Air interference strategy optimization method based on decision tree

Abstract

The invention discloses an air interference strategy optimization method based on a decision tree, and the method comprises the following steps: S1, data collection and preprocessing: collecting enemy plane information, seeker parameters, interference resource information and electromagnetic environment information, and carrying out the data cleaning and normalization processing, so as to obtain the feature input of a decision tree model; s2, constructing a decision tree model: based on a CART algorithm, splitting nodes by using a Gini index, and generating the decision tree model containing key features; s3, interference strategy generation: according to real-time battlefield data input, generating an adaptive interference strategy through a decision tree model; and S4, interference parameter binding and optimization: performing parameter regulation and control of frequency, power, delay and duration on the generated interference strategy to ensure that the effect of the interference strategy is optimal in a battlefield environment. Through the steps of data collection, decision tree model construction, air interference strategy generation and optimization and the like, the decision ability of an air defense system and the adaptability of an interference strategy are improved.

Description

Technical Field

[0001] The invention belongs to the technical field of electronic countermeasures, and in particular relates to an air jamming strategy optimization method based on a decision tree. Background Art

[0002] In modern air defense operations, facing complex and ever-changing air threats, effective air jamming strategies are needed. However, traditional air jamming strategy formulation methods often rely on experience or simple rules and fail to fully consider the interactive effects of multiple factors, such as enemy aircraft type, flight attitude, radar parameters, jammer performance, and complex electromagnetic environment. This results in poor jamming effects in actual operations and difficulty in dealing with diverse air threats. The following limitations are often shown: (1) Single factor decision-making problem: Traditional jamming strategies are often formulated based on a single or a few factors, such as enemy aircraft type or distance, ignoring the interaction between multiple factors. In a complex battlefield environment, factors such as the enemy aircraft's flight attitude, radar parameters, and seeker working status will affect the jamming effect. The single factor decision-making method cannot meet the needs of modern operations. (2) Low efficiency in strategy optimization: Existing methods often rely on manual intervention and experience accumulation to adjust jamming strategies, resulting in a lengthy and inefficient decision-making process. Especially in the ever-changing battlefield environment, inefficient strategy optimization methods often lead to the inability of interference effects to adapt to actual needs in a timely manner. (3) Insufficient adaptability and robustness: Traditional interference strategies lack sufficient adjustment capabilities when information is inaccurate or the environment changes, resulting in poor robustness of air defense strategies and difficulty in coping with complex situations in actual combat. Summary of the invention

[0003] Technical purpose: To solve the above technical problems, the present invention proposes a method for optimizing air jamming strategies based on a decision tree, which improves the decision-making ability of the air defense system and the adaptability of the jamming strategy.

[0004] Technical solution: To achieve the above technical objectives, the present invention adopts the following technical solution:

[0005] A method for optimizing an air jamming strategy based on a decision tree, characterized in that it comprises the following steps:

[0006] S1. Data collection and preprocessing: Collect enemy aircraft information, seeker parameters, jamming resource information and electromagnetic environment information, and perform data cleaning and normalization to obtain feature inputs for the decision tree model;

[0007] S2. Decision tree model construction: Based on the CART algorithm, the Gini index is used to split the nodes to generate a decision tree model containing key features;

[0008] S3, jamming strategy generation: Generate adaptive jamming strategies through decision tree models based on real-time battlefield data input;

[0009] S4. Interference parameter binding and optimization: The generated interference strategy is subjected to parameter adjustment of frequency, power, delay and duration to ensure the optimal effect of the interference strategy in the battlefield environment.

[0010] The technical solution of the present invention is further defined as follows: the collection of enemy aircraft information, seeker parameters, jamming equipment characteristics and electromagnetic environment information specifically includes:

[0011] Enemy aircraft information used to analyze the enemy aircraft's action intentions and its response to jamming strategies: enemy aircraft basic flight parameters, including model, speed, altitude, and heading; radar performance information including: frequency, power, scanning mode, etc.

