come, first served" evaluation and interception principle. However, in large-scale regional air defense, due to the large number of targets and different threat values, with limited defense resources, resources should be concentrated to intercept high-value targets first to achieve maximum resource utilization. At this time, threat assessment is transformed into a multi-attribute decision-making problem. The multi-attribute decision-making method quantifies the multi-attribute values, sets weight vectors for different attributes, and finally aggregates the multi-attributes through a monotonic weighting function to obtain a single value that represents the threat level of the target. This method overly relies on the setting and modification of attribute weights. During combat, commanders need to consider the attribute characteristics of the target in different combat scenarios and battlefield situations. However, this method is overly abstract, obscuring the game nature of combat problems and completely transforming military problems into mathematical problems. As a result, the final evaluation results are too one-sided and simple, unable to reflect the characteristics of multiple attributes of the target, making it difficult to provide commanders with a comprehensive basis for decision-making. When making decisions, commanders need more than just a simple threat assessment value, but decision-making information such as target attack capabilities (target type, mobility, interference capabilities), enemy intention judgment (flight altitude, formation attributes, route shortcuts), etc. Summary of the invention

[0004] The purpose of the present invention is to solve the above problems and provide a large-scale target threat assessment method in air defense operations. Aiming at large-scale complex combat scenarios, a threat assessment method for improving the skyline method selection is proposed. First, a clustering method is designed, and a parameter adaptive density grid clustering method based on fuzzy theory is proposed to process large-scale targets, so as to facilitate subsequent rapid threat assessment; secondly, a multi-level skyline selection method is proposed, which avoids the complex process of weight setting of traditional threat assessment methods and eliminates many interference factors such as personal preferences in the weight setting process; finally, a multi-index representation method of radar chart is established to provide a basis for commanders to make targeted decisions.

[0005] The above technical objectives of the present invention are achieved through the following technical solutions: a method for assessing the threat of large-scale targets in air defense operations, the method comprising the following steps:

[0006] S1: A parameter adaptive density grid clustering method based on fuzzy theory is used to cluster large scene targets;

[0007] A K-dimensional clustering method based on density grid is adopted to adaptively improve the algorithm parameters, and the concept of fuzzy logic is introduced through the ink drop diffusion algorithm, so that the parameter adaptive density grid clustering method based on fuzzy theory can simultaneously utilize the advantages of density-based and grid-based clustering and fuzzy logic to cluster the complex and large amount of target data that need to be processed;

[0008] S2: A multi-level skyline selection method is used to determine the threat level of multiple targets;

[0009] S3: Establish a multi-index representation method for radar charts to provide commanders with more comprehensive target information, present various attributes in a unified manner, and provide a basis for commanders to make targeted decisions.

[0010] The present invention is further configured as follows: the S1 is a K-dimensional clustering method based on a density grid, and the concept of fuzzy logic is introduced through an ink drop diffusion algorithm to obtain a fuzzy-based density grid clustering method, which specifically includes the following steps:

[0011] S11: Assume that the data input is k-dimensional and is specified in space as:

[0012] S=S1×S2×...×S K (1)

[0013] Where S i represents the i-th dimension, KIDS divides S into k-dimensional k-space units, and each S i Divided into P i units, The data space S is divided into Units, each unit is called a density unit and is denoted by g, g = 1, ..., N, where N is the number of density units, and data is mapped to density units g:

[0014]

[0015] g d It refers to the density unit of the mapping data record, where d = 1, ..., L is the number of mapping records;

[0016] S12: After mapping the data to the grid cells, apply the diffusion operator; define the neighborhood radius to spread the data in the neighborhood; for each g d , perform the diffusion process according to the following steps, and repeat L times;

[0017] (1) Selected g d One of the cells, this cell uses Then select the closest g d ,Right now:

[0018]

[0019] To give and g d The distance between Small values ​​of are considered to have a central The neighborhood radius is:

[0020]

[0021] (2) The data points are distributed within the radius in the form of ink droplet diffusion patterns. Find the density unit within the neighborhood radius. The neighborhood radius is:

[0022]

[0023] It is assumed that each data point contains information about its exact location and its neighborhood; except Unit, diffusion effect also appears In the surrounding density units;

[0024] (3) The density vector (D) is introduced and the diffusion operator is used to extend the influence of the data point to a radius of The effect is stored in the vector D, i.e. The effect of all grid cells in the neighborhood is equal to the vector D

[0025] (4) Each component of vector D is updated after the data is propagated around, that is:

[0026]

[0027] τ(.) is the grid update function, which initially sets the density vectors of all density cells to zero;

[0028] Assume that the data points are randomly distributed and there are no clusters, that is as well as have random and independent distributions; for each Follows a Gaussian distribution; random variable The probability density of is given by where μ is the mean or expected value of the distribution and σ is 2 is the variance:

[0029]

[0030]

[0031]

[0032] N is the number of density units, K is the number of attributes; The probability density function (pdf) is expressed as

[0033] f Y (y) = (N-1)g(y)(1-G(y)) N-2 (10)

[0034] in, and G(y) is the Gaussian cumulative distribution function (cdf);

[0035] (4) Determine D by threshold and select D so that each density unit with a value less than the threshold becomes zero;

[0036]

[0037] The present invention is further configured as follows: the improvement of the parameter-adaptive KIDS clustering algorithm in S1 comprises the following steps:

