A decision tree-based aerial combat dogfight roll maneuver auxiliary decision-making method

Abstract

The application discloses a kind of based on decision tree's air combat dog fight roll maneuver auxiliary decision method, belong to aviation technology field, including: S1 obtains the battlefield situation data related to enemy and our two sides warplane and the roll angle data of our warplane;S2 battlefield situation data and roll angle data are preprocessed;S3 the battlefield situation data after being preprocessed, roll angle data are respectively used as input and output to train decision tree algorithm, obtain air combat dog fight roll maneuver auxiliary decision model;S4 in the process of single machine close air combat fight, the battlefield situation data related to enemy and our two sides warplane collected is preprocessed after using the same mode and input to air combat dog fight roll maneuver auxiliary decision model, obtains the roll angle category for assisting our warplane to carry out roll maneuver.The application can provide maneuver roll angle suggestion to different battlefield situation, with the characteristics of easy to realize, strong explainability and high consistency with human intuitive thinking.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of aviation technology, in particular to a dogfight roll maneuvering auxiliary decision-making method based on a decision tree. BACKGROUND

[0002] Air combat usually includes medium and long distance, near distance and other task scenarios. In medium and long distance air combat, both sides of enemy and me use radar guided missiles as weapons, and perform tasks such as situation awareness and prediction, missile positioning and launching, and enemy missile avoidance according to airborne sensors / early warning aircraft and other equipment. In near distance air combat, the weapon commonly used by pilots is infrared guided missile and aircraft gun. However, due to the influence of missile envelope, aircraft gun range and shooting power, etc., the tail-chasing attack is the most effective attack method in near distance air combat (i.e. dogfight). However, dogfighting has the characteristics of high dynamic, strong real-time and large overload, and the pilot's operation requirement is extremely high. At present, various countries have carried out a lot of training for air combat dogfight. The well-known foreign military dogfight training institutions include the United States navy fighter weapons school (nicknamed: TOP GUN), and the Chinese navy aviation also uses DCS World and other air combat simulation software with good simulation degree to carry out a lot of training work.

[0003] Due to the limitations of human physiological limits, psychological load and other factors, in air combat dogfight, pilots will inevitably "make mistakes", that is, to perform air combat with non-optimal performance maneuvering decision. At present, with the rapid development of unmanned aerial vehicle technology, the sustained optimization decision performance advantage of unmanned aerial vehicle is more prominent, which is not affected by physiological, psychological and other factors. A lot of research on unmanned aerial vehicle technology for autonomous air combat has been carried out, and certain achievements have been made. However, the current autonomous unmanned air combat technology has not considered the influence of weapon use principles, ethics and other factors, and there are few mass-produced unmanned aerial vehicles with autonomous air combat capability. Considering the weapon use principles, ethics and other issues related to air combat dogfight, pilots cannot be completely eliminated in future air combat dogfight.

[0004] Considering the excellent performance of data-driven methods for sustained optimization decision, some scholars have proposed to develop a dogfight auxiliary system, so as to provide real-time air combat dogfight suggestions for pilots by using data-driven methods. However, on the one hand, the performance of these methods still needs to be improved; on the other hand, in the face of high dynamic and strong real-time environment of dogfight, these methods also need to have good and intuitive understandable performance, in order to improve the human-machine trust degree of pilots and related algorithms, ensure the algorithm performance advantage of related methods, realize the "1+1>2" effect in the scene of human-machine coexistence and cooperation.

[0005] The decision tree is a machine learning algorithm for solving classification problems. The decision tree algorithm adopts a tree structure and layer-by-layer reasoning to achieve the final classification. The decision tree algorithm usually includes three types of elements: root node, internal node and leaf node. Among them, the root node contains the sample set, the internal node corresponds to feature extraction and attribute test, and the leaf node represents the result of classification. In the classification task, the root node inputs data, the internal node judges based on the attribute value to determine the next internal node of layer-by-layer, and finally reaches the leaf node to output the classification result. The essence of the decision tree algorithm is a supervised learning algorithm based on if-then logical reasoning structure, and the attribute value and attribute test method of the related node are obtained through data-driven training. Because of the easy implementation, strong interpretability and high consistency with human intuitive thinking of the decision tree, it is necessary to carry out air combat dog fight maneuver classification research based on the decision tree algorithm.