[0012] Seeker parameters used to help the decision tree model determine the most appropriate jamming strategy, including information related to the enemy seeker's operating status, characteristics, and current environment;

[0013] Electromagnetic environment information for selecting appropriate interference methods based on complex electromagnetic environments: including the electromagnetic noise level in the current war zone, the distribution of other radio signals, etc.;

[0014] Interference resource information to evaluate the availability and adaptability of current interference resources: including the model, interference power, and frequency range of our jammers.

[0015] Furthermore, the decision tree model building step includes:

[0016] S21. Use information gain to calculate the importance of each feature, so as to select the key features that have the greatest impact on the generation of the decision tree, including the enemy aircraft's radar frequency, missile distance, seeker working mode, electromagnetic environment complexity, and interference resource remaining power, to form a feature subset;

[0017] S22. Use the CART algorithm to calculate the Gini index of each feature in the feature subset, select the feature with the highest information gain as the split node, and generate a node splitting path; construct a decision path based on different seeker working systems and signal forms until reaching the leaf node to generate a specific interference strategy.

[0018] Furthermore, the step S22 is specifically implemented as follows:

[0019] (1) When constructing a decision tree, the CART (Classification and Regression Tree) algorithm is used as the basic algorithm of the model. At each splitting node, the CART algorithm recursively calculates the Gini index and selects the eigenvalue with the lowest impurity for splitting;

[0020] (2) When generating a node, a loss function based on the misclassification rate is used to evaluate the splitting effect, thereby deciding whether to continue splitting or to treat the current node as a leaf node;

[0021] (3) Starting from the root node, each layer selects the optimal features for splitting according to the Gini index, and gradually constructs a decision path; for each leaf node on the path, a specific interference strategy output is finally generated.

[0022] Furthermore, the interference strategy generation step in step S3 is specifically as follows:

[0023] S31, receiving real-time data from sensors and external intelligence, inputting it into the root node of the decision tree model, and triggering the automatic decision-making process of the model;

[0024] S32, according to the characteristic value of the real-time data, the strategy generation module splits the node path of the decision tree in sequence, and screens downward from the root node layer by layer until reaching the leaf node;

[0025] S33. When the path reaches the leaf node, if the required interference style is single interference, the interference type that best suits the current battlefield needs is directly selected; in the case of multiple targets or complex threats, the leaf node outputs a combined interference strategy.

[0026] Furthermore, step S3 of generating an interference strategy also includes an interference parameter optimization step, specifically:

[0027] (1) Adjust the jamming frequency and power to match the frequency range of the enemy aircraft's detection signal;

[0028] (2) Control interference delay and duration.

[0029] Furthermore, in the interference parameter optimization step, the Monte Carlo sorting method generates a large number of strategy samples by simulating a variety of possible interference parameter combinations to evaluate the effectiveness of each strategy in the current battlefield environment. The specific application process is as follows:

[0030] Generate interference parameter samples: In the "interference parameter optimization" step, the Monte Carlo method is used to sample different interference parameter combinations to generate a series of interference strategy samples;

[0031] Effect evaluation: Evaluate the effect of each sampling strategy and simulate the performance of the strategy under different environmental variables;

[0032] Select the optimal strategy combination: Through the evaluation results of a large amount of sample data, the Monte Carlo method finally selects the interference parameter combination with the best effect for the current strategy.

[0033] Furthermore, after the interference strategy is generated in step S33, the interference strategy is output: an interference strategy instruction including detailed parameters such as specific interference mode, frequency, power, duration, etc. is output.

[0034] Beneficial effects:

[0035] The method of the present invention improves the decision-making ability of the air defense system and the adaptability of the interference strategy through the steps of data collection, decision tree model construction, air interference strategy generation and optimization, etc.; after the real-time combat data is input, the interference strategy can be automatically selected and optimized, and the efficiency and accuracy of strategy generation are effectively improved; the defect that the traditional method cannot effectively and comprehensively consider various air interference related factors, accurately analyzes the relationship between enemy aircraft characteristics, our interference resources and complex environment, and realizes air interference strategy decision-making based on multiple factors; in a complex strategy space, the optimal or near-optimal air interference strategy is quickly found, and tedious manual trial and error and calculation are avoided, and the timeliness of strategy generation is improved to adapt to the ever-changing battlefield environment; the air interference strategy can be adapted to different combat scenarios and air threat changes, and the effectiveness of the interference strategy can be guaranteed even when some information is inaccurate or the environment changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 A technical flow chart of the air interference strategy optimization method based on a decision tree provided by the present invention.