[0038] S1a: First, the following definitions are made:

[0039] Definition 1: The discreteness H of the j-th dimension data j , H j is the data dispersion of the j-th dimension of the data set, which can be calculated by the following formula

[0040]

[0041] S j and X_means j are the standard deviation and mean of the j-th dimension data respectively;

[0042] Definition 2: The standardized discreteness of the data set; its H s is the standardized discreteness of the data set, which can be calculated by the following formula

[0043]

[0044] Among them, d is the dimension of the data object in the dataset, H s The smaller the value, the greater the dispersion of the data set;

[0045] According to the above definition, it can be determined that the dimension segmentation parameter m can be adaptively calculated by the following formula:

[0046]

[0047] In the formula, d is the dimension of the data object in the data set, and N is the amount of data in the sample of the data set. This method can calculate the segmentation parameter m that is suitable for different data sets according to their data distribution, thereby determining the number of density units N.

[0048]

[0049] where max j 、min j are the maximum and minimum values ​​of the attribute value of the jth dimension in the data set s, respectively;

[0050] S1b: After calculating the dimension segmentation parameters, the data space is segmented by m to obtain the number of density units N. After the data is mapped to the unit grid, the density threshold D is used. th To determine whether the update requirement of the density vector D is met;

[0051] X i The grid cell where it is located is a sparse grid cell and X i The grid cell where it is located is not within the ink droplet diffusion range of the dense grid cell, then it is considered that X i is the noise data.

[0052] The present invention is further configured as follows: the multi-level skyline selection method in S2 selects a skyline member S from a given evaluation object set T as a subset S of the evaluation object set T, and then performs the following process: after the system is initialized, the first target in the set T is placed into the skyline set S i , the element t in T and S i The elements s in the matrix are compared one by one, and a dominance mode is selected according to the comparison result. The dominance modes include t dominating s, t and s do not dominate each other, and s dominates t. The t dominance s mode is that t is added to S i , s joins P, the non-dominant mode between t and s is t joins S i, the way s dominates t is that t is added to P and s is retained; after the domination is completed, let T be P, let P be empty, and judge whether T is empty. If T is empty, the process ends; if T is not empty, the process is repeated.

[0053] The present invention is further configured as follows: the multi-index representation method of the radar chart in S3 comprises the following steps:

[0054] S21: firstly, parameter quantization is performed to calculate the quantized values ​​of various attributes of the target, wherein the target attributes include target speed, route shortcut, jamming capability, stealth performance, and weapon damage capability;

[0055] S22: divide the skyline by multi-level skyline algorithm;

[0056] S23: Obtain the first-level, second-level and third-level skyline member graphs through skyline division.

[0057] In summary, the present invention has the following beneficial effects:

[0058] (1) In view of the characteristics of large-scale combat scenarios, such as large number of targets, multiple types, and extremely complex air situation, a parameter adaptive density grid clustering method based on fuzzy theory is introduced. This method combines the advantages of density clustering and grid clustering, improves the clustering efficiency and adaptability to different data scenarios, and introduces the ink droplet diffusion algorithm to realize multi-dimensional fuzzy clustering. At the same time, the algorithm parameters are adaptively improved during the clustering process to further improve the robustness of the algorithm. It works well on complex data sets. In the comparative experiments with other algorithms, it is verified that the method of the present invention has a faster clustering speed, which provides support for subsequent threat assessment.

[0059] (2) In order to solve the problems of traditional threat assessment methods, such as the difficulty in objectively evaluating the weight setting and the incompleteness of the evaluation results, the skyline dominance selection theory is introduced, and a multi-level skyline selection method is proposed on this basis. This method eliminates the difficulty of weight setting and avoids the secondary error transmission from motion state to threat degree and then to threat level.

[0060] (3) As the traditional skyline advantage selection method is difficult to clearly display the evaluation results in various dimensions at the same time, a representation method based on radar chart is designed to help commanders allocate targets according to different combat objectives and achieve the maximum combat objective.

[0061] (4) By assuming a regional joint air defense combat scenario and conducting a comparative simulation analysis with the traditional threat assessment method, target allocation and fire interception were carried out based on the threat assessment method designed by the present invention. Ultimately, the deduction results showed outstanding performance in four aspects: the consumption of the red interceptor, the interception hit rate, the degree of damage to the defended key points, and the interception of the blue cruise missile.

[0062] In modern warfare with high game confrontation, high environmental complexity and high real-time responsiveness, threat assessment has a significant impact on combat decision-making. The design method of the present invention solves the problem of low efficiency of threat assessment in the face of huge amounts of information to a certain extent. At the same time, in view of the rapidly changing battlefield situation and the rapid changes in target status, the design method of the present invention improves the adaptability of the model to complex air situations. It can effectively shorten the entire command and control process time and improve the command and control capability to a certain extent. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 is a flow chart of KIDS in an embodiment of the present invention;

[0064] Figure 2 It is a diffusion process step in the embodiment of the present invention;

[0065] Figure 3 is the relationship between the threshold and the number of clusters in the embodiment of the present invention;

[0066] Figure 4 It is the original distribution of the Chainlink dataset diagram in the embodiment of the present invention;

[0067] Figure 5 This is the KIDS clustering effect in the embodiment of the present invention;

[0068] Figure 6 It is a comparison of the operation of the algorithms in the embodiments of the present invention;