[0006] Considering that the pilot often adopts BFM maneuver in the execution of close-range air combat task, that is, according to the situation of the enemy and the friendly forces on the battlefield to select the appropriate aircraft roll angle, and apply the normal overload to point the nose to the enemy as much as possible, it can be seen that one of the cores of the BFM maneuver is to determine the appropriate roll angle of the aircraft. Therefore, it is actually urgent to develop an air combat dog fight roll maneuver decision method based on the decision tree algorithm. SUMMARY

[0007] In view of the problems of the existing technology that the aircraft roll angle auxiliary decision method in air combat is not easy to implement, poor interpretability and low consistency with human intuitive thinking, the purpose of the present application is to provide an air combat dog fight roll maneuver auxiliary decision method based on a decision tree.

[0008] To achieve the above purpose, the technical scheme of the present application is:

[0009] In the first aspect, the present application provides an air combat dog fight roll maneuver auxiliary decision method based on a decision tree, comprising the following steps:

[0010] S1, obtaining a plurality of battlefield situation data related to the enemy and friendly aircraft, and obtaining the roll angle data of the friendly aircraft;

[0011] S2, preprocessing the battlefield situation data and the roll angle data;

[0012] S3, constructing a decision tree algorithm, training the decision tree algorithm by taking the preprocessed battlefield situation data and roll angle data as input and output respectively, and obtaining an air combat dog fight roll maneuver auxiliary decision model;

[0013] S4, in the process of single aircraft close combat, the collected battlefield situation data related to both sides of the warplane is preprocessed in the same way and input into the air combat dog fight roll maneuver auxiliary decision model, so as to obtain the roll angle category for assisting the warplane to perform roll maneuver.

[0014] Preferably, in step S2, the step of preprocessing the roll angle data comprises:

[0015] The roll angle data is defined in the range of-180°-180°.

[0016] The roll angle data is classified by clustering algorithm, and the optimal category number, inter-class distance and class center point of the roll angle category are determined according to the classification result.

[0017] The roll angle category corresponding to each roll angle data is recorded, so as to be used as the output of the decision tree algorithm in step S3.

[0018] Preferably, the step of determining the optimal category number, inter-class distance and class center point of the roll angle category according to the classification result comprises:

[0019] A plurality of different category numbers are searched, and for each category number, the class center point, intra-class distance and inter-class distance are calculated based on the kmeans clustering analysis algorithm respectively.

[0020] The optimal category number, inter-class distance and class center point of the roll angle category are determined according to the proportion value of intra-class distance and inter-class distance, and the change trend of the proportion value of intra-class distance and inter-class distance.

[0021] Preferably, in step S2, the step of preprocessing the battlefield situation data comprises normalizing the battlefield situation data.

[0022] Preferably, in step S1, the battlefield situation data includes: relative position yetoo of the enemy aircraft in the y-axis direction of the aircraft body coordinate, relative position zetoo of the enemy aircraft in the z-axis direction of the aircraft body coordinate, relative position xetoo of the enemy aircraft in the x-axis direction of the aircraft body coordinate, straight-line distance Distance of the enemy aircraft to the aircraft, closure speed magnitude Closure_v of the enemy aircraft to the aircraft, angle ATA of the line-of-sight direction of the enemy aircraft to the aircraft to the x-axis of the aircraft body coordinate system, rate of change Dot(ATA) of the angle ATA, angle AA of the line-of-sight direction of the enemy aircraft to the aircraft to the x-axis of the enemy aircraft body coordinate system, rate of change Dot(AA) of the angle AA, angle HCA of the x-axis direction of the enemy aircraft body coordinate system to the x-axis direction of the aircraft body coordinate system, rate of change Dot(HCA) of the angle HCA, angle of attack AOA of the aircraft, normal acceleration Ny of the aircraft, altitude Alt_ground_level of the aircraft, and speed magnitude vtas of the aircraft.

[0023] Preferably, in step S1, the battlefield situation data related to the enemy and friendly aircraft and the roll angle data of the friendly aircraft are acquired in the process of single-aircraft close-range air combat training of the enemy and friendly aircraft based on BFM maneuver conditions.