[0037] Figure 2 An example diagram of building a decision tree model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0039] Embodiment 1

[0040] This technical solution is based on a decision tree model, which aims to use machine learning methods to solve multi-factor decision-making problems. By building a decision tree model, we can comprehensively consider the characteristics of enemy aircraft, jamming resources and electromagnetic environment factors, and quickly find the best in a complex strategy space to generate an effective air jamming strategy.

[0041] In this method, the decision tree selects features and splits nodes based on the collected enemy features (such as aircraft model, radar frequency, etc.), electromagnetic environment parameters, our jamming capabilities, etc., so as to generate decision paths layer by layer and finally output the optimal jamming strategy at the leaf node. This strategy generation method can dynamically adapt to complex battlefield environments and automatically adjust the jamming plan according to real-time input data.

[0042] The specific implementation steps are as follows:

[0043] 1. Data Collection and Preprocessing

[0044] 1. Data collection module: Collects multi-dimensional data required for air jamming strategies from a variety of sensors and intelligence sources, covering key characteristics of both the enemy and our side.

[0045] Includes: Enemy aircraft information: collects basic flight parameters such as enemy aircraft model, speed, altitude, heading, and radar performance information such as radar frequency, power, scanning mode, etc. These data are used to analyze the enemy aircraft's action intentions and its response to interference strategies.

[0046] Seeker parameters: include information related to the working state, characteristics and current environment of the enemy seeker, which is used to help the decision tree model determine the most appropriate jamming strategy. Specifically: Seeker working state: the current working mode of the seeker, such as search, tracking, imaging, locking, etc. These states determine the type and strength of the jamming strategy. Seeker system: the working principle and signal characteristics of the seeker, including active radar, passive radar, infrared, laser and other different systems, and different systems have different response sensitivities to jamming. Signal form: the type of signal transmitted or received by the seeker, such as pulse wave, continuous wave, frequency modulation signal, etc. Different signal forms have an impact on the selection of jamming frequency and jamming style. Detection band: the frequency or band (such as X-band, L-band, mid-infrared, near-infrared, etc.) of the seeker, which determines the matching requirements of the jamming frequency range. Detection distance and threat level: according to the current distance between the seeker and the jamming device, and the degree of its threat to us (high, medium, low threat level), it helps to optimize the priority and jamming method of the jamming strategy. The collection and analysis of these seeker parameters help the decision tree model to be more targeted in strategy selection, ensuring that the generated interference strategy can maximize the effect and adapt to the real-time battlefield environment.

[0047] Electromagnetic environment information: Obtain information such as the electromagnetic noise level in the current war zone and the distribution of other radio signals, so as to select appropriate interference methods based on the complex electromagnetic environment.

[0048] Interference resource information: including the model, interference power, frequency range and other parameters of our jammer to evaluate the availability and adaptability of current interference resources.

[0049] The data collection in this embodiment will be based on the changes in the real-time battlefield environment. The system will update the data in each time window to ensure that the decision tree model makes strategic inferences and adjustments based on the latest intelligence.

[0050] 2. Data preprocessing module: cleans and normalizes the collected data, specifically:

[0051] The raw data is cleaned and normalized to unify parameters of different units and magnitudes, such as converting signal parameters of different frequency units to the same standard range.

[0052] For missing data and abnormal data, interpolation method is used to complete or outlier elimination method is used to ensure the integrity and accuracy of the data, enhance the robustness of the model, improve data quality, and provide a reliable data foundation for the subsequent training and application of decision tree models.

[0053] Smoothing is adopted for dynamic data such as electromagnetic environment noise level to avoid strategy deviation caused by noise fluctuations.