[0069] Figure 7 is a flow chart of an improved multi-level skyline selection method in an embodiment of the present invention;

[0070] Figure 8 is a first-level skyline member representation diagram in an embodiment of the present invention;

[0071] Fig. 9 is a secondary skyline member representation diagram in an embodiment of the present invention;

[0072] Fig.10 is a three-level skyline member representation diagram in an embodiment of the present invention;

[0073] Fig.11 It is a blue team approach reconnaissance map in an embodiment of the present invention;

[0074] Fig.12 It is a diagram of the interception implemented by the red team in the embodiment of the present invention;

[0075] Fig.13 is a first-level skyline member graph in a simulation experiment in an embodiment of the present invention;

[0076] Fig.14 is a secondary skyline member graph in a simulation experiment in an embodiment of the present invention;

[0077] Fig.15 is a three-level skyline member graph in a simulation experiment in an embodiment of the present invention;

[0078] Fig.16 is a four-level skyline member graph in a simulation experiment in an embodiment of the present invention;

[0079] Fig.17 In the embodiment of the present invention, the target is intercepted according to the evaluation result of scheme 1;

[0080] Fig.18 In the embodiment of the present invention, the target is intercepted according to the evaluation result of scheme 2;

[0081] Fig.19 is a comparison chart of interceptor consumption in an embodiment of the present invention;

[0082] Fig. 20 is a comparison chart of interceptor hit rates in an embodiment of the present invention;

[0083] Fig.21 is a comparison diagram of the degree of damage to the key defense area in the embodiment of the present invention;

[0084] Fig. 22 It is a comparison of the interception of the blue cruise missile in the embodiment of the present invention. DETAILED DESCRIPTION

[0085] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme of the present invention will be further described in detail below in conjunction with the embodiments of the present invention and the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in the field without creative work should fall within the scope of protection of the present invention.

[0086] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will be described in detail below in conjunction with the embodiments.

[0087] Example:

[0088] 1. If Figure 1 , Figure 2 As shown, a method for assessing large-scale target threats in air defense operations comprises the following steps:

[0089] S1: A parameter adaptive density grid clustering method based on fuzzy theory is used to cluster large scene targets;

[0090] A K-dimensional clustering method based on density grid is adopted to adaptively improve the algorithm parameters, and the concept of fuzzy logic is introduced through the ink drop diffusion algorithm, so that the parameter adaptive density grid clustering method based on fuzzy theory can simultaneously utilize the advantages of density-based and grid-based clustering and fuzzy logic to cluster the complex and large amount of target data that need to be processed.

[0091] 1. Fuzzy-based density grid clustering method

[0092] In fuzzy logic, sets are defined by their members and their membership in the set. This embodiment introduces the Ink Drop Spread (IDS) algorithm proposed by Shouraki et al. and processes K-dimensional data on this basis to achieve fuzzy processing of data.

[0093] In K-dimensional Ink Drop Spread (KIDS), a fuzzy set is assigned to each data point and all data point values ​​are fuzzified into membership functions. In the fuzzy subset, its elements are the neighborhood of the data point, and the membership decreases when the neighborhood radius increases. Therefore, not only the exact location of the data in the feature space is important, but also the neighboring areas are important because the data may also appear in the neighboring areas. This method is called diffusion. One of the main methods using the diffusion operator is the ink drop spread method (IDS).

[0094] Ink Drop Spread (IDS) is a soft computing method in which a Multiple Input Single Output (MISO) system is divided into single-input single-output subsystems, in each subsystem the output performance is examined and displayed on a two-dimensional screen called the IDS, and then the general behavior of the system is obtained from the overlap of these subsystems.

[0095] As shown in Table 1 and Figure 1 As shown in the figure, in KIDS, in order to solve the problem of high-dimensional data clustering, an improvement is made on the IDS method, and the MISO system is not divided into SISO subsystems. The data points are distributed in the K-dimensional feature space instead of the two-dimensional space. Diffusion and aggregation are performed in the K-dimensional space.

[0096] Table 1 KIDS algorithm process

[0097]

[0098]

[0099] Since it is difficult to keep all the original data, different methods and algorithms have been developed to minimize the amount of data to be stored. In this regard, KIDS divides the multidimensional data space into different cells and forms clusters in this k-dimensional grid. The grid cells are called density cells because their density ultimately determines the shape and position of the clusters.

[0100] S11: Assume that the data input is k-dimensional and is specified in space as:

[0101] S=S1×S2×...×S K (1)

[0102] Where S i represents the i-th dimension, KIDS divides S into k-dimensional k-space units, and each S i Divided into P i units, The data space S is divided into Units, each unit is called a density unit and is denoted by g, g = 1, ..., N, where N is the number of density units, and data is mapped to density units g:

[0103]

[0104] g d It refers to the density unit of the mapping data record, where d = 1, ..., L is the number of mapping records;

[0105] S12: After mapping the data to the grid cells, apply the diffusion operator; define the neighborhood radius to spread the data in the neighborhood; for each g d , perform the diffusion process according to the following steps, and repeat L times;

[0106] (1) Selected g d One of the cells, this cell uses Then select the closest g d ,Right now:

[0107]

[0108] To give and g d The distance between Small values ​​of are considered to have a central The neighborhood radius is:

[0109]