[0024] Preferably, in step S3, the decision tree algorithm is constructed based on the XGBOOST algorithm library.

[0025] In a second aspect, the present application also provides a decision tree-based air combat dogfight roll maneuver auxiliary decision system, comprising:

[0026] An acquisition module is configured to collect battlefield situation data related to enemy and friendly aircraft and roll angle data of the friendly aircraft.

[0027] A preprocessing module is configured to preprocess the battlefield situation data and the roll angle data.

[0028] A training module is configured to train a decision tree algorithm and obtain an air combat dogfight roll maneuver auxiliary decision model.

[0029] An execution module is configured to input the collected battlefield situation data related to enemy and friendly aircraft into the air combat dogfight roll maneuver auxiliary decision model to obtain a roll angle category for assisting the friendly aircraft in roll maneuver.

[0030] In a third aspect, the present application also provides an electronic device, comprising a memory storing executable program codes and a processor coupled with the memory; wherein the processor invokes the executable program codes stored in the memory to execute the method as described above.

[0031] In a fourth aspect, the present application also provides a computer readable storage medium storing a computer program, which, when executed by a processor, performs the method as described above.

[0032] With the above technical solution, the present application provides a dogfight roll maneuver auxiliary decision-making method based on a decision tree, which is oriented to the dogfight problem and provides a dogfight maneuver roll angle suggestion in real time by using a decision tree algorithm for different battlefield situations, and has the characteristics of easy implementation, strong interpretability and high consistency with human intuitive thinking. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 A method flowchart of the dogfight roll maneuver auxiliary decision-making method based on a decision tree in the present application;

[0034] Figure 2 An example schematic diagram of various maneuver actions used in close-range air combat flight;

[0035] Figure 3 A time length schematic diagram of the present application in which our aircraft respectively fight against three types of enemy aircraft in air combat;

[0036] Figure 4 A proportional value schematic diagram of the intra-class distance and the inter-class distance obtained after the clustering algorithm solves the roll angle data in the present application;

[0037] Figure 5 A change trend schematic diagram of the proportional value of the intra-class distance and the inter-class distance obtained after the clustering algorithm solves the roll angle data in the present application;

[0038] Figure 6 A roll angle classification result schematic diagram under the condition that the category number is 6 in the present application;

[0039] Figure 7 A structure schematic diagram of the decision tree network proposed in the present application;

[0040] Figure 8 A structure schematic diagram of the dogfight roll maneuver decision-making system based on a decision tree in the present application;

[0041] Figure 9 A structure schematic diagram of an electronic device in the present application. DETAILED DESCRIPTION

[0042] The specific embodiments of the present application will be further described below with reference to the drawings. It should be noted that the description of these embodiments is intended to help understand the present application, but does not constitute a limitation of the present application. In addition, the technical features involved in the various embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0043] Embodiment one

[0044] A dogfight roll maneuver decision-making method based on a decision tree, as shown in Figure 1 , the method comprises steps S1-S4.

[0045] S1, obtain a plurality of battlefield situation data related to the enemy and our warplanes, and obtain the roll angle data of our warplane.

[0046] This embodiment is based on the DCS World air combat simulation software for data collection and analysis, which has high authenticity in simulating the flight dynamics, avionics system, fire control system, etc. of the fighter. It has been used for flight training of some military branches and has achieved good results.

[0047] This embodiment is aimed at one-on-one air combat within the line of sight, and the enemy and our warplanes are both operated by human pilots with good air combat skills. The roll angle data collection of our warplane is mainly aimed at more experienced pilots. In close-range air combat, the flight characteristics, maneuvering performance, cockpit field of view, etc. of the enemy and our warplanes have a great influence on flight maneuver decision-making and flight control, so this embodiment's data collection investigates four different types of warplanes, as shown in Table 1. Based on previous research results, domestic and foreign research, and evaluations of pilots from multiple military branches, the simulation software used in this application can achieve good simulation of the four types of aircraft in terms of flight and fire control.