[0054] 2. Decision Tree Model Construction

[0055] 1. Feature selection: From the preprocessed data, the importance of each feature is calculated using information gain, so as to screen out the key features that have the greatest influence on the generation of the decision tree. These features serve as the input nodes of the decision tree model and determine the hierarchical structure and splitting method of the tree.

[0056] Based on the interference principle and the combat application principle of equipment, feature selection is carried out to screen out key features, including the enemy aircraft's radar frequency, missile distance, seeker working mode, electromagnetic environment complexity and residual power of interference resources, to form a feature subset.

[0057] After feature selection, each feature is input into the decision tree model according to the weight to determine the priority of the split node and the hierarchical structure of the model, so as to optimize the response speed and strategy accuracy of the decision tree.

[0058] Through feature selection, the system can dynamically adjust strategic decision parameters based on real-time intelligence, laying a data foundation for generating interference strategies that adapt to the battlefield environment.

[0059] 2. Model training and node generation: Use historical air jamming combat data and simulation data as training samples to train and generate a decision tree model.

[0060] (1) When constructing a decision tree, the CART (Classification and Regression Tree) algorithm is used as the basic algorithm of the model. At each splitting node, the CART algorithm recursively calculates the Gini index and selects the eigenvalue with the lowest impurity for splitting.

[0061] The algorithm selects the splitting node through the Gini index, which is used to measure the impurity of the data set, indicating the probability that two samples are randomly selected from the data set and their category labels are inconsistent. For a data set D containing K categories, the calculation formula for its Gini index is:

[0062]

[0063] where p k It represents the proportion of k in the data set D. The value range of the Gini coefficient is between [0,1]. The smaller the value, the higher the purity of the data set, that is, the larger the proportion of samples belonging to the same category. Therefore, the feature with the smallest Gini index is selected when splitting the node.

[0064] (2) When generating a node, a loss function based on the misclassification rate is used to evaluate the splitting effect, thereby deciding whether to continue splitting or to keep the current node as a leaf node.

[0065] Loss function based on misclassification rate:

[0066] C α (T) = C(T) + αT |

[0067] Among them, C(T) is the misclassification rate of the model for training data, |T| is the number of leaf nodes in the decision tree, and α is the adjustment parameter.

[0068] When the loss function value is small, it means that the current splitting effect is good; when the loss function value reaches a certain threshold, the splitting stops and a leaf node is generated.

[0069] (3) Starting from the root node, each layer selects the optimal feature for splitting according to the Gini index, and gradually builds a decision path. For each leaf node on the path, a specific interference strategy output is finally generated.

[0070] When the termination condition is reached, the leaf node is labeled with a specific strategy. For example, strategy labels such as "adopt intensity-drag jamming" or "switch to medium-frequency broadband jamming" are used as the output of the leaf node.

[0071] (4) Optimize the model through pruning algorithms to avoid overfitting. When pruning, the misclassification rate loss function is considered and branches with less influence are deleted to make the model more versatile and efficient.

[0072] As attached Figure 2 As shown, an example of completing the decision tree establishment based on the Gini index is:

[0073] (1) Root node selection: The establishment of the decision tree starts from the root node, and the "seeker working system" is selected as the first split node because this condition has an important influence on the selection of the jamming strategy and has a low Gini coefficient. Therefore, the root node is split based on the "seeker working system" to classify the working status of the enemy seeker.

[0074] (2) Split to generate internal nodes:

[0075] Split according to missile distance interval: After the seeker works, the node is split according to the distance between the missile and the jammer. Here, multiple distance intervals are divided, such as [0, 30), [30, 50), [50, 100), [100, 200), [200, 500), etc. Each interval corresponds to a child node, reflecting the requirements of different distances for jamming methods.

[0076] Further splitting based on the number of missiles: Next, the node is subdivided into two subnodes based on the number of missiles identified. For example, in some branches, the number of missiles can be divided into single missile and multiple missile situations, so that corresponding interference strategies can be adopted according to the degree of threat.

[0077] (3) Refine node decision:

[0078] Refinement of the seeker working state: At certain distance nodes, the seeker is further split according to its specific working state (such as search, tracking, imaging, coherent pulse SAR system). Under different working states, the system has different requirements for jamming strategies, and these splits ensure the accuracy of the jamming strategy.