[0110] (2) The data points are distributed within the radius in the form of ink droplet diffusion patterns. Find the density unit within the neighborhood radius. The neighborhood radius is:

[0111]

[0112] It is assumed that each data point contains information about its exact location and its neighborhood; except Unit, diffusion effect also appears In the surrounding density units;

[0113] (3) The density vector (D) is introduced and the diffusion operator is used to extend the influence of the data point to a radius of The effect is stored in the vector D, i.e. The effect of all grid cells in the neighborhood is equal to the vector D

[0114] (4) Each component of vector D is updated after the data is propagated around, that is:

[0115]

[0116] τ(.) is the grid update function, which initially sets the density vectors of all density cells to zero;

[0117] Assume that the data points are randomly distributed and there are no clusters, that is as well as have random and independent distributions; for each Follows a Gaussian distribution; random variable The probability density of is given by where μ is the mean or expected value of the distribution and σ is 2 is the variance:

[0118]

[0119]

[0120]

[0121] N is the number of density units and K is the number of attributes. The minimum distance between data points predicts whether it is a random data distribution or not. If the data points are random variables, then they are almost uniformly distributed in the k-dimensional space, otherwise, the points in the space are more concentrated and the shortest distance between the points decreases. Therefore, The square of the minimum distance is a random variable, and its probability density function (pdf) is expressed as:

[0122] f Y (y) = (N-1)g(y)(1-G(y)) N-2 (10)

[0123] in, and G(y) is the Gaussian cumulative distribution function (cdf);

[0124] If we consider the probability of wrong decision (P fd ), then R th It can be inferred from the following formula:

[0125]

[0126] R th It is used to determine the distance between connected points and isolated points. If for data point R min、gds Greater than R th , then the point is likely to be randomly selected and not in any cluster. By choosing an appropriate P fd , we can get R th .

[0127] According to the selected data set, select P fd To achieve an appropriate threshold level (e.g., P fd =10 -4 ). Therefore, if the minimum distance between two points is less than R th =R min、gds , then they belong to a cluster and are therefore not randomly distributed. In addition, R min、gds The smaller it is, the more likely it is that two points are in the same cluster. Therefore, the update function seems to change with R min、gds A good choice for this function could be Since this function may not be optimal, other functions may be used in future research.

[0128] (4) Determine D by threshold and select D so that each density unit with a value less than the threshold becomes zero;

[0129]

[0130] Through (4), anomalies are identified and eliminated. In addition, the threshold is a control parameter for detecting disjoint clusters.

[0131] Diffusion process Figure 2 A simple example is shown in Figure 1. In this example a two-dimensional dataset with 4 samples is used. Figure 2 shows the first step of the diffusion process, where for each g d , find the closest g d , and select its distance as the neighborhood radius R min、gds We will use τ(g ds ) attributed to gds (yellow symbol) and its R min、gds This is called a drop pattern.

[0132] As droplet patterns overlap, the overlapping portions become increasingly effective. Vector D represents these effects.

[0133] 2. The improvement of the parameter adaptive KIDS clustering algorithm includes the following steps:

[0134] S1a: First, the following definitions are made:

[0135] Definition 1: The discreteness H of the j-th dimension data j , H j is the data dispersion of the j-th dimension of the data set, which can be calculated by the following formula

[0136]

[0137] S j and X_means j are the standard deviation and mean of the j-th dimension data respectively;

[0138] Definition 2: The standardized discreteness of the data set; its H s is the standardized discreteness of the data set, which can be calculated by the following formula

[0139]

[0140] Among them, d is the dimension of the data object in the dataset, H s The smaller the value, the greater the dispersion of the data set;

[0141] According to the above definition, it can be determined that the dimension segmentation parameter m can be adaptively calculated by the following formula:

[0142]

[0143] In the formula, d is the dimension of the data object in the data set, and N is the amount of data in the sample of the data set. This method can calculate the segmentation parameter m that is suitable for different data sets according to their data distribution, thereby determining the number of density units N.

[0144]

[0145] where max j 、min j are the maximum and minimum values ​​of the attribute value of the jth dimension in the data set s, respectively;

[0146] S1b: After calculating the dimension segmentation parameters, the data space is segmented by m to obtain the number of density units N. After the data is mapped to the unit grid, the density threshold D is used. thTo determine whether the update requirement of the density vector D is met. The determination of the density threshold can effectively filter out the influence of noise data and define noise data.

[0147] X i The grid cell where it is located is a sparse grid cell and X i The grid cell where it is located is not within the ink droplet diffusion range of the dense grid cell, then it is considered that X i For noise data

[0148] The amount of noise data obtained under different density thresholds is different. Similarly, due to different data types, the noise data in different data environments is also different. It is necessary to select a suitable density threshold according to different data environments to obtain a better processing effect. The method of this embodiment is to count the number of noise samples corresponding to different density thresholds before clustering, draw a density threshold-noise number curve, and select the best density threshold through the curve.

[0149] Figure 3 shows how changing the threshold level affects the number of clusters. The plot shows that for different values ​​of N, the threshold level should not be large (less than 0.2).

[0150] 3. Simulation analysis and verification

[0151] The Chainlink dataset is used for verification. This dataset is a challenging dataset because it has interconnected transactions and places extremely high demands on the clustering ability of the algorithm. The clustering results are as follows: Figure 4 , Figure 5 shown.