[0048] During the data collection process of this embodiment, the pilots fly air combat using BFM maneuvers. Traditional BFM maneuvers include high-g, low-g, horizontal / vertical scissors flight, etc. Some examples of BFM maneuvers collected in this embodiment are shown in Figure 2 , where (a) is high-g, (b) is break S maneuver, (c) is pure pursuit, (d) is horizontal scissors flight, (e) is vertical scissors flight, (f) is somersault maneuver, and (g) is barrel roll maneuver. During data collection, a more experienced pilot flies the F warplane (i.e. our warplane), and another pilot flies the other three types of warplanes (i.e. enemy warplanes), and performs three air combat maneuvers respectively to facilitate data collection. As shown in Figure 3As shown, it is the total duration of the data collected in this embodiment (about 360 flight minutes in total), and the time of three air combat fights is kept balanced. Among them, the initial distance of the enemy and our warplanes is 10 km, the initial speed is 0.75 Mach, and the initial (radio) height is selected as 4500 meters (M1), 7500 meters (M2), and 600 meters (S) respectively.

[0049] Table 1 - Basic parameters of four types of warplanes

[0050]

[0051] In this embodiment, there are 15 battlefield situation data in total, including: the relative position yetoo of the enemy plane in the y-axis direction of the body coordinate of the host plane, the relative position zetoo of the enemy plane in the z-axis direction of the body coordinate of the host plane, the relative position xetoo of the enemy plane in the x-axis direction of the body coordinate of the host plane, the straight-line distance Distance from the enemy plane to the host plane, the closure speed Closure_v of the enemy plane to the host plane, the angle ATA of the line-of-sight direction of the enemy plane to the host plane and the x-axis of the body coordinate system of the host plane, the rate of change Dot(ATA) of the angle ATA of the line-of-sight direction of the enemy plane to the host plane and the x-axis of the body coordinate system of the host plane, the angle AA of the line-of-sight direction of the enemy plane to the host plane and the x-axis of the body coordinate system of the enemy plane, the rate of change Dot(AA) of the angle AA of the line-of-sight direction of the enemy plane to the host plane and the x-axis of the body coordinate system of the enemy plane, the angle HCA of the x-axis direction of the body coordinate system of the enemy plane and the x-axis direction of the body coordinate system of the host plane, the rate of change Dot(HCA) of the angle HCA of the x-axis direction of the body coordinate system of the enemy plane and the x-axis direction of the body coordinate system of the host plane, the angle of attack AOA of the host plane, the normal acceleration Ny of the host plane, the altitude Alt_ground_level of the host plane, and the speed vtas of the host plane.

[0052] The data directly collected in Table 2 is shown in Table 2, so the data in Table 2 needs to be converted to obtain the battlefield situation data described above.

[0053] Table 2 - Data types of enemy and our warplanes

[0054]

[0055]

[0056] It can be understood that, in order to avoid overfitting in the decision tree supervised learning process in the subsequent steps, the data set is divided in this embodiment, wherein 2 / 3 of the data (about 240 flight minutes) is used as training data, and 1 / 3 of the data (about 120 flight minutes) is used as test data. Since the flight time of M1, M2 and S fighters in the original data set is relatively balanced, the proportion of the three types of fighters in the final training data and test data is also roughly balanced, that is, the training data of each type of aircraft is about 80 flight minutes, and the test data is about 40 flight minutes.

[0057] S2, preprocessing the battlefield situation data and the roll angle data.

[0058] In this embodiment, the step of preprocessing the battlefield situation data includes normalizing the battlefield situation data.

[0059] In this embodiment, the step of preprocessing the roll angle data includes:

[0060] S21, defining the roll angle data in the range of -180°-180°.

[0061] S22, classifying the roll angle data by a clustering algorithm, and determining the optimal number of roll angle categories, inter-class distance and class center point according to the classification result.

[0062] That is, the roll angle data is discretized in this embodiment, and the classified roll angle data after discretization is used as the output of the decision tree algorithm.

[0063] The step of determining the optimal number of roll angle categories, inter-class distance and class center point according to the classification result includes:

[0064] Iterative search for a plurality of different category numbers, and for each category number, the class center point, intra-class distance and inter-class distance are calculated based on the kmeans clustering analysis algorithm respectively; the optimal number of roll angle categories, inter-class distance and class center point are determined comprehensively according to the ratio value of intra-class distance and inter-class distance and the change trend of the ratio value of intra-class distance and inter-class distance.