[0079] Interference type selection: At the leaf node of the refined node, the system further subdivides the interference type, such as selecting "single interference" or "combined interference", and deploys appropriate interference for each target based on the specific situation.

[0080] (4) Interference style and parameter output: In the leaf nodes of the decision tree, according to the interference style classification, multiple interference methods such as broadband noise and aiming frequency interference are selected, including a total of 16 different interference methods. According to actual needs, the system will output specific interference parameter combinations at the leaf nodes, and finally realize the precise deployment of air interference strategies. The CART algorithm of the decision tree is constructed, the Gini index is used to complete the split, and hierarchical decision-making is used to adapt to complex and changing air threats, ensuring the effectiveness and dynamic response capabilities of the interference strategy.

[0081] 3. Optimization of air jamming strategies

[0082] 1. Real-time data input

[0083] The strategy generation module receives real-time data from sensors and external intelligence from the decision tree model. Real-time data includes enemy target parameters (such as the working status of the seeker, missile distance, quantity, etc.) and environmental information (such as electromagnetic noise level). After receiving the real-time data, the system will input it into the root node of the decision tree model to trigger the model's automatic decision-making process.

[0084] 2. Decision path selection

[0085] Path traversal: Based on the eigenvalues ​​of real-time data, the strategy generation module splits along the node path of the decision tree in sequence, filtering from the root node layer by layer until it reaches the leaf node. Each node selects the optimal branch based on the specific eigenvalue to ensure that the strategy generation is highly matched with the current battlefield situation.

[0086] Feature matching: At each node, the module compares the specific values ​​or feature categories of the data. For example, when the "seeker status" is in tracking mode, it enters a specific branch, and when the "distance" is within a specific range, it continues to split along the corresponding path.

[0087] 3. Strategy combination generation

[0088] Single interference strategy selection: When the path reaches the leaf node, if the required interference style is single interference, the strategy generation module directly selects the interference type that best suits the current battlefield needs, such as "low-frequency broadband interference" or "speed drag interference".

[0089] Combined jamming strategy generation: In the case of multiple targets or complex threats, the leaf node outputs a combined jamming strategy. The module combines multiple jamming methods (such as broadband noise + speed drag) according to the current situation, and adjusts the combined jamming intensity and frequency through weighted parameters to adapt to the specific threat level.

[0090] 4. Interference parameter optimization

[0091] Frequency and power control: To achieve the best interference effect, the strategy generation module further optimizes specific parameters after the strategy is generated, adjusts the interference frequency, power and duration, and ensures that the interference effect is accurate and controllable.

[0092] Delay and synchronization control: For multiple jamming methods, the module adjusts the start time and duration of each jamming to achieve a synergistic effect. For example, while performing distance towing jamming on long-range missiles, broadband noise is applied to close-range missiles.

[0093] In the process of interference parameter optimization, the Monte Carlo sorting method generates a large number of strategy samples by simulating multiple possible combinations of interference parameters (such as frequency, power, duration, etc.) to evaluate the effectiveness of each strategy in the current battlefield environment. The specific application process is as follows:

[0094] Generate interference parameter samples: In the "interference parameter optimization" step, the Monte Carlo method is used to sample different interference parameter combinations to generate a series of interference strategy samples.

[0095] Effect evaluation: Conduct effect evaluation on each sampling strategy, simulate the performance of the strategy under different environmental variables, and determine whether it meets the current air interference requirements.

[0096] Select the optimal strategy combination: Through the evaluation results of a large amount of sample data, the Monte Carlo method finally selects the interference parameter combination with the best effect for the current strategy, thereby achieving strategy optimization.

[0097] The result of Monte Carlo sorting is the recommended air jamming strategy for the current scenario, such as jamming timing, method, and parameters. The selection of this strategy path is based on pre-built decision conditions, enabling the system to quickly respond to changes in complex battlefield environments and achieve optimal air jamming effects.

[0098] 5. Generate policy instruction output

[0099] The module finally outputs interference strategy instructions containing detailed parameters such as specific interference mode, frequency, power, duration, etc.