[0152] The running time of this algorithm is compared with that of FUALM, DBSCAN, DENCLUE and FN-DBSCAN with respect to the number of clusters, as shown in Figure 6 As shown, as the number of clusters increases, the execution time of all methods does not increase and remains almost unchanged. Figure 6 It is also determined that the execution time of DBSCAN, FN-DBSCAN, and DENCLUE is at least one order of magnitude higher than that of KIDS and FUALM.

[0153] Experiments have shown that the clustering method of the present invention has obvious advantages in running speed and accuracy.

[0154] S2: A multi-level skyline selection method is used to determine the threat level of multiple targets;

[0155] Multi-level skyline selection method such as Figure 7As shown, by selecting a skyline member S from a given evaluation object set T as a subset S of the evaluation object set T, the following process is then performed: After the system is initialized, the first target in the set T is placed into the skyline set S i , the element t in T and S i The elements s in the matrix are compared one by one, and a dominance mode is selected according to the comparison result. The dominance modes include t dominating s, t and s do not dominate each other, and s dominates t. The t dominance s mode is that t is added to S i , s joins P, the non-dominant mode between t and s is t joins S i , the way s dominates t is that t is added to P and s is retained; after the domination is completed, let T be P, let P be empty, and judge whether T is empty. If T is empty, the process ends; if T is not empty, the process is repeated

[0156] S3: Establish a multi-index representation method for radar charts to provide commanders with more comprehensive target information, present various attributes in a unified manner, and provide a basis for commanders to make targeted decisions;

[0157] The multi-index representation method includes the following steps:

[0158] S21: firstly, parameter quantization is performed to calculate the quantized values ​​of various attributes of the target, wherein the target attributes include target speed, route shortcut, jamming capability, stealth performance, and weapon damage capability;

[0159] S22: divide the skyline by multi-level skyline algorithm;

[0160] S23: Obtain the first-level, second-level and third-level skyline member graphs through skyline division.

[0161] The present invention describes the multi-index representation method of radar chart in detail through an example:

[0162] (1) Scenario Design

[0163] Assume that in an air defense operation, there are 10 incoming targets. After the sensor resources identify and track them, the target information is obtained as shown in Table 1:

[0164] Table 2 Incoming target information

[0165] Target speed Route shortcuts Interference ability Stealth performance Weapon damage capability <![CDATA[t1]]> 1.0 9.5 generally weak powerful <![CDATA[t2]]> 2.0 8.7 generally generally powerful <![CDATA[t3]]> 2.5 8 generally generally Very strong <![CDATA[t4]]> 1.2 10 Very strong Very strong weak <![CDATA[t5]]> 1.6 9.6 Very strong Very strong weak <![CDATA[t6]]> 2.3 8 Very strong Very strong weak <![CDATA[t7]]> 1.1 6.5 powerful weak powerful <![CDATA[t8]]> 1.3 5.9 powerful weak Very strong <![CDATA[t9]]> 0.8 5.7 powerful weak Very strong <![CDATA[t 10 ]]> 0.9 5.6 Very strong weak powerful

[0166] The steps for using the multi-level skyline method to conduct threat assessment on 10 incoming targets are as follows:

[0167] Step 1 first performs the parameter quantization method to calculate the quantitative values ​​of each attribute of the target. The specific method is as follows.

[0168] 1) Target speed: The higher the speed of the air raid target, the lower the accuracy and stability of the radar tracking it, and the lower the possibility of interception. In this embodiment, the membership function of the target is expressed as:

[0169]

[0170] Among them, k v is the target speed gain coefficient, which is generally taken as the typical value k v =-0.1Mach -2 .

[0171] 2) Route shortcut: The threat level and attack intention of the target are reflected by the route shortcut. If the route shortcut is smaller, the target's attack intention is more obvious and the threat level is greater. The present invention expresses the membership function of the target route shortcut as

[0172]

[0173] k p is the route shortcut attenuation coefficient, which reflects the attack threat range of the target and is generally taken as k p =5×10 -3 km -2 .

[0174] 3) Route shortcuts: Interference capability: When the target interferes with our sensors, it will put extremely high demands on the search and tracking difficulty of our radar. It is also an important indicator for judging the degree of threat assessment and is quantified as shown in the following table.

[0175] Table 3 Electronic jamming capability parameters

[0176]

[0177] 4) Stealth performance: The stealth performance of the target poses a great challenge to our detection capability. The higher the stealth performance of the target, the more difficult it is to detect, and the stronger the target's penetration and survivability capabilities are, which will pose a sudden threat to us.

[0178] Table 4 Stealth performance parameters

[0179]

[0180] 5) Weapon destructive capability

[0181] The greater the destructive power of the target weapon, the greater the threat to our defense. The quantitative results are shown in the following table.

[0182] Table 5 Weapon damage capability parameters

[0183]

[0184] The target initial decision matrix is ​​obtained by parameter calculation

[0185]

[0186] Step 2 uses a multi-level skyline algorithm to divide the skyline. The division result is:

[0187] U=(3,2,1,3,2,1,3,2,1)

[0188] The smaller the number, the higher the level. The members of the first-level skyline are U1 = {t3, t6, t9, t 10}, the second-level skyline members are U2 = {t2, t5, t8}, and the third-level skyline members are U3 = {t1, t4, t7}.