[0065] For example, in the case of 2-16 categories, the corresponding class center points, intra-class distances and inter-class distance data are calculated respectively, and the ratio value of intra-class distance and inter-class distance and the change trend of the ratio value of intra-class distance and inter-class distance are as shown in Figure 4 、 Figure 5 It can be seen that the smaller the ratio value of intra-class distance and inter-class distance, the more the data in each class is gathered and the more it is separated from other classes. Therefore, after comprehensive consideration, the optimal number of categories is selected as 6, as shown in Figure 6 which shows a roll angle category division diagram when the number of categories is 6.Figure 6 In the figure, the top represents the position of the roll angle of 0°, the right deviation is positive and the left deviation is negative, the value of the roll angle is distributed on the unit circle with the origin as the center, and the six segments on the unit circle represent the classification results of clustering the roll angle data. The "+" symbol in each segment represents the class center point of the roll angle category, wherein the horizontal coordinate represents the sine value of the roll angle and the vertical coordinate represents the cosine value of the roll angle.

[0066] S23, record the roll angle category corresponding to each roll angle data, so as to be used as the output of the decision tree algorithm in step S3.

[0067] In the embodiment, the roll angle value is divided into six category numbers (angle value intervals), because in actual air combat, our aircraft does not need to accurately control the specific value of the roll angle, and it is already relatively accurate as long as the roll angle is roughly in a certain interval, so for the output result of the model, it is only required to output a range interval (roll angle category).

[0068] S3, construct a decision tree algorithm, train the decision tree algorithm by taking the preprocessed battlefield situation data and roll angle data as input and output respectively, and obtain an air combat roll maneuvering auxiliary decision-making model.

[0069] The embodiment constructs a decision tree algorithm based on the XGBOOST algorithm library, as shown in FIG. 2, which shows a structural schematic diagram of the decision tree algorithm constructed in the embodiment. Figure 7

[0070] In the embodiment, the key parameters of the decision tree algorithm mainly include three kinds of learning rate, number of decision trees and maximum depth of the tree. Through training and testing of the input and output data, the accuracy of the prediction of the output data by the decision tree is compared. It can be concluded that when the learning rate is 0.1, the number of decision trees is 100, and the maximum depth of the tree is 10, the decision tree model achieves the best accuracy. The important parameter settings of the decision tree are shown in Table 3, and the other parameter values of the decision tree adopt the default values.

[0071] Table 3-Important parameter table of decision tree algorithm network

[0072] Parameter Name Parameter Meaning Setting Value learning_rate Learning rate of the model produced at each iteration 0.1 n_estimators Number of decision trees 100 max_depth Maximum depth of the tree 10 min_child_weight Minimum weight of leaf nodes 1 gamma Minimum loss function decrease required for node splitting 0.5 subsample Proportion of random sampling for each tree 0.9 objective Given loss function 'multi:softmax' num_class Number of classes, used with'multi:softmax' 8 random_state Random seed number 27

[0073] S4, in the process of single machine close-range air combat, the battlefield situation data related to the enemy and our aircraft collected is preprocessed in the same way as step S1 and then input into the air combat roll maneuvering auxiliary decision-making model, so as to obtain the roll angle category used for assisting our aircraft to perform roll maneuvering.

[0074] ​In the step of testing the model, the embodiment verifies the accuracy of the air combat dogfight roll maneuver auxiliary decision-making model through the confusion matrix. Since the roll angle categories of our aircraft are divided into 6 categories in the embodiment, that is, the output results of the model are divided into 6 categories, therefore, the confusion matrix is a 6x6 matrix. In the confusion matrix, each row represents the actual classification result, and each column represents the predicted classification result output by the model, and the diagonal line represents the percentage of the predicted classification result consistent with the actual classification result. Finally, after using the optimized decision tree algorithm network, the test results are shown in Table 4.