[0100] The generated interference strategy is transmitted to the execution unit for real-time deployment of air interference strategy, thereby achieving the effect of dynamically responding to enemy threats and flexibly coping with complex environments.

[0101] The specific steps are as follows:

[0102] The first step is to input the real-time data of the jammer. For seekers of different systems, the jammer can be selected according to the table below.

[0103] Table 1 Jamming equipment selection matrix

[0104]

[0105] By referring to the jammer selection matrix provided in Table 1, specific jammer matching solutions are provided for different seeker types and target distances. This selection table can help quickly decide which equipment to use against the enemy's seeker, and dynamically adjust the strategy when conditions change (such as target distance, seeker type).

[0106] Step 2: Decision Path Selection

[0107] According to the types of enemy radar seekers, infrared seekers, etc., as well as different missile distances, select appropriate jamming equipment, such as chaff corridor jamming, cross-eye jamming, laser directional jamming, etc.

[0108] In the matrix, equipment is matched by distance range (such as ≤100km, ≤20km, etc.) to ensure that the selected equipment meets the current threat level. For example, the infrared seeker selects "chaff jammer" when the range is 10-15km, and combines laser jammer to improve the effect when the range is ≤10km.

[0109] Step 3: Interference strategy selection

[0110] Obtain the working status of the enemy's seeker. The seeker system includes different types of seekers such as radar, infrared, and laser. Different seekers have different sensitivities to interference and the applicability of interference strategies. Working status and signal form: For example, the seeker may be in search or tracking mode, and the signal form may be pulse, continuous wave, etc. These characteristics directly affect the choice of strategy.

[0111] The detection information obtained by the main radar (actively transmitting signals) and the passive radar (receiving enemy signals) is used to determine the detection capability, distance and threat level of the enemy seeker. For laser seekers, the band (such as near infrared, mid-infrared) and power directly affect the effectiveness of our strategy. For example, a high-power seeker in the near-infrared band may give priority to interference methods with higher energy dissipation.

[0112] Table 2 Infrared seeker jamming strategy

[0113]

[0114] Table 3 Active jamming strategy for active radar seeker

[0115]

[0116] Table 4 Active jamming strategies for passive radar seekers

[0117] Itemization Seeker working status Active jammer style 1 search High repetition rate pulse interference 2 track Combination of Suppression Jamming and Deception Jamming 3 Interference failure Combination of Suppression Jamming and Deception Jamming

[0118] Table 5 Active jamming strategy of composite seeker

[0119] Itemization Seeker working status Active jammer style 1 search High repetition rate pulse interference 2 track Combination of Suppression Jamming and Deception Jamming 3 Interference failure Combination of Suppression Jamming and Deception Jamming

[0120] Lookup table to match jamming strategy: According to the threat level, type of seeker and signal characteristics, the corresponding jamming strategy table is searched to ensure that the strategy is targeted at the current threat. For example, for the high threat level of infrared seeker, a combination jamming strategy of chaff jamming + thermal decoy may be selected. For the medium threat level of radar seeker, aiming frequency jamming or broadband noise jamming can be selected. Strategy combination: In a high threat situation, the table may return a combination jamming strategy, such as laser + chaff combination, to enhance the jamming effect.

[0121] By evaluating the threat level and selecting the corresponding interference strategy through table lookup, it is ensured that the system selects the most effective interference method according to the characteristics of different seekers, can quickly respond to complex threat situations, and improve the accuracy and adaptability of the strategy generation module.

[0122] Step 4: Interference parameter optimization

[0123] When the enemy's radar performance indicators are identified, especially after the enemy's radar pulse sequence is obtained, the jamming strategy can be selected in a targeted manner to jam the enemy's radar. The blue team's jamming strategy implementation decision, that is, the ECM tactics in electronic countermeasures, should be based on the main purpose, whether it is active or passive, the deployment and use method, the type of platform used, the type of jamming the enemy's radar, or a combination of them. After determining the threat level based on the detection results of the active and passive radars, the corresponding jamming strategy is selected by looking up the table, and the jamming type and jamming parameters are issued.