[0189] Step 3: Figures 8 to 10 As shown, the first-level, second-level and third-level skyline member graphs are obtained through skyline division.

[0190] The graphic display method applicable to radar charts can provide commanders with more comprehensive target information and present various attributes in a unified manner, making it easier for commanders to make combat decisions that are more in line with the battlefield environment.

[0191] 2. Simulation and Verification

[0192] In order to verify the practical application effect of this method, it is necessary to conduct simulation verification in combat experiments. This embodiment uses a combat simulation platform to design combat scenarios, conduct combat experiments, conduct simulation verification of the method of the present invention, and conduct a relatively complete analysis of the model method of the present invention based on the simulation results, and conduct effective verification.

[0193] 1. Operational scenario setting

[0194] The large-scale regional joint combat scenario is mainly based on the air defense of key points. In view of the complexity of the air situation, it is necessary to set the combat intentions of both sides, enemy and friendly situation information, and combat mission settings.

[0195] (1) Blue Team Scenario

[0196] 1) Blue's intention

[0197] The Blue Side has assembled large-scale naval and air forces in the coastal areas of the Red Side. Its naval and air force combat intentions are: relying on the local naval and air force advantages, using large-scale cruise missiles, drone swarms, high-performance fighters, bombers and early warning aircraft to carry out multi-directional, multi-batch, saturation strikes on important strategic targets such as the Red Side's airports, command and control centers, and air defense positions, thereby weakening our combat power.

[0198] 2) Composition of participating forces

[0199] The air combat force consists of 48 third-generation fighters, 24 fourth-generation fighters, 12 carrier-based electronic attack aircraft, 3 early warning aircraft, 2 stealth bombers, 450 swarm drones, and 400 cruise missiles.

[0200] 3) Tactics and tactics

[0201] When the Blue Army launches an air strike against the Red Army, it first uses saturation cruise missiles and drone swarm formations to consume the Red Army's air defense resources. It then assembles fighter jets, bombers, etc. to penetrate the Red Army's depth under the cover of early warning aircraft, electronic jammers, etc., and uses air-to-ground missiles to attack strategic targets.

[0202] (2) Red Team Scenario

[0203] 1) Red’s intention

[0204] The combat intentions of the air defense forces in the Red area are mainly concentrated on ground-based and sea-based air defense forces to protect the important targets of the Red side and crush the combat intentions of the Blue side.

[0205] 2) Composition of participating forces

[0206] Ground air defense forces: equipped with 12 sets of medium- and long-range surface-to-air missiles, 3 sets of short-range air defense missiles, and 1 set of anti-radiation missiles.

[0207] Sea-based air defense force: mainly equipped with 4 destroyers of various types.

[0208] Air defense force: 30 fighter jets of various types.

[0209] Early warning support force: 2 early warning aircraft

[0210] 2. Simulation analysis

[0211] In order to verify the target threat assessment effectiveness of the method proposed in the present invention in the confrontation game scenario, a comparative analysis was conducted between the threat assessment performed by the method of the present invention and the traditional multi-attribute threat assessment method in this scenario according to the combat assumptions.

[0212] Solution 1: Large-scale target threat assessment based on the improved skyline method. First, the large-scale targets are grouped using the parameter adaptive density grid clustering method based on fuzzy theory. Then, the threat assessment of the targets is performed using the improved multi-level skyline selection method, and the radar chart method is used to provide the commander with a basis for decision-making.

[0213] Option 2: Perform threat assessment directly through traditional multi-attribute decision-making methods.

[0214] Evaluation indicators: First, analyze the threat assessment efficiency of the two schemes, then allocate targets and intercept according to the two schemes, and analyze the final red team interception effectiveness.

[0215] (1) Situation Analysis

[0216] The simulation experiment was performed on a Core i7-7800X, CPU 3.50 GHz, 16 GB RAM, and Windows 11 experimental platform.

[0217] Pre-war deployment is carried out through a simulation platform, and the red side's detection and air interception range after deployment meets the basic requirements of regional joint air defense operations.

[0218] During the simulation, in Scheme 1, the red side gives the threat assessment result according to the threat assessment method designed by the present invention and then performs target allocation and fire interception operations, while the blue side uses the AI ​​operation provided by the simulation platform; in Scheme 2, the red and blue armies are controlled by AI according to the preset engagement regulations, and the training level of both sides is set to "top".

[0219] The blue team conducts close reconnaissance and the red team conducts interception. Figure 11-12 shown.

[0220] (2) Operational simulation process

[0221] The simulation system was deduced under two sets of schemes respectively. The following is the deduction process.

[0222] The deduction process of scenario 1

[0223] First, for large-scale scenarios, the command and control granularity of the regional command and control system is relatively large, and a large number of targets need to be clustered and processed. The enemy air targets are processed by the KIDS clustering method based on parameter adaptation proposed in this invention. Since the attribute performance of each target in each cluster is also different, we use the average value of each attribute index of the cluster to represent the index of each attribute of the cluster. The processing results are shown in Table 6:

[0224] Table 6 Target information of the attacking group

[0225]

[0226]

[0227]

[0228] The evaluation indicators are parameterized and used as input for the subsequent multi-level skyline method.