[0075] Table 4 - Network prediction results (confusion matrix)

[0076]

[0077] In practical application, for the roll angle classification problem in single aircraft close-range air combat, the air combat dogfight roll maneuver auxiliary decision-making model proposed in the embodiment can effectively meet the maneuver auxiliary decision-making task as long as the output roll angle category is in the same or adjacent position as the actual roll angle data. Therefore, in the test step, the embodiment defines three levels of accuracy for the predicted classification results output by the model, and the definition method is shown in Table 5, wherein the second level of accuracy can also be called right bias accuracy, and the third level of accuracy includes right bias and left bias based on the operation habit of pilots usually rolling to the right.

[0078] Table 5 - Three-level accuracy definition

[0079]

[0080] Based on the definition of three levels of accuracy of the predicted classification results, in the above test classification results, the first level of accuracy is close to 3 / 4 (74.06%), and the third level of accuracy is more than 96%, as shown in Table 6, it can be seen that the air combat maneuver auxiliary decision-making method proposed in the embodiment has good accuracy performance, and is expected to effectively provide good roll auxiliary decision-making for pilot air combat flight maneuver in complex close-range air combat battlefield situation.

[0081] Table 6 - Prediction accuracy of decision tree algorithm network

[0082] Accuracy Rating Accuracy (%) Primary 74.06 Secondary 85.08 Tertiary 96.19

[0083] Embodiment Two

[0084] An air combat dogfight roll maneuver auxiliary decision-making system based on a decision tree, as shown in Figure 8 , comprising:

[0085] An acquisition module for collecting battlefield situation data related to enemy and friendly aircraft and roll angle data of the friendly aircraft;

[0086] a pretreatment module configured to pretreat the battlefield situation data and the roll angle data;

[0087] a training module configured to train the decision tree algorithm and obtain an air combat dogfight roll maneuver auxiliary decision model;

[0088] and an execution module configured to input the collected battlefield situation data related to the enemy and friendly aircraft into the air combat dogfight roll maneuver auxiliary decision model to obtain a roll angle category for assisting the friendly aircraft to perform roll maneuver.

[0089] Embodiment Three

[0090] An electronic device, as shown in the accompanying drawings, comprises a memory storing executable program codes and a processor coupled with the memory; wherein the processor invokes the executable program codes stored in the memory to execute the method steps disclosed in the above embodiments. Figure 9

[0091] Embodiment Four

[0092] A computer storage medium, wherein the computer storage medium stores a computer program, and the computer program is run by a processor to execute the method steps disclosed in the above embodiments.

[0093] The present application is described with reference to the flowcharts and / or block diagrams according to the methods, devices (systems), and computer program products of embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to produce a machine, so that the instructions executed by the computer or other programmable data processing devices generate a device that implements the functions specified in the flowcharts and / or block diagrams. Figure One The functions specified in one flow or multiple flows and / or blocks Figure One The functions specified in one flow or multiple flows and / or blocks

[0094] These computer program instructions can also be stored in a computer-readable memory that can direct the computer or other programmable data processing devices to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including instruction devices that implement the functions specified in the flowcharts and / or block diagrams. Figure One The functions specified in one flow or multiple flows and / or blocks Figure One The functions specified in one flow or multiple flows and / or blocks

[0095] ​These computer program instructions can also be loaded into a computer or other programmable data processing devices, so that a series of operational steps are performed on the computer or other programmable data processing devices to generate a computer implemented process, so that the instructions executed on the computer or other programmable data processing devices provide a process for implementing the functions specified in the flowchart Figure One one flow or multiple flows and / or blocks Figure One one block or multiple blocks.

[0096] The embodiments of the present application described above are merely intended to illustrate the present application, but the present application is not limited to the described embodiments. Various changes, modifications, replacements and variations of the embodiments can be made by those skilled in the art without departing from the principles and spirits of the present application, and still fall within the protection scope of the present application.