[0124] Table 6 Decisions on interference with infrared seeker

[0125]

[0126]

[0127] This table shows the jamming equipment and methods applicable to infrared seekers of different bands in various situations, such as infrared decoys and laser directional jamming. This table helps the system quickly match the appropriate jamming method after identifying the band of the infrared seeker.

[0128] Table 7 Interference decision for active seeker

[0129]

[0130]

[0131] This table lists applicable jamming styles (such as broadband noise, aiming frequency jamming, false target jamming) according to the working state of the active radar seeker (such as search, tracking, imaging) and signal form (pulse signal, continuous wave, etc.). The system can accurately select the jamming strategy according to the working state of the enemy's seeker.

[0132] Table 8: Decisions on passive seeker jamming

[0133]

[0134] For different states of passive radar seekers (such as search and tracking), this table provides corresponding jamming strategy options, such as high repetition rate pulse jamming and suppression deception combined jamming. This provides a specific jamming strategy reference for passive radar seekers.

[0135] Table 9: Decision making on interference with composite seeker

[0136]

[0137]

[0138] For composite seekers, this table recommends corresponding jamming combinations according to different working states (such as search and tracking) to adapt to different composite threats. The suppression and deception jamming combinations listed in the table help enhance the response effect against multi-function seekers.

[0139] IV. Strategy Evaluation and Optimization

[0140] In the process of implementing the recommended jamming strategy, feedback information is collected in real time, such as jamming effect evaluation indicators (the degree of decline in enemy aircraft radar detection accuracy, enemy aircraft communication interruption, etc.). The strategy is evaluated based on the feedback information. If the strategy effect is found to be unsatisfactory, the new data is fed back to the decision tree model for incremental learning, and the model is adjusted and optimized. By introducing mechanisms such as strategy weights and effect scores, the decision path is dynamically optimized, the accuracy of strategy optimization is further improved, and the adaptability and stability of the jamming strategy in different battlefield situations are ensured.

[0141] At the beginning of decision tree construction, an initial weight can be assigned to each strategy. According to the interference effect evaluation of the selected interference strategy, the weight of the strategy in the decision tree can be adjusted. A common method is to use proportional adjustment, that is, to adjust the weight according to the ratio of the strategy's effect score to the total effect scores of all strategies.

[0142] For example, let the effectiveness score of strategy i be Si, and the sum of the scores of all strategies be The confidence weight of the strategy is

[0143] In the real-time stage of the interference strategy, the weights of the decision tree interference strategy are incrementally learned through the feedback of the interference effect, which further improves the effectiveness of strategy optimization.

[0144] The specific steps are as follows:

[0145] The first step is to input new data into the decision tree model. The decision tree model will update its internal structure based on the new data, such as adjusting the rules for node splitting and increasing or decreasing the depth of the tree.

[0146] The second step is to evaluate the performance of the model through cross-validation during the incremental learning process to ensure that the model still has good generalization ability after learning new data.

[0147] The basic formula for cross validation can be expressed as:

[0148]

[0149] Where CV represents the cross-validation error, N represents the number of data samples, and y i represents the actual observed value of the i-th sample, Represents the model prediction value of the i-th sample.

[0150] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the above embodiments do not limit the present invention in any form, and any technical solution obtained by equivalent replacement or equivalent transformation falls within the protection scope of the present invention.

Claims

1. A method for optimizing air jamming strategy based on decision tree, characterized in that: The following steps are involved: S1. Data collection and preprocessing: Collect enemy aircraft information, seeker parameters, jamming resource information and electromagnetic environment information, and perform data cleaning and normalization to obtain feature inputs for the decision tree model; S2. Decision tree model construction: Based on the CART algorithm, the Gini index is used to split the nodes to generate a decision tree model containing key features; S3, jamming strategy generation: Generate adaptive jamming strategies through decision tree models based on real-time battlefield data input; S4. Interference parameter binding and optimization: The generated interference strategy is subjected to parameter adjustment of frequency, power, delay and duration to ensure the optimal effect of the interference strategy in the battlefield environment.