[0229] Solve the target threat level through the multi-level skyline method

[0230] U=(4,3,1,2,1,1,2,1,1,1,3,2,1,4,3,1)

[0231] Among them, the members of the first-level skyline are U1 = {t3, t5, t6, t8, t9, t 10 ,t 13 ,t 16}, the members of the secondary skyline are U2 = {t4, t7, t 12}, the members of the third-level skyline are U3={t2,t 15 ,t 11}, the members of the fourth-level skyline are U3={t1,t 14}.

[0232] Figure 1.10: Skyline members at all levels Figure 13-16 shown.

[0233] The Red Team allocates and intercepts targets based on the multi-level skyline output results. The real-time interception process is as follows: Fig.17 shown.

[0234] The combat consumption of the Red Army in Scheme 1 is shown in Table 7, and the impact on the Red Army’s strategic targets is shown in Table 8.

[0235] Table 7 Red Army Combat Consumption Situation in Scheme 1

[0236]

[0237]

[0238] Table 8: The situation of the Red Army's strategic targets being hit in Plan 1

[0239]

[0240] The blue side's combat losses in Scheme 1 are shown in Table 9, and the blue side's combat consumption is shown in Table 10

[0241] Table 9: Blue Force Battle Damage Situation for Scheme 1

[0242]

[0243] Table 10: Blue Force Combat Consumption Situation in Scheme 1

[0244]

[0245] 2) Option 2

[0246] Solution 2 uses the first-come-first-served and multi-attribute decision-making method to rank the threats of the targets (the ranking results are shown in Table 11), and allocates interceptions based on the ranking results. The interception diagram is shown in Fig.18 shown.

[0247] Table 11 List of traditional target threat assessments

[0248]

[0249]

[0250] The combat consumption of the Red Army in Plan 2 is shown in Table 12, and the impact on the Red Army’s strategic targets is shown in Table 13.

[0251] Table 12 Red Army Combat Consumption Situation in Scheme 2

[0252]

[0253] Table 13: Situation 2: The impact of the Red Army's strategic targets

[0254]

[0255] The blue side’s combat damage and consumption in Plan 2 is shown in Figure 14, and the blue side’s combat consumption is shown in Table 15.

[0256] Table 14: Blue Force Battle Damage Situation in Scheme 2

[0257]

[0258] Table 15: Blue Force Combat Consumption Situation in Scheme 2

[0259]

[0260]

[0261] (3) Analysis of Threat Assessment Results

[0262] According to the comparison of the Red Army's combat damage and consumption in Scheme 1 and Scheme 2, it can be concluded that the Red Army consumes less resources for allocating interception operations under the technology proposed in the present invention, that is, the parameter adaptive clustering and multi-level skyline selection and radar chart illustration method based on the KIDS method, and the strategic locations are less vulnerable to attacks, and the most important command center is less vulnerable to attacks, thus better maintaining the overall combat effectiveness.

[0263] From Table 12, Fig.19 and Fig. 20It can be found that the number of air-to-air missiles consumed in Scheme 2 is very high, accounting for a large proportion of the total ammunition consumption, and the hit rate of advanced medium and long-range air defense missiles is extremely low. According to the previous behavioral analysis, it is mainly due to the use of first-come-first-served and weighted multi-attribute decision-making methods under the conditions of Scheme 2, which unilaterally sorts the threats of air strike targets, and then guides the aviation fighters to actively intercept the blue party’s large-scale cruise missiles after detecting them, and launches a large number of air-to-air missiles, which leads to an air combat disadvantage in the face of the blue party’s fighter jets’ penetration; and the advanced medium and long-range surface-to-air missiles also launch missiles at the blue party’s cruise missiles when tracking them, and fall into a disadvantage when facing subsequent high-value targets of the blue party, which seriously restricts its combat efficiency. As shown in Tables 10 and 14, the red party under the control of Scheme 1 has a stronger strike capability against the blue party’s air strike platform and can cause greater damage; combined with Tables 10, 14 and Fig. 22 , the Red Army combat units under the control of Scheme 1 have a higher success rate in intercepting the Blue Army's guided weapons, consume less ammunition, and have higher combat effectiveness.

[0264] like Fig.19 , Fig. 20 , the combat unit under Scheme 1 consumes less ammunition when the number of hits on incoming targets is similar, and the number of interceptor self-destructions is significantly reduced compared to Scheme 2, which improves the efficiency of interceptor resource utilization. Fig.21 Compared with the situation of the Red Army's strategic locations being attacked, it is obvious that the design method of the present invention can achieve better threat assessment results when facing large-scale, multi-batch, and multi-directional incoming targets in current scenario combat, thereby providing a better allocation basis for subsequent target allocation and fire interception, and ultimately improving interception efficiency, combat cost-effectiveness, and defense intensity.

[0265] This specific embodiment is merely an explanation of the present invention and is not a limitation of the present invention. After reading this specification, those skilled in the art may make non-creative modifications to the present embodiment as needed. However, as long as they are within the scope of the claims of the present invention, they are protected by the patent law.

Claims

1. A method for assessing the threat of large-scale targets in air defense operations, characterized by: The method comprises the following steps: S1: A parameter adaptive density grid clustering method based on fuzzy theory is used to cluster large scene targets; A K-dimensional clustering method based on density grid is adopted to adaptively improve the algorithm parameters, and the concept of fuzzy logic is introduced through the ink drop diffusion algorithm, so that the parameter adaptive density grid clustering method based on fuzzy theory can simultaneously utilize the advantages of density-based and grid-based clustering and fuzzy logic to cluster the complex and large amount of target data that need to be processed; S2: A multi-level skyline selection method is used to determine the threat level of multiple targets; S3: Establish a multi-index representation method for radar charts to provide commanders with more comprehensive target information, present various attributes in a unified manner, and provide a basis for commanders to make targeted decisions.