Claims

1. A decision tree-based aerial combat dogfight roll maneuver assistance decision method, characterized in that: The method comprises the following steps: S1, acquiring a plurality of battlefield situation data related to both sides of the warplane and acquiring the roll angle data of the warplane of the own side; S2, preprocessing the battlefield situation data and the roll angle data; S3, constructing a decision tree algorithm, training the decision tree algorithm by taking the preprocessed battlefield situation data and roll angle data as input and output respectively, and obtaining an air combat roll maneuver auxiliary decision model; S4, in the process of single machine close-range air combat, the collected battlefield situation data related to both sides of the warplane is preprocessed in the same way and then input into the air combat roll maneuver auxiliary decision model, so as to obtain a roll angle category for assisting the warplane of the own side to perform roll maneuver; In step S2, the preprocessing of the roll angle data comprises: defining the roll angle data in the range of -180°-180°; classifying the roll angle data by a clustering algorithm, determining the optimal category number, inter-class distance and class center point of the roll angle category according to the classification result; recording the roll angle category corresponding to each roll angle data, so as to be used as the output of the decision tree algorithm in step S3; the step of determining the optimal category number, inter-class distance and class center point of the roll angle category according to the classification result comprises: iteratively searching for a plurality of different category numbers, and for each category number, respectively calculating the class center point, intra-class distance and inter-class distance based on the kmeans clustering analysis algorithm; comprehensively determining the optimal category number, inter-class distance and class center point of the roll angle category according to the proportion value of the intra-class distance and the inter-class distance and the change trend of the proportion value of the intra-class distance and the inter-class distance.

2. The method of claim 1, wherein: In step S2, the preprocessing of the battlefield situation data comprises normalizing the battlefield situation data.

3. The method of claim 1, wherein: In step S1, the battlefield situation data comprises the relative position yetoo of the enemy plane in the y-axis direction of the machine body coordinate, the relative position zetoo of the enemy plane in the z-axis direction of the machine body coordinate, the relative position xetoo of the enemy plane in the x-axis direction of the machine body coordinate, the straight-line distance Distance from the enemy plane to the machine, the closure speed Closure_v of the enemy plane to the machine, the angle ATA between the line-of-sight direction of the enemy plane to the machine and the x-axis of the machine body coordinate system, the change rate Dot(ATA) of the angle ATA, the angle AA between the line-of-sight direction of the enemy plane to the machine and the x-axis of the enemy plane body coordinate system, the change rate Dot(AA) of the angle AA, the angle HCA between the x-axis direction of the enemy plane body coordinate system and the x-axis direction of the machine body coordinate system, the change rate Dot(HCA) of the angle HCA, the angle of attack AOA of the machine, the normal overload Ny of the machine, the altitude Alt_ground_level of the machine, and the speed vtas of the machine.

4. The method of claim 1, wherein: In step S1, in the process of single close combat training of both sides of warplanes based on BFM maneuvering conditions, several battlefield situation data related to both sides of warplanes and roll angle data of our warplanes are acquired.

5. The method of claim 1, wherein: In step S3, the decision tree algorithm is constructed based on the XGBOOST algorithm library.

6. A decision tree based aerial combat dogfight roll maneuver aiding decision system, characterized in that: The method comprises the steps of: an acquisition module, configured to collect battlefield situation data related to both sides of warplanes and roll angle data of our warplanes; a preprocessing module, configured to preprocess the battlefield situation data and the roll angle data; a training module, configured to train a decision tree algorithm and obtain an air combat roll maneuvering auxiliary decision model; and an execution module, configured to input the collected battlefield situation data related to both sides of warplanes into the air combat roll maneuvering auxiliary decision model to obtain a roll angle category for assisting our warplanes to perform roll maneuvering; wherein the preprocessing module preprocesses the roll angle data by: defining the roll angle data in the range of -180°-180°; classifying the roll angle data by a clustering algorithm, and determining the optimal category number, inter-class distance and class center point of the roll angle category according to the classification result; recording the roll angle category corresponding to each roll angle data, so as to be used as the output of the decision tree algorithm in step S3; the step of determining the optimal category number, inter-class distance and class center point of the roll angle category according to the classification result comprises: iteratively searching for multiple different category numbers, and for each category number, respectively calculating the class center point, intra-class distance and inter-class distance based on the kmeans clustering analysis algorithm; comprehensively determining the optimal category number, inter-class distance and class center point of the roll angle category according to the proportion value of the intra-class distance and the inter-class distance and the change trend of the proportion value.

7. An electronic device, comprising: The computer program is executed by the processor to perform the method of any one of claims 1-5.

8. A computer readable storage medium storing a computer program, characterized in that: The computer program is executed by the processor to perform the method of any one of claims 1-5.

Images