2. The air interference strategy optimization method based on decision tree according to claim 1 is characterized in that: The collection of enemy aircraft information, seeker parameters, jamming equipment characteristics and electromagnetic environment information specifically includes: Enemy aircraft information used to analyze the enemy aircraft's action intentions and its response to jamming strategies: enemy aircraft basic flight parameters, including model, speed, altitude, and heading; radar performance information including: frequency, power, scanning mode, etc. Seeker parameters used to help the decision tree model determine the most appropriate jamming strategy, including information related to the enemy seeker's operating status, characteristics, and current environment; Electromagnetic environment information for selecting appropriate interference methods based on complex electromagnetic environments: including the electromagnetic noise level in the current war zone, the distribution of other radio signals, etc.; Interference resource information to evaluate the availability and adaptability of current interference resources: including the model, interference power, and frequency range of our jammers.

3. The air interference strategy optimization method based on decision tree according to claim 1 is characterized in that: The decision tree model building step comprises: S21. Use information gain to calculate the importance of each feature, so as to select the key features that have the greatest impact on the generation of the decision tree, including the enemy aircraft's radar frequency, missile distance, seeker working mode, electromagnetic environment complexity, and interference resource remaining power, to form a feature subset; S22. Use the CART algorithm to calculate the Gini index of each feature in the feature subset, select the feature with the highest information gain as the split node, and generate a node splitting path; construct a decision path based on different seeker working systems and signal forms until reaching the leaf node to generate a specific interference strategy.

4. The method for optimizing air interference strategy based on decision tree according to claim 3 is characterized in that: The specific implementation of step S22 is as follows: (1) When constructing a decision tree, the CART (Classification and Regression Tree) algorithm is used as the basic algorithm of the model. At each splitting node, the CART algorithm recursively calculates the Gini index and selects the eigenvalue with the lowest impurity for splitting; (2) When generating a node, a loss function based on the misclassification rate is used to evaluate the splitting effect, thereby deciding whether to continue splitting or to treat the current node as a leaf node; (3) Starting from the root node, each layer selects the optimal features for splitting according to the Gini index, and gradually constructs a decision path; for each leaf node on the path, a specific interference strategy output is finally generated.

5. The method for optimizing air interference strategy based on decision tree according to claim 4 is characterized in that: The interference strategy generation step in step S3 is specifically as follows: S31, receiving real-time data from sensors and external intelligence, inputting it into the root node of the decision tree model, and triggering the automatic decision-making process of the model; S32, according to the characteristic value of the real-time data, the strategy generation module splits the node path of the decision tree in sequence, and screens downward from the root node layer by layer until reaching the leaf node; S33. When the path reaches the leaf node, if the required interference style is single interference, the interference type that best suits the current battlefield needs is directly selected; in the case of multiple targets or complex threats, the leaf node outputs a combined interference strategy.

6. The method for optimizing air interference strategy based on decision tree according to claim 5, characterized in that: Step S3, the interference strategy generation step, also includes an interference parameter optimization step, specifically: (1) Adjust the jamming frequency and power to match the frequency range of the enemy aircraft's detection signal; (2) Control interference delay and duration.

7. The method for optimizing air interference strategy based on decision tree according to claim 6, characterized in that: In the interference parameter optimization step, the Monte Carlo sorting method generates a large number of strategy samples by simulating multiple possible interference parameter combinations to evaluate the effectiveness of each strategy in the current battlefield environment. The specific application process is as follows: Generate interference parameter samples: In the "interference parameter optimization" step, use the Monte Carlo method to sample different interference parameter combinations to generate a series of interference strategy samples; Effect evaluation: Evaluate the effect of each sampling strategy and simulate the performance of the strategy under different environmental variables; Select the optimal strategy combination: Through the evaluation results of a large amount of sample data, the Monte Carlo method finally selects the interference parameter combination with the best effect for the current strategy.

8. The method for optimizing air interference strategy based on decision tree according to claim 5 is characterized in that: After the interference strategy is generated in step S33, the interference strategy is output: an interference strategy instruction including detailed parameters such as specific interference mode, frequency, power, duration, etc. is output.

Images