2. The method for assessing large-scale target threats in air defense operations according to claim 1 is characterized by: The S1 is a K-dimensional clustering method based on a density grid, which introduces the concept of fuzzy logic through an ink drop diffusion algorithm to obtain a fuzzy density grid clustering method, which specifically includes the following steps: S11: Assume that the data input is k-dimensional and is specified in space as: S=S1×S2×...×S K (1) Where S i represents the i-th dimension, KIDS divides S into k-dimensional k-space units, and each S i Divided into P i units, The data space S is divided into Units, each unit is called a density unit and is denoted by g, g = 1, ..., N, where N is the number of density units, and data is mapped to density units g: g d It refers to the density unit of the mapping data record, where d = 1, ..., L is the number of mapping records; S12: After mapping the data to the grid cells, apply the diffusion operator; define the neighborhood radius to spread the data in the neighborhood; for each g d , perform the diffusion process according to the following steps, and repeat L times; (1) Selected g d One of the cells, this cell uses Then select the closest g d ,Right now: To give and g d The distance between Small values ​​of are considered to have a central The neighborhood radius is: (2) The data points are distributed within the radius in the form of ink droplet diffusion patterns. Find the density unit within the neighborhood radius. The neighborhood radius is: It is assumed that each data point contains information about its exact location and its neighborhood; except Unit, diffusion effect also appears In the surrounding density units; (3) The density vector (D) is introduced and the diffusion operator is used to extend the influence of the data point to a radius of The effect is stored in the vector D, i.e. The effect of all grid cells in the neighborhood is equal to the vector D (4) Each component of vector D is updated after the data is propagated around, that is: τ(.) is the grid update function, which initially sets the density vectors of all density cells to zero; Assume that the data points are randomly distributed and there are no clusters, that is as well as have random and independent distributions; for each Follows a Gaussian distribution; random variable The probability density of is given by where μ is the mean or expected value of the distribution and σ is 2 is the variance: N is the number of density units, K is the number of attributes; The probability density function (pdf) is expressed as f Y (y)=(N-1)g(y)(1-G(y)) N-2 (10) in, and G(y) is the Gaussian cumulative distribution function (cdf); (4) Determine D by threshold and select D so that each density unit with a value less than the threshold becomes zero; 3. The method for assessing large-scale target threats in air defense operations according to claim 2 is characterized by: Improvement of the parameter-adaptive KIDS clustering algorithm in S1 The following steps are involved: S1a: First, the following definitions are made: Definition 1: The discreteness H of the j-th dimension data j , H j is the data dispersion of the j-th dimension of the data set, which can be calculated by the following formula S j and X_means j are the standard deviation and mean of the j-th dimension data respectively; Definition 2: The standardized discreteness of the data set; its H s is the standardized discreteness of the data set, which can be calculated by the following formula Among them, d is the dimension of the data object in the dataset, H s The smaller the value, the greater the dispersion of the data set; According to the above definition, it can be determined that the dimension segmentation parameter m can be adaptively calculated by the following formula: In the formula, d is the dimension of the data object in the data set, and N is the amount of data in the sample of the data set. This method can calculate the segmentation parameter m that is suitable for different data sets according to their data distribution, thereby determining the number of density units N. where max j 、min j are the maximum and minimum values ​​of the attribute value of the jth dimension in the data set s, respectively; S1b: After calculating the dimension segmentation parameters, the data space is segmented by m to obtain the number of density units N. After the data is mapped to the unit grid, the density threshold D is used. th To determine whether the update requirement of the density vector D is met; X i The grid cell where it is located is a sparse grid cell and X i The grid cell where it is located is not within the ink droplet diffusion range of the dense grid cell, then it is considered that X i is the noise data.

4. The method for assessing large-scale target threats in air defense operations according to claim 1 is characterized by: The multi-level skyline selection method in S2 selects a skyline member S from a given evaluation object set T as a subset S of the evaluation object set T, and then performs the following process: After the system is initialized, the first target in the set T is placed in the skyline set S i , the element t in T and S i The elements s in the matrix are compared one by one, and a dominance mode is selected according to the comparison result. The dominance modes include t dominating s, t and s do not dominate each other, and s dominates t. The t dominance s mode is that t is added to S i , s joins P, the non-dominant mode between t and s is t joins S i , the way s dominates t is that t joins P and s is retained; After the control is completed, let T be P, let P be empty, and determine whether T is empty. If T is empty, the process ends; if T is not empty, the process is repeated.

5. The method for assessing large-scale target threats in air defense operations according to claim 1 is characterized by: The multi-index representation method of the radar chart in S3 comprises the following steps: S21: firstly, parameter quantization is performed to calculate the quantized values ​​of various attributes of the target, wherein the target attributes include target speed, route shortcut, jamming capability, stealth performance, and weapon damage capability; S22: divide the skyline by multi-level skyline algorithm; S23: Obtain the first-level, second-level and third-level skyline member graphs through skyline division.

Images