Unmanned aerial vehicle intelligent autonomous decision-making method and device, electronic equipment, readable storage medium and program product

Abstract

The embodiment of the invention discloses an intelligent autonomous decision-making method and device for an unmanned aerial vehicle, electronic equipment, a readable storage medium and a program product. The method comprises the following steps: training by using a tactical behavior sample set to obtain a tactical behavior prediction model; acquiring action parameter information of the enemy plane in a preset time sequence, and determining a next maneuver sequence feature of the enemy plane according to the action parameter information and the tactical behavior prediction model; constructing a situation analysis model based on space occupation of our aircrafts and enemy aircrafts; determining an occupation comprehensive advantage value and an occupation comprehensive threat value of the our aircraft relative to the enemy aircraft based on the situation analysis model, and determining a radar comprehensive advantage value and a radar comprehensive threat value; and according to the occupancy comprehensive advantage value, the occupancy comprehensive threat value, the radar comprehensive advantage value and the radar comprehensive threat value, making an optimal confrontation decision for the next maneuver sequence feature of the enemy plane. According to the technical scheme, the efficiency and accuracy of autonomous intelligent decision making can be effectively improved on the basis of improving the ability of autonomous intelligent decision making.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent decision-making for unmanned aerial vehicles, and particularly to an intelligent autonomous decision-making method, device, electronic device, readable storage medium, and program product for unmanned aerial vehicles. Background Art

[0002] From the current technical level, the demand for intelligent unmanned aerial vehicles with "reconnaissance and strike" capabilities is rapidly increasing. Usually, the combat environment is judged through the radar monitoring of the unmanned aerial vehicle and the thinking of the user. Therefore, the commonly used method is that in the specific combat process, the manned aircraft and the unmanned aerial vehicle cooperate to implement combat operations and transmit them out. Precious time will be delayed. The modern battlefield is unpredictable and the information changes rapidly. The entire decision-making output is not ideal. The unmanned aerial vehicle system will have higher thinking ability and autonomous decision-making ability in the future complex battlefield environment, which will inevitably become the future trend. How to quickly make an intelligent response strategy is a problem that needs to be solved. Therefore, there is an urgent need for an intelligent autonomous decision-making method for unmanned aerial vehicles to solve the above technical problems. Summary of the Invention

[0003] In view of this, the present invention provides an intelligent autonomous decision-making method, device, electronic device, readable storage medium, and program product for unmanned aerial vehicles, which can improve the ability of autonomous intelligent decision-making in a complex battlefield environment and effectively improve the efficiency and accuracy of autonomous intelligent decision-making.

[0004] According to an aspect of the present invention, an embodiment of the present invention provides an intelligent autonomous decision-making method for unmanned aerial vehicles, and the method includes:

[0005] Obtain a pre-constructed tactical behavior sample set, and use the tactical behavior sample set to train a behavior prediction model to obtain a trained tactical behavior prediction model;

[0006] Obtain the action parameter information of the enemy aircraft in a preset time series, and determine the next maneuver sequence feature of the enemy aircraft according to the action parameter information and the tactical behavior prediction model;

[0007] Construct a situation analysis model based on the spatial occupancy of our aircraft and the enemy aircraft; wherein, the situation analysis model includes: an angle situation analysis model, a distance situation analysis model, a speed situation analysis model, and a height situation analysis model;

[0008] Determine the occupancy comprehensive advantage value and occupancy comprehensive threat value of our aircraft relative to the enemy aircraft based on the situation analysis model;

[0009] Determine the radar comprehensive advantage value and radar comprehensive threat value of the airborne radar of our aircraft relative to the airborne radar of the enemy aircraft;

[0010] Make an optimal countermeasure decision on the next maneuver sequence feature of the enemy aircraft based on the occupancy comprehensive advantage value, the occupancy comprehensive threat value, the radar comprehensive advantage value, and the radar comprehensive threat value.

[0011] According to another aspect of the present invention, an unmanned aerial vehicle intelligent autonomous decision-making device is also provided in an embodiment of the present invention. The device includes:

[0012] A data acquisition module, configured to acquire a pre-constructed tactical behavior sample set, and use the tactical behavior sample set to train a behavior prediction model to obtain a trained tactical behavior prediction model;

[0013] An enemy aircraft maneuver sequence prediction module, configured to acquire action parameter information of the enemy aircraft in a preset time series, and determine the next maneuver sequence feature of the enemy aircraft according to the action parameter information and the tactical behavior prediction model;

[0014] A situation analysis establishment module, configured to construct a situation analysis model based on the spatial occupancy of our aircraft and the enemy aircraft; wherein, the situation analysis model includes: an angle situation analysis model, a distance situation analysis model, a speed situation analysis model, and an altitude situation analysis model;

[0015] An occupancy comprehensive value determination module, configured to determine an occupancy comprehensive advantage value and an occupancy comprehensive threat value of our aircraft relative to the enemy aircraft based on the situation analysis model;

[0016] A radar comprehensive value determination module, configured to determine a radar comprehensive advantage value and a radar comprehensive threat value of the airborne radar of our aircraft relative to the airborne radar of the enemy aircraft;

[0017] A countermeasure decision determination module, configured to make an optimal countermeasure decision on the next maneuver sequence feature of the enemy aircraft according to the occupancy comprehensive advantage value, the occupancy comprehensive threat value, the radar comprehensive advantage value, and the radar comprehensive threat value.

[0018] According to another aspect of the present invention, an embodiment of the present invention also provides an electronic device, and the electronic device includes:

[0019] At least one processor;

[0020] And a memory communicatively connected to the at least one processor; wherein,

[0021] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the unmanned aerial vehicle intelligent autonomous decision-making method according to any embodiment of the present invention.

[0022] According to another aspect of the present invention, an embodiment of the present invention further provides a computer-readable storage medium storing computer instructions for causing a processor to implement the unmanned aerial vehicle intelligent autonomous decision-making method according to any embodiment of the present invention when executed.

[0023] Through the above technical solution of the embodiment of the present invention, a tactical behavior prediction model is trained through a tactical behavior sample set, and the next maneuver sequence feature of the enemy aircraft is determined according to the action parameter information of the enemy aircraft in a preset time series and the tactical behavior prediction model. On this basis, a situation analysis model is constructed based on the spatial occupancy of our aircraft and the enemy aircraft. Based on this situation analysis model, the comprehensive occupancy advantage value and comprehensive occupancy threat value of our aircraft relative to the enemy aircraft are determined, and the comprehensive radar advantage value and comprehensive radar threat value of the airborne radar of our aircraft relative to the airborne radar of the enemy aircraft are determined. Thus, an optimal countermeasure decision is made on the next maneuver sequence feature of the enemy aircraft according to the comprehensive occupancy advantage value, comprehensive occupancy threat value, comprehensive radar advantage value, and comprehensive radar threat value, which can improve the ability of autonomous intelligent decision-making in a complex battlefield environment and effectively improve the efficiency and accuracy of autonomous intelligent decision-making.

[0024] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present invention, nor is it used to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0026] Figure 1 It is a flowchart of an unmanned aerial vehicle intelligent autonomous decision-making method provided by an embodiment of the present invention;

[0027] Figure 2 It is a perspective view schematic diagram of a horizontal right turn action provided by an embodiment of the present invention;

[0028] Figure 3 It is a perspective view schematic diagram of a horizontal left turn action provided by an embodiment of the present invention;

[0029] Figure 4 It is a perspective view schematic diagram of a large-angle dive turn action provided by an embodiment of the present invention;

[0030] Figure 5 It is a perspective view schematic diagram of a large-angle pull-up turn action provided by an embodiment of the present invention;

[0031] Figure 6 A schematic diagram of a hovering escape maneuver provided by an embodiment of the present invention;

[0032] Figure 7 A flowchart of constructing a situation analysis model in a method for intelligent autonomous decision-making of a drone provided by an embodiment of the present invention;

[0033] Figure 8 A schematic diagram of an air combat position situation provided by an embodiment of the present invention;

[0034] Figure 9 A schematic diagram of a radar coordinate system of an airborne radar of a friendly aircraft and a body coordinate system of an enemy aircraft provided by an embodiment of the present invention;

[0035] Figure 10 A flowchart of another unmanned aerial vehicle intelligent autonomous decision-making method provided by an embodiment of the present invention;

[0036] Figure 11 A structural block diagram of an intelligent autonomous decision-making device for a drone provided by an embodiment of the present invention;

[0037] Figure 12 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0038] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. 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 this field without creative work should fall within the scope of protection of the present invention.

[0039] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0040] In one embodiment,Figure 1 The flowchart of an intelligent autonomous decision-making method for an unmanned aerial vehicle provided by an embodiment of the present invention. This embodiment is applicable to the situation of making intelligent confrontation decisions against enemy aircraft in an air battle. This method can be executed by an intelligent autonomous decision-making device for an unmanned aerial vehicle, which can be implemented in the form of hardware and / or software, and can be configured in an electronic device. As Figure 1 shown, the method includes:

[0041] S110. Obtain a pre-constructed tactical behavior sample set, and use the tactical behavior sample set to train a behavior prediction model to obtain a trained tactical behavior prediction model.

[0042] Among them, the behavior prediction model can be a recurrent neural network model for processing sequence data. This recurrent neural network model can be a neural network architecture such as a GRU gated recurrent unit or an LSTM long short-term memory neural network. In this embodiment, the recurrent neural network model can be based on the MobileNet-V1 network architecture.

[0043] In this embodiment, after obtaining the tactical behavior sample set, the tactical behavior sample set is used as a training sample set to train the behavior prediction model, so as to obtain a trained tactical behavior prediction model for predicting the next maneuver sequence feature of the enemy aircraft. In some embodiments, data preprocessing is performed on each sample in the tactical behavior sample set, and the target sample set is input into the behavior prediction model to optimize the model parameters until the model parameters reach the optimal or the predefined loss function reaches the minimum, so as to obtain a trained tactical behavior prediction model. In some embodiments, the gradient descent method, Bayesian optimization method, genetic algorithm, etc. can be used to optimize the model parameters, and the model parameters can include the learning rate, regularization, etc. Of course, to facilitate reducing the computational amount, the structure of the model can also be correspondingly improved. Exemplarily, depth-wise separable convolution is used to reduce the amount of computation and the number of parameters; a depthwise separable convolution structure is formed by using DWCONV+(1×1CONV) to reduce the model parameters and forward inference.

[0044] In one embodiment, the construction of the tactical behavior sample set includes: obtaining the motion trajectory information of the unmanned aerial vehicle at a historical moment, determining the basic actions of the unmanned aerial vehicle according to the motion trajectory information; forming maneuver actions from the basic actions, and forming tactical behaviors based on at least two consecutive time series of maneuver actions; wherein, the tactical behavior represents the tactical intention; a continuous maneuver sequence of the unmanned aerial vehicle with typical tactical behaviors is established based on a preset aircraft dynamics model, a preset flight envelope, and the tactical intention to form a tactical behavior library, and the tactical behavior library is used as the tactical behavior sample set.

[0045] Among them, the maneuver sequence features of typical tactical behaviors at least include: no tactical maneuver, dive sharp turn, spiral escape maneuver, small-radius loop, snake maneuver, bell maneuver, Herbst maneuver, Immelmann maneuver, cobra maneuver, vertical snake maneuver, and zoom sharp turn; each maneuver sequence feature corresponds to corresponding maneuver parameter values; the maneuver parameter values at least include: altitude value range, speed value range, pitch angle value, yaw angle value, roll angle value.

[0046] In this embodiment, the preset aircraft dynamics model is a nonlinear simulation model of the aircraft, which can be a simulation model of a specific type of UAV selected by the user, including the structure and parameters of the aircraft. The parameters may include but are not limited to mass, reference wingspan, barometric altitude, moment of inertia, inertia tensor, etc. This model can be written using Matlab and S functions. The C language used in the S function needs to be compiled before running. You can enter "mex + file name" in the command line.

[0047] The preset flight envelope in this embodiment refers to a closed geometric figure that uses parameters such as flight speed, altitude, overload, and environmental temperature as coordinates to represent the flight range of the aircraft and the aircraft usage limit conditions. The flight envelope can include the level flight speed envelope, speed overload envelope, and gust overload envelope. In this embodiment, the motion trajectory information refers to the trajectory points of the motion, which are the original particle trajectories obtained by the radar. Each radar detection point can be regarded as a particle, and its motion trajectory is represented by a time series. The motion trajectory point is a motion trajectory point in a three-dimensional space.

[0048] In this embodiment, the basic actions of the UAV can be determined based on the motion trajectory information, and then the maneuvering actions can be formed from the basic actions. The tactical behavior is composed of at least two consecutive time-series maneuvering actions, which can represent the tactical intention. Thus, a continuous maneuvering sequence of the UAV for typical tactical behaviors is established based on the preset aircraft dynamics model, preset flight envelope, and tactical intention to form a tactical behavior library, and the tactical behavior library is used as a tactical behavior sample set. The basic maneuvering actions in this embodiment mainly include level flight, pull-up, dive, left turn, right turn, and combinations of the basic maneuvering actions. Finally, the maneuvering actions to be recognized are determined as: level flight, large-angle dive, large-angle pull-up, general dive, dive left turn, dive right turn, pull-up left turn, pull-up right turn, horizontal right turn, horizontal left turn, general pull-up, a total of 11 types of maneuvering actions. Thus, the tactical behavior is composed of at least two consecutive time-series maneuvering actions. Among them, the tactical behavior represents the tactical intention. It can be understood that the tactical behavior of a fighter can be decomposed into a maneuvering sequence composed of several basic maneuvering action units, and each tactical behavior reflects a tactical intention. Based on the aircraft dynamics model, flight envelope, and tactical intention, a continuous maneuvering sequence of the aircraft for typical tactical behaviors is established to form a tactical behavior library. In this embodiment, for the recognition and intention understanding of the tactical behavior, the flight maneuvers of the fighter in space and the continuous maneuvers within a certain time period can be observed to judge the basic behavior type of the aircraft. At the same time, the enemy's tactical behavior can be judged according to factors such as the distance between the enemy and us and the position occupancy, and the opponent's intention can be inferred.

[0049] It should be noted that after the basic actions form the maneuvering actions, the maneuvering actions need to be perspectively presented. It can be understood that the performance of the actions in the three-dimensional space is presented in a visual way, so as to help understand the spatial relationship and execution effect of the actions, so as to more intuitively analyze and evaluate the rationality and effect of the actions, and assist in decision-making and optimization. Exemplarily, for better understanding of the action diagram after the basic actions form the maneuvering actions, Figure 2 is a perspective schematic diagram of a horizontal right turn action provided by an embodiment of the present invention; Figure 3 is a perspective schematic diagram of a horizontal left turn action provided by an embodiment of the present invention; Figure 4 is a perspective schematic diagram of a large-angle dive turn action provided by an embodiment of the present invention; Figure 5 is a perspective schematic diagram of a large-angle pull-up turn action provided by an embodiment of the present invention. Similarly, after the tactical behavior is composed of at least two consecutive time-series maneuvering actions, before establishing a continuous maneuvering sequence of the UAV for typical tactical behaviors to form a tactical behavior library, the sample tactical actions also need to be perspectively presented. Similarly, the performance of the actions in the three-dimensional space is presented in a visual way. For better understanding of this sample tactical action perspective, Figure 6 is an action schematic diagram of a spiral escape maneuver provided by an embodiment of the present invention.

[0050] In this embodiment, the maneuver sequence characteristics of typical tactical behaviors and the corresponding maneuver parameter values are stored in the maneuver action library in the form of knowledge. Through the tactical requirements of beyond-visual-range and medium-range air combat, a typical maneuver sequence feature table is constructed, which describes the timing process of the maneuver sequence with altitude, speed, and the three-direction angles of pitch, yaw, and roll as the main features. Among them, each basic maneuver can be realized by combining simpler atomic actions (such as left pull-up, altitude holding, speed holding, heading holding, etc.) in a certain time sequence. In this embodiment, the maneuver action sequence definition of the fighter aircraft in beyond-visual-range air combat is used for illustration, and the maneuver action sequence definition of the fighter aircraft in beyond-visual-range air combat is shown in Table 1. In this embodiment, the typical maneuver sequence characteristics may include typical maneuver names, basic descriptions, altitude descriptions, speed descriptions, and relevant descriptions of the three-direction angles of pitch, yaw, and roll. Exemplarily, the typical maneuver name is the spiral escape maneuver; the basic description is to quickly change the heading to get rid of the tail chase; the altitude is 3000m - 5000m; the speed is 0.4 - 0.6Ma; the three-direction angles of pitch, yaw, and roll are to turn in a circle with the minimum radius and 360°; another example, the typical maneuver name is the bell-shaped maneuver; the basic description is to quickly pull up, when the pitch is close to 90 degrees, hold for 4 seconds, then the aircraft pitches backward with the back facing down, and finally turns to exit the maneuver and resumes the level flight state (turning 360° in a circle); the altitude is below 3000m; the speed is 0.2 - 0.6Ma; the three-direction angles of pitch, yaw, and roll are pitch 80° ± 10°, and then the elevation angle continues to increase without deflection or roll, and turns in a circle in the vertical direction.

[0051] Table 1: Definition of the maneuver action sequence of the fighter aircraft in beyond-visual-range air combat

[0052]

[0053] S120. Obtain the action parameter information of the enemy aircraft in the preset time series, and determine the next maneuver sequence characteristics of the enemy aircraft according to the action parameter information and the tactical behavior prediction model.

[0054] Among them, the action parameter information in the preset time series can be the action parameters corresponding to each time in a certain continuous time. The action parameter information at least includes altitude, speed, pitch angle, yaw angle, roll angle, longitude, and latitude. The preset time series is a continuous time series.

[0055] In this embodiment, the next maneuver sequence feature can be understood as the maneuver sequence feature that the enemy aircraft may correspond to at the next moment. This maneuver sequence feature is in the form of a maneuver action, and this maneuver action corresponds to a corresponding maneuver action category. The next maneuver sequence feature may include, but is not limited to, various types such as a climbing snap roll and a spiral breakaway maneuver. The next maneuver sequence feature corresponds to corresponding maneuver parameter values, and the maneuver parameter values may include an altitude value, a speed value, a pitch angle value, a yaw angle value, and a roll angle value.

[0056] In this embodiment, the action parameter information of the enemy aircraft in a preset time series can be used to determine the next maneuver sequence feature of the enemy aircraft according to the action parameter information and the tactical behavior prediction model. In some embodiments, the action parameter information can be preprocessed, and then the preprocessed action parameter information can be directly input into the tactical behavior prediction model to predict the next maneuver sequence feature of the enemy aircraft; in other embodiments, the decision-making process of the enemy aircraft pilot can be simulated through a reinforcement learning model, and an optimal maneuver can be generated based on a reward function (such as "occupying a favorable position" or "evading a missile"). In other embodiments, a game theory model can also be used. First, it is assumed that the enemy aircraft adopts a Nash equilibrium strategy to predict its optimal response action. This embodiment does not limit this here.

[0057] S130. Construct a situation analysis model based on the spatial positions of our aircraft and the enemy aircraft; wherein, the situation analysis model includes: an angle situation analysis model, a distance situation analysis model, a speed situation analysis model, and an altitude situation analysis model.

[0058] Among them, the spatial position can also be called the air combat position, which can be understood as the position of our aircraft and the enemy in the confrontation. The key positions of both sides in the three-dimensional space can be controlled or predicted through maneuver actions to gain tactical advantages (such as an attack window, a defense barrier, or an energy advantage). It can be understood that when both sides conduct confrontation, they need to consider the positions they occupy. For example, in an air combat, the higher the standing position, the better the safety.

[0059] In this embodiment, the airborne radar of our aircraft is used to detect the enemy aircraft to determine the spatial position of the enemy aircraft relative to our aircraft. According to the spatial position, the attitude angle of the enemy aircraft relative to the incident beam of the airborne radar of our aircraft at each moment is determined, and then based on the attitude angle, the radar cross section (RCS value) of the enemy aircraft exposed to the airborne radar of our aircraft at each moment is determined. On this basis, the spatial situation is analyzed according to the RCS value of the enemy aircraft to construct a situation analysis model, and the situation analysis model includes: an angle situation analysis model, a distance situation analysis model, a speed situation analysis model, and an altitude situation analysis model; specifically, the situation analysis model is expressed by the formula: ; where, where, represents the angle situation analysis model, is represented as a speed situation analysis model, is represented as a distance situation analysis model, is represented as an altitude situation analysis model; , , and respectively represent the weight coefficients corresponding to the angle situation analysis model, speed situation analysis model, distance situation analysis model, and altitude situation analysis model.

[0060] In some other embodiments, a UAV air combat confrontation movement model can be constructed based on the positions, speeds, and attitudes of the UAVs on both sides of the air combat. Then, the UAV air combat confrontation movement model is combined with the air combat situation elements to establish an air combat situation assessment model based on the dominant function method; in some other embodiments, the combat advantage index parameters of our UAV relative to the enemy UAV, the threat situation index parameters of the enemy UAV relative to our UAV, and the target value index parameters can also be calculated first, so as to calculate the comprehensive situation function according to the combat advantage index parameters, threat situation index parameters, and target value index parameters. This embodiment is not limited herein.

[0061] S140. Determine the occupancy comprehensive advantage value and occupancy comprehensive threat value of our aircraft relative to the enemy aircraft based on the situation analysis model.

[0062] Among them, the occupancy comprehensive advantage value can be understood as the advantage value of the space occupancy of our aircraft relative to the enemy aircraft, which can represent the advantage probability of our aircraft in space occupancy. This advantage can include angle advantage, distance advantage, altitude advantage, and speed advantage; the occupancy comprehensive threat value can be understood as the threat value of the space occupancy of our aircraft relative to the enemy aircraft, which can represent the threat probability of the enemy aircraft to our aircraft in space occupancy. Similarly, this threat can include angle threat, distance threat, altitude threat, and speed threat.

[0063] In this embodiment, through the weight allocation method based on the correlation between indicators (Criteria Importance Through Intercriteria Correlation, CRITIC), the variable weight coefficients corresponding to altitude, angle, distance, and speed in the situation analysis model can be solved first, so as to obtain the occupancy comprehensive advantage value and occupancy comprehensive threat value of our aircraft relative to the enemy aircraft based on the variable weight coefficients corresponding to altitude, angle, distance, and speed; it can be understood that corresponding weights are assigned to altitude, angle, distance, and speed, and these weights can be adjusted according to tactical requirements, and the occupancy comprehensive advantage value and occupancy comprehensive threat value of our aircraft relative to the enemy aircraft are obtained by weighted summation.

[0064] S150. Determine the radar comprehensive advantage value and radar comprehensive threat value of the airborne radar of our own aircraft relative to that of the enemy aircraft.

[0065] Among them, the radar comprehensive advantage can be understood as the advantage value of the airborne radar of our own aircraft relative to that of the enemy aircraft; the radar comprehensive threat value can be understood as the threat value of the airborne radar of our own aircraft relative to that of the enemy aircraft. In this embodiment, the main factors affecting the radar comprehensive threat value and radar comprehensive advantage value in air-to-air combat are: the maximum radar detection range (antenna gain, transmission power), radar detection ability (signal-to-noise ratio), and radar cross section.

[0066] In this embodiment, the airborne radar of our own aircraft is used to detect the enemy aircraft to determine the spatial position of the enemy aircraft relative to our own aircraft. According to the spatial position, the attitude angle of the incident beam of the enemy aircraft relative to the airborne radar of our own aircraft at each moment is determined. Based on the attitude angle, the RCS value of the enemy aircraft is determined. On this basis, according to the power received by the radar antenna and the generated noise power, the signal-to-noise ratio is determined to characterize the radar monitoring ability. According to the signal-to-noise ratio and the RCS value of the enemy aircraft, the maximum detection range of the airborne radar of our own aircraft is determined. Thus, through simulation software, the empirical value results corresponding to the maximum detection range, signal-to-noise ratio, and RCS value are simulated respectively. Each empirical value result is used as the corresponding weight coefficient. Finally, by calculating the weighted value, the radar comprehensive advantage value and radar comprehensive threat value of the airborne radar of our own aircraft relative to that of the enemy aircraft are obtained.

[0067] S160. Make an optimal countermeasure decision on the next maneuver sequence characteristics of the enemy aircraft according to the occupancy comprehensive advantage value, occupancy comprehensive threat value, radar comprehensive advantage value, and radar comprehensive threat value.

[0068] Among them, the optimal countermeasure decision can be understood as the tactical action of the countermeasure decision finally executed by our own aircraft.

[0069] In this embodiment, through the occupancy comprehensive advantage value and occupancy comprehensive threat value under spatial occupancy, as well as the radar comprehensive advantage value and radar comprehensive threat value under the radar, an optimal countermeasure decision is made on the next maneuver sequence characteristics of the enemy aircraft; in some embodiments, the comprehensive advantage value is determined through the occupancy comprehensive advantage value and radar comprehensive advantage value, and the comprehensive threat value is determined according to the occupancy comprehensive threat value and radar comprehensive threat value. Then, corresponding thresholds are preset for the comprehensive threat value and comprehensive advantage value. Thus, the final countermeasure action is carried out according to the set thresholds. Of course, in addition to setting thresholds, it can also be determined whether there is interference through the comprehensive threat value and comprehensive advantage value. If there is interference, the optimal countermeasure action can be set to a level flight action. In other embodiments, it is also possible to establish a UAV air combat countermeasure strategy model to conduct UAV air combat confrontation. This embodiment does not limit this here.

[0070] In the above technical solution of the embodiment of the present invention, a tactical behavior prediction model is trained through a tactical behavior sample set, and the next maneuver sequence feature of the enemy aircraft is determined according to the action parameter information of the enemy aircraft under a preset time series and the tactical behavior prediction model. On this basis, a situation analysis model is constructed based on the spatial occupancy of our aircraft and the enemy aircraft. Based on this situation analysis model, the comprehensive occupancy advantage value and comprehensive occupancy threat value of our aircraft relative to the enemy aircraft are determined, and the comprehensive radar advantage value and comprehensive radar threat value of the airborne radar of our aircraft relative to the airborne radar of the enemy aircraft are determined. Thus, an optimal countermeasure decision is made on the next maneuver sequence feature of the enemy aircraft according to the comprehensive occupancy advantage value, comprehensive occupancy threat value, comprehensive radar advantage value and comprehensive radar threat value, which can improve the ability of autonomous intelligent decision-making in a complex battlefield environment and effectively improve the efficiency and accuracy of autonomous intelligent decision-making.

[0071] In one embodiment, Figure 7 FIG. is a flowchart of constructing a situation analysis model in an unmanned aerial vehicle intelligent autonomous decision-making method provided by an embodiment of the present invention. On the basis of the above embodiments, the trained tactical behavior prediction model is obtained by training the behavior prediction model with a tactical behavior sample set, the next maneuver sequence feature of the enemy aircraft is determined according to the action parameter information and the tactical behavior prediction model, and the construction of the situation analysis model based on the spatial occupancy of our aircraft and the enemy aircraft is further refined.

[0072] As Figure 7 shown, in the unmanned aerial vehicle intelligent autonomous decision-making method in this embodiment, constructing a situation analysis model may specifically include the following steps:

[0073] S710. Obtain a pre-constructed tactical behavior sample set, and perform zero-padding on the time series of each sample in the tactical behavior sample set to make the time series lengths of all samples consistent to obtain processed samples; wherein, the tactical behavior sample set includes the maneuver sequence features of typical tactical behaviors and the maneuver parameter values corresponding to each maneuver sequence feature.

[0074] In this embodiment, the flight time of each tactical action is not uniform. For example, if 0.5 s is defined as a time series, and most samples are statistically analyzed, 60% of the samples have a total action time near 40 s, while very few are near 20 s and 120 s. Since the samples input into the model need to be time series of a fixed length, therefore, it is necessary to perform zero-padding on the time series of each sample in the tactical behavior sample set to make the time series lengths of all samples consistent to obtain processed samples. It can be understood that the samples are padded with zeros to the maximum time series. Exemplarily, the maximum time series is 150 s and there are 300 sequences. For a 120 s sample, it is equivalent to padding the action to 150 s, and all the data are filled with 0, so as to meet the requirements of fixed input of the model.

[0075] S720. Perform data augmentation on the processed samples to obtain the samples after data augmentation.

[0076] In this embodiment, due to operating habits or changes in some environments in the real scenario, there will be some deviations in the slope and angle of attack during each flight. For the model, if this historical information is not input during the training process, the prediction result will not be accurate enough. The main problem is to improve the generalization ability of the model, but too strong generalization ability will bring insufficient memory ability of the model, and a balance point needs to be found. Therefore, data augmentation is used in actual training, Gaussian white noise is added, and by adding small perturbations, the model training sample library is increased to improve the generalization ability of the model.

[0077] S730. For the samples after data augmentation, insert a level flight action after the end of each sample's action to initialize the next sample's action, and obtain the target sample set.

[0078] In this embodiment, for the samples after data augmentation, the UAV maneuvering actions to be recognized are continuous data, and the model needs to recognize the starting point and ending point of the action for continuous prediction. The model cannot obtain this flag bit or information. Therefore, a level flight action is inserted after the end of each sample's tactical action in the training samples. By recognizing continuous level flights, the recognition of the next tactical action is initialized.

[0079] S740. Input the target sample set into the behavior prediction model to optimize the model parameters until the model parameters reach the optimal value or the predefined loss function reaches the minimum, and obtain the trained tactical behavior prediction model.

[0080] Among them, the behavior prediction model is a recurrent neural network model; the recurrent neural network model is based on the MobileNet-V1 network architecture; the convolutional layer of the recurrent neural network model consists of DWCONV+(1×1CONV) to form a depthwise separable convolution structure, reducing the model parameters and forward inference.

[0081] In this embodiment, the activation function in the behavior prediction model is RELU, which can improve the forward inference calculation speed. In addition, the SDCA module, DWCONV+1*1CONV+, after improving the model structure, can greatly reduce the amount of calculation through improvement. The attention module can ensure the model learning and memory ability of the structure with low computational complexity, reduce the network width, and add Attention+Shortcut after the depthwise separable convolution structure, so as not to cause gradient dissipation. Using RELU, for negative inputs, the gradient of RELU is 0. Due to the existence of shortcut, there is no need to do more protection mechanisms, and the calculation speed is improved first.

[0082] In this embodiment, when training the behavior prediction model, the target sample set is input into the improved behavior prediction model to optimize the model parameters until the model parameters reach the optimal or the predefined loss function reaches the minimum, and the trained tactical behavior prediction model is obtained. Exemplarily, after initializing the model parameters (random normal distribution), in the first training, the samples of this batch are trained for 1000 times to stop iteration, and a basic effect model is obtained. The basic model is loaded, and the iteration stops when the threshold model parameters reach the optimal. The current neural network model stops iteration when the loss is less than 10. Then the model is loaded again, and the iteration stops when the loss is less than 5. Then the model is combined with the decision model, and according to the simulation data results, the problems existing in the model are determined, the samples of this problem are increased, and the model is continuously trained and optimized.

[0083] S750. Obtain the action parameter information of the enemy aircraft in the preset time series, perform data preprocessing on the action parameter information, and input the preprocessed action parameter information into the tactical behavior prediction model to predict the next maneuver sequence characteristics of the enemy aircraft.

[0084] In this embodiment, the ways to perform data preprocessing on the action parameter information may include, but are not limited to, zero-padding operation, sample enhancement operation, inserting level flight actions, etc.

[0085] In this embodiment, data preprocessing is performed on the action parameter information of the enemy aircraft in the preset time series, and the preprocessed action parameter information is input into the tactical behavior prediction model to predict the next maneuver sequence characteristics of the enemy aircraft, so that our aircraft can make the next intelligent decision by judging the behavior, the spatial occupancy of our aircraft and the enemy aircraft, the radar advantage, etc.

[0086] S760. Use the onboard radar of our aircraft to detect the enemy aircraft to determine the spatial position of the enemy aircraft relative to our aircraft.

[0087] In this embodiment, the onboard radar of our aircraft will detect the position of the enemy aircraft in real time. The detected spatial position relationship may include: our aircraft is above the enemy aircraft, our aircraft is below the enemy aircraft, and our aircraft and the enemy aircraft are on the same horizontal line. In this embodiment, for better understanding the air combat occupancy situation between our aircraft and the enemy, Figure 8 is a schematic diagram of an air combat occupancy situation provided by an embodiment of the present invention. As Figure 8 shown, A is our combat aircraft, and T is the airspace target aircraft (i.e., the enemy aircraft), and are the velocity vectors corresponding to our aircraft and the enemy aircraft respectively, is the heading angle. It can be seen that when our aircraft and the enemy aircraft are on the same horizontal plane, the attitude angle of our aircraft exposed to the enemy aircraft's radar is equal to the azimuth angle. In an air combat, the fighter aircraft first faces the detection threat of the enemy aircraft's radar. How to select an appropriate approach method according to the situation of both sides to ensure that it is exposed to the enemy radar threat in a better attitude and reduce the detection probability of the enemy aircraft is the main means to achieve stealth. On the other hand, the static RCS characteristics of a fighter aircraft are a relatively complex variable, which needs to be measured in a real environment or anechoic chamber, or simulated and calculated. Under the conditions of air combat confrontation, the attitude angle of the enemy aircraft exposed to our aircraft's airborne radar changes in real time, and the RCS shows strong dynamics.

[0088] S770. Determine the attitude angle of the enemy aircraft relative to the incident beam of our aircraft's airborne radar at each moment according to the spatial position; wherein, the attitude angle includes: azimuth angle and pitch angle.

[0089] In this embodiment, the attitude angle of the moving target relative to the incident wave of the airborne radar is solved according to the spatial position of the airborne radar and the target. Specifically, Figure 8 is taken as an example to establish a coordinate system. Figure 9 is a schematic diagram of the radar coordinate system of our aircraft's airborne radar and the body coordinate system of the enemy aircraft provided by an embodiment of the present invention. As Figure 9 shown. is the origin of the coordinate system of our aircraft's airborne radar. is the coordinate system of our aircraft's airborne radar (taken as the spherical rectangular coordinate system at the position of the carrier aircraft, the x-axis points due east, the y-axis points due north, and the z-axis is vertically upward). r is the straight-line distance from the origin of the airborne radar to the enemy aircraft. The position coordinates of the enemy aircraft in the coordinate system of our aircraft's airborne radar are denoted as ( ). is the origin of the coordinate system in the body coordinate system of the enemy aircraft. is the body coordinate system of the enemy aircraft. is the projection of the enemy aircraft on the xy plane of our aircraft's coordinate system, which can be used to calibrate the angle. During the movement of the enemy aircraft, it is exposed to the airborne radar beam in a certain attitude. Therefore, it is necessary to solve the azimuth angle and pitch angle of our aircraft's radar line of sight in the body coordinate system of the enemy aircraft. Given any point (x, y, z) in the coordinate system of our aircraft's airborne radar, its coordinates in the body coordinate system of the enemy aircraft are denoted as , then the conversion relationship between the airborne radar coordinate system and the body coordinate system of the enemy aircraft is: , where is the rotation matrix from the body coordinate system of the enemy aircraft to the airborne radar coordinate system. When the position coordinates of the detection carrier aircraft in the body coordinate system of the enemy aircraft are , where , then the position of the detection carrier aircraft in the body coordinate system of the enemy aircraft and its coordinate form are expressed as: , where , Azimuth angle, and depression angle; all of the above variables change dynamically with time. By calculating the azimuth angle and elevation angle of the radar incident beam at each moment, the RCS value of the enemy aircraft can be obtained. This RCS sequence represents the RCS value of the target exposed to the airborne radar at each moment when the target moves along a certain trajectory and attitude.

[0090] S780. Determine the RCS value of the enemy aircraft based on the attitude angle; where the RCS value represents the RCS value of the enemy aircraft exposed to the airborne radar of our aircraft at each moment.

[0091] In this embodiment, the RCS value is also called the radar signature and is an index to measure the degree to which an object can be detected by radar. The larger the RCS, the easier it is for the object to be detected. The RCS value of the enemy aircraft exposed to the airborne radar of our aircraft at each moment can be determined based on the attitude angle.

[0092] S790. Analyze the spatial situation based on the RCS value of the enemy aircraft to construct a situation analysis model.

[0093] In this embodiment, the spatial situation is analyzed through the RCS value of the enemy aircraft to construct a situation analysis model. Specifically, the spatial situation refers to judging the advantages and disadvantages based on the motion state information and RCS value between our aircraft and the enemy aircraft. The key to an air battle is to keep the enemy aircraft within the attack area of our aircraft while avoiding our aircraft entering the attack area of the enemy aircraft. There are four influencing factors that have a greater impact on the attack area range: angle, speed, distance, and altitude. Based on this, a spatial situation advantage function is established.

[0094] In this embodiment, the situation analysis model is expressed by the formula: ; where represents the angle situation analysis model, represents the speed situation analysis model, represents the distance situation analysis model, represents the altitude situation analysis model; , , and represent the weight coefficients corresponding to the angle situation analysis model, speed situation analysis model, distance situation analysis model, and altitude situation analysis model, respectively.

[0095] Among them, the angle situation analysis model includes: azimuth angle advantage and approach angle advantage , which are respectively expressed by the formulas: ; in the formula, represents the azimuth angle advantage; represents the approach angle advantage; , respectively represent the weights of the azimuth angle and the approach angle, ; where,

[0096] , ; in the formula, represents the azimuth angle advantage; represents the approach angle advantage; represents the approach angle, represents the azimuth angle, represents the maximum search azimuth angle, represents the maximum off-axis launch angle, represents the non-escape cone angle, where, ; the velocity situation analysis model is expressed as: when , ; when , the velocity situation analysis model is expressed as: ; in the formula, represents the velocity of our own aircraft; represents the velocity of the enemy aircraft; represents the optimal air combat velocity of our own aircraft; the distance situation analysis model is expressed as: , in the formula, represents the distance advantage; represents the distance between our own aircraft and the enemy aircraft; represents the maximum detection distance of the radar, represents the maximum attack distance, represents the non-escape maximum distance, represents the non-escape minimum distance, represents the minimum attack distance, where, ; the altitude situation analysis model is expressed by the formula: ; in the formula, represents the altitude advantage; represents the altitude of our own aircraft; represents the altitude of the enemy aircraft; represents the optimal occupancy altitude of our own aircraft.

[0097] In this embodiment, there is a coupling relationship between the azimuth angle and the approach angle in the influence on the air combat situation. The main influencing factors of the angle situation function are the target azimuth angle and the target approach angle , , , the smaller, the larger, the larger the range of the air-to-air missile attack area and the better the angle situation.

[0098] In one embodiment, Figure 10The flowchart of another intelligent autonomous decision-making method for an unmanned aerial vehicle provided by an embodiment of the present invention. On the basis of the above embodiments, this embodiment further refines the training of a trained tactical behavior prediction model by using a tactical behavior sample set for a behavior prediction model, determining the next maneuver sequence characteristics of an enemy aircraft according to action parameter information and the tactical behavior prediction model, and constructing a situation analysis model based on the spatial occupancy of our aircraft and the enemy aircraft.

[0099] As Figure 10 shown, in the intelligent autonomous decision-making method for an unmanned aerial vehicle in this embodiment, constructing a situation analysis model may specifically include the following steps:

[0100] S1010. According to the preset CRITIC method, inversely solve the first variable weight coefficients corresponding to each first index in the situation analysis model, and determine the sum of the products of each first variable weight coefficient and the corresponding first index, so as to obtain the comprehensive threat value of our aircraft relative to the enemy aircraft; where the first index includes: altitude, angle, distance, and speed; where the first variable weight coefficient is an occupancy threat coefficient.

[0101] Among them, the preset CRITIC method is an objective weight allocation method, which determines the weights of each index by analyzing the comparison intensity (standard deviation) and conflict (correlation) between the indexes. It can be understood that the weights are determined by combining the conflict between the indexes and the influence of data variation on the index weights. In this embodiment, in the air combat occupancy evaluation, the weights of the comprehensive advantage value and the comprehensive threat value can be dynamically adjusted. Comparison intensity: The larger the standard deviation of the index, the higher the degree of data dispersion, and a higher weight should be given. Conflict: The lower the correlation between the indexes, the less information overlap, and the higher the weight should be. Comprehensive weight: Combine the comparison intensity and conflict to avoid subjective preference.

[0102] In this embodiment, inversely can be understood as the smaller the value of the first index, the better. Exemplarily, the positive index 1: ; Positive can be understood as the larger the value of the first index, the better. Exemplarily, the inverse index 1: .

[0103] In this embodiment, during the combat process from far to near between the two sides in air combat, the weight values of various threat and advantage factors change with the changes of conditions such as the occupancy and maneuver of the two sides. Therefore, it is necessary to achieve dynamic weight changes. In this embodiment, since the method of using the preset CRITIC method to solve the first variable weight coefficients corresponding to each first index in the situation analysis model and the second weight coefficient is the same, it should be noted that although both use the preset CRITIC method for solution, when calculating the threat value, inverse processing is adopted, that is, it is desired that the index is as small as possible; when calculating the advantage value, positive processing is adopted, that is, it is desired that the index is as large as possible.

[0104] In this embodiment, the preset CRITIC method is used to solve the first variable weight coefficients corresponding to height, angle, distance, and speed in the situation analysis model. The specific steps may include: 1) Construct a situation evaluation matrix; where the situation evaluation matrix includes the index value of the j-th index at the i-th moment, and j takes 1, 2, 3, 4, where 1 represents the angle, 2 represents the speed, 3 represents the distance, and 4 represents the height; 2) Normalize the situation evaluation matrix to eliminate the influence of index types and dimensions, that is, when calculating the threat value, reverse processing is used, that is, the smaller the index, the better; when calculating the advantage value, forward processing is used, that is, the larger the index, the better. For example, the forward factor 1: ; The reverse factor 1: . 3) For the normalized situation evaluation matrix, the conflict between indicators and the comparison intensity between data are represented by the correlation coefficient and the standard deviation, and the comprehensive information volume of the indicators is evaluated; 4) The constant weight vector between indicators can be determined by evaluating the comprehensive information volume of the indicators, and the state variable weight vector is constructed. The variable weight vector is obtained by the Hadamard of the state variable weight vector and the constant weight vector.

[0105] More specifically, 1) Construct a situation evaluation matrix, which consists of a real-time evaluation state information set composed of t moments and an index set composed of p situation evaluation indicators . The j-th index value at the i-th moment is , then the situation evaluation matrix is expressed as: ; 2) Normalize the evaluation matrix: Normalize the benefit type, cost type, and fixed type indicators to eliminate the influence of index types and dimensions. Benefit indicators: , Cost type indicators: , Fixed type indicators: , where: is the best value of the fixed type indicator. After normalization, the situation matrix obtained is . It should be noted that when calculating the threat value, reverse processing is used, that is, the smaller the index, the better. 3) Calculate the comprehensive information volume between evaluation indicators: The conflict between indicators and the comparison intensity between data are represented by the correlation coefficient and the standard deviation , and the comprehensive information volume between evaluation indicators is: , where: , is the average value of the j-th indicator, , where, is the k-th indicator, is the j-th index. 4) Determine the constant weight of the evaluation index: The constant weight vector between the indexes can be determined through the comprehensive information quantity of the evaluation index as: , where: , and the corresponding variable weight vector W(X) can be expressed as: , where is the variable weight of the j-th situation evaluation index. The main method of variable weight is to construct a state variable weight vector to realize the weighted adjustment of the state and avoid state imbalance. The state variable weight vector can be expressed as a mapping relationship: , where , and the variable weight vector can be obtained through the Hadamard product of the state variable weight vector and the constant weight vector w: . The core of variable weight solution lies in constructing an equilibrium function according to the influence degrees of angle, speed, distance, and altitude on the air combat situation, and calculating the state variable weight vector through the derivative of the equilibrium function. Its calculation formula is as follows: , where: B(x) represents the equilibrium function of the situation evaluation index x. As the engagement distance between the two sides approaches, the influence degrees of the angle, speed, and altitude indexes on the situation become greater and greater. From medium range to short range, both sides are within the detection range of each other, and it is necessary for the fighter to maneuver and take positions to form an attack condition. A positive correlation non-linear function is used for incentive variable weight, and the equilibrium function is expressed as: , j = 1, 2, 3, 4, where is the variable weight factor, j = 1, 2, 3, 4, is the incentive variable weight, is the penalty variable weight, and the value of j in the formula is each threat factor, where 1 is the angle, 2 is the speed, 3 is the distance, and 4 is the altitude. On the other hand, the influence degrees of the distance and combat ability indexes on the situation become smaller and smaller as the distance shrinks. The key to determining the outcome of the air combat lies in preferentially forming an attack condition, and a non-linear function is used for penalty variable weight. The equilibrium function is expressed as: . The tactical intention situation belongs to a dynamic process, which reflects the position-taking process and potential trend of the fighter. Based on this analysis, the comprehensive equilibrium function is expressed as: . The state variable weight vectors of each index are expressed as: . To sum up, the variable weights of each index are expressed as: .

[0106] In this embodiment, after obtaining the first variable weight coefficients corresponding to the distance, altitude, angle, and speed respectively, calculate the weight sum between the distance, altitude, angle, and speed and the corresponding first variable weight coefficients respectively to obtain the comprehensive threat value of the position-taking of our aircraft relative to the enemy aircraft.

[0107] S1020. According to the preset CRITIC method, forward solve the second variable weight coefficients corresponding to each first index in the situation analysis model, and determine the sum of the products of each second variable weight coefficient and the corresponding first index, so as to obtain the comprehensive occupancy advantage value of our aircraft relative to the enemy aircraft; wherein, the second variable weight coefficient is the occupancy advantage coefficient.

[0108] In this embodiment, since the method of using the preset CRITIC method to solve the second variable weight coefficients corresponding to each first index in the situation analysis model is the same as the method of solving the first weight coefficient above, this embodiment will not give specific explanations. In this embodiment, after obtaining the second variable weight coefficients corresponding to distance, altitude, angle and speed respectively, calculate the weight sum between distance, altitude, angle and speed and the corresponding second variable weight coefficients respectively, so as to obtain the comprehensive occupancy advantage value of our aircraft relative to the enemy aircraft.

[0109] S1030. Determine the signal-to-noise ratio according to the power received by the radar antenna of our aircraft and the generated noise power; wherein, the signal-to-noise ratio characterizes the radar monitoring ability.

[0110] In this embodiment, the signal-to-noise ratio can be determined by the power received by the radar antenna and the generated noise power, and this signal-to-noise ratio characterizes the detection ability of the radar. Specifically, first determine the radiation power per unit solid angle when the target intercepts the incident wave power and then uniformly radiates it into the entire space, so as to determine the power received by the radar antenna based on the received echo and the radiation power per unit solid angle, and finally determine the signal-to-noise ratio through the power received by the radar antenna and the generated noise power.

[0111] S1040. Obtain the RCS value of the enemy aircraft, and determine the maximum detection distance of the airborne radar of our aircraft according to the signal-to-noise ratio and the RCS value of the enemy aircraft.

[0112] In this embodiment, the maximum detection distance of the airborne radar of our aircraft is determined according to the signal-to-noise ratio and the RCS value of the enemy aircraft. In this embodiment, the calculation method of this maximum detection distance is the calculation method in the prior art, and this embodiment will not give specific introduction here.

[0113] S1050. Take the maximum detection distance, signal-to-noise ratio and RCS value as the second indicators, and simulate and calculate the first empirical value results and the second empirical value results corresponding to each second indicator according to the preset digital combat simulator DCS software. Take the first empirical value results as the corresponding first weight coefficients respectively, and take the second empirical value results as the corresponding second weight coefficients respectively.

[0114] Wherein, the first weight coefficient is the radar threat coefficient corresponding to each second indicator; the second weight coefficient is the radar advantage coefficient corresponding to each second indicator.

[0115] Among them, the preset digital combat simulator DCS software is a simulation platform in the prior art that focuses on the simulation of fighter jets, helicopters, etc. The first empirical value results and the second empirical value results corresponding to the maximum detection range, signal-to-noise ratio, and RCS value can be obtained through simulation. The first empirical value result and the second empirical value result are used as the corresponding first weight coefficient and second weight coefficient respectively to calculate the radar threat value and the radar advantage value. It can be understood that the first empirical value result and the second empirical value result are the radar threat empirical value result and the radar advantage empirical value result obtained during the simulation process of the engagement between our aircraft and the enemy aircraft. The radar threat empirical value result is used as the first weight coefficient, and the radar advantage empirical value result is used as the second weight coefficient.

[0116] In this embodiment, in DCS, the empirical value results of the maximum detection range, signal-to-noise ratio (SNR), and RCS value of the radar are affected by balance. In this embodiment, the onboard radar of our aircraft is used to detect the enemy aircraft to determine the spatial position of the enemy aircraft relative to our aircraft. According to the spatial position, the azimuth angle and elevation angle of the incident beam of the enemy aircraft relative to the onboard radar of our aircraft at each moment are determined. Thus, the RCS value of the enemy aircraft is determined based on the azimuth angle and elevation angle. The maximum detection range, signal-to-noise ratio, and RCS value are used as the second indicators, and the first empirical value results and the second empirical value results corresponding to each second indicator are simulated and calculated according to the preset digital combat simulator DCS software. The first empirical value results are used as the corresponding first weight coefficients respectively, and the second empirical value results are used as the corresponding second weight coefficients respectively.

[0117] S1060. Determine the sum of the first products of each first weight coefficient and the corresponding second indicator to obtain the radar comprehensive advantage value of the onboard radar of our aircraft relative to the onboard radar of the enemy aircraft, and determine the sum of the second products of each second weight coefficient and the corresponding second indicator to obtain the radar comprehensive threat value of the onboard radar of our aircraft relative to the onboard radar of the enemy aircraft.

[0118] In this embodiment, after obtaining the weight coefficients of the maximum detection range, signal-to-noise ratio (SNR), and RCS value, calculate the sum of the first products of each first weight coefficient and the corresponding second indicator to obtain the radar comprehensive advantage value of the onboard radar of our aircraft relative to the onboard radar of the enemy aircraft, and calculate the sum of the second products of each second weight coefficient and the corresponding second indicator to obtain the radar comprehensive threat value of the onboard radar of our aircraft relative to the onboard radar of the enemy aircraft.

[0119] S1070. Determine the comprehensive advantage value based on the occupancy comprehensive advantage value and the radar comprehensive advantage value, and determine the comprehensive threat value based on the occupancy comprehensive threat value and the radar comprehensive threat value.

[0120] In this embodiment, the average value can be obtained by solving the sum of the occupancy comprehensive advantage value and the radar comprehensive advantage value, and then the average value is used as the comprehensive advantage value. Similarly, the average value is obtained by solving the sum of the occupancy comprehensive threat value and the radar comprehensive threat value, so as to obtain the comprehensive threat value.

[0121] S1080. When the comprehensive threat value is greater than or equal to the preset first threat threshold and the comprehensive advantage value is less than the preset first advantage threshold, the confrontation decision is the first type of decision.

[0122] Among them, both the preset first threat threshold and the preset first advantage threshold are within the range of 0 - 1, and they are thresholds set by the user according to experience or requirements. The first type of decision is an escape decision. Exemplarily, the preset first threat threshold is 0.7 and the preset first advantage threshold is 0.6.

[0123] In this embodiment, when the comprehensive threat value is greater than or equal to the preset first threat threshold and the comprehensive advantage value is less than the preset first advantage threshold, it is determined that the corresponding confrontation decision at present is an escape decision. This escape decision can be selected from a preset tactical behavior library, and the preset tactical behaviors can include various tactical behaviors such as diving and sharp turning, circling and breaking maneuvers, small - radius loops, snake - shaped maneuvers, and bell - shaped maneuvers. Each tactical behavior corresponds to corresponding maneuver parameter values, and the maneuver parameter values can include, but are not limited to, angle values in the three directions of pitch, yaw, and roll, altitude value, and speed value, etc.

[0124] S1090. When the comprehensive threat value is greater than or equal to the preset second threat threshold, less than the preset first threat threshold, and the comprehensive advantage value is less than the preset second advantage threshold, the confrontation decision is the second type of decision.

[0125] Among them, the preset first threat threshold is greater than the preset second threat threshold, and both are within the range of 0 - 1. They are also values defined by the user according to experience or requirements. The preset first advantage threshold is greater than the preset second advantage threshold. Exemplarily, the preset second threat threshold is 0.5 and the preset second advantage threshold is 0.5.

[0126] In this embodiment, when the comprehensive threat value is greater than or equal to the preset second threat threshold, less than the preset first threat threshold, and the comprehensive advantage value is less than the preset second advantage threshold, the confrontation decision is the second type of decision, and this second type of decision is a defense decision, which can also be selected from the preset tactical behavior library.

[0127] S10100. When the comprehensive threat value is less than the preset third threat threshold and the comprehensive advantage value is less than the preset first advantage threshold, the confrontation decision is the third type of decision.

[0128] Among them, the preset second threat threshold is greater than the preset third threat threshold; the third type of decision is an attack decision. Exemplarily, the preset third threat threshold is 0.3.

[0129] In this embodiment, when the comprehensive threat value is less than the preset third threat threshold and the comprehensive advantage value is less than the preset first advantage threshold, the confrontation decision is the third type of decision, and it is also possible to select from the preset tactical behavior library.

[0130] S10110. Select the optimal confrontation decision from the first type of decision, or the second type of decision, or the third type of decision.

[0131] In this embodiment, the optimal confrontation decision can be selected from the first type of decision, or the second type of decision, or the third type of decision according to the maneuver parameters corresponding to the next maneuver sequence characteristics of the enemy aircraft. It can be understood that according to the angle value, altitude value, speed value, etc. corresponding to the next maneuver sequence characteristics of the enemy aircraft, one type in the major categories of attack, defense, and escape can be selected, and then according to the current situation, the optimal small classification action is determined after determining the major category. For example: Attack: Pull up at a large angle; Defense: Vertical snake; Escape: Circle and break away. It should be noted that in addition to the above major categories of attack, defense, and escape, there may also be cases of interference types. At this time, level flight is selected to execute the confrontation. Specifically, the height, speed, and spatial position moved within a unit time of the defined tactical maneuver action can be used as the selection basis. For example, the heights of both the bell-shaped maneuver and the zooming sharp turn are below 3000m. First, the altitude is raised to adjust the spatial altitude occupancy advantage. However, the bell-shaped maneuver is for defense, and the zooming sharp turn is for escape, which are different confrontation decisions.

[0132] The above technical solution of the embodiment of the present invention trains a tactical behavior prediction model through a tactical behavior sample set, determines the next maneuver sequence characteristics of the enemy aircraft according to the action parameter information of the enemy aircraft in the preset time series and the tactical behavior prediction model. On this basis, a situation analysis model is constructed based on the spatial occupancy of our aircraft and the enemy aircraft. Based on this situation analysis model, the occupancy comprehensive advantage value and occupancy comprehensive threat value of our aircraft relative to the enemy aircraft are determined, and the radar comprehensive advantage value and radar comprehensive threat value of the airborne radar of our aircraft relative to the airborne radar of the enemy aircraft are determined. Thus, the optimal confrontation decision is made for the next maneuver sequence characteristics of the enemy aircraft according to the occupancy comprehensive advantage value, occupancy comprehensive threat value, radar comprehensive advantage value, and radar comprehensive threat value, which can improve the ability of autonomous intelligent decision-making in a complex battlefield environment and effectively improve the efficiency and accuracy of autonomous intelligent decision-making.

[0133] In one embodiment, Figure 11The following is a structural block diagram of an intelligent autonomous decision-making device for an unmanned aerial vehicle provided by an embodiment of the present invention. This device is applicable to the situation when training a model for DC arc detection, and can be implemented by hardware / software. It can be configured in an electronic device to implement an intelligent autonomous decision-making method for an unmanned aerial vehicle in an embodiment of the present invention. As Figure 11 shown, the device includes: a data acquisition module 1110, an enemy aircraft maneuver sequence prediction module 1120, a situation analysis and establishment module 1130, an occupancy comprehensive value determination module 1140, a radar comprehensive value determination module 1150, and an adversarial decision-making determination module 1160.

[0134] Among them, the data acquisition module 1110 is used to obtain a pre-constructed tactical behavior sample set, and use the tactical behavior sample set to train a behavior prediction model to obtain a trained tactical behavior prediction model;

[0135] The enemy aircraft maneuver sequence prediction module 1120 is used to obtain the action parameter information of the enemy aircraft in a preset time series, and determine the next maneuver sequence feature of the enemy aircraft according to the action parameter information and the tactical behavior prediction model;

[0136] The situation analysis and establishment module 1130 is used to construct a situation analysis model based on the spatial occupancy of our aircraft and the enemy aircraft; among them, the situation analysis model includes: an angle situation analysis model, a distance situation analysis model, a speed situation analysis model, and an altitude situation analysis model;

[0137] The occupancy comprehensive value determination module 1140 is used to determine the occupancy comprehensive advantage value and occupancy comprehensive threat value of our aircraft relative to the enemy aircraft based on the situation analysis model;

[0138] The radar comprehensive value determination module 1150 is used to determine the radar comprehensive advantage value and radar comprehensive threat value of the airborne radar of our aircraft relative to the airborne radar of the enemy aircraft;

[0139] The adversarial decision-making determination module 1160 is used to make an optimal adversarial decision on the next maneuver sequence feature of the enemy aircraft according to the occupancy comprehensive advantage value, the occupancy comprehensive threat value, the radar comprehensive advantage value, and the radar comprehensive threat value.

[0140] In an embodiment of the present invention, an enemy aircraft maneuver sequence prediction module trains a tactical behavior prediction model through a tactical behavior sample set, and determines the next maneuver sequence feature of the enemy aircraft based on the action parameter information of the enemy aircraft in a preset time series and the tactical behavior prediction model. On this basis, a situation analysis establishment module constructs a situation analysis model based on the spatial occupancy of our aircraft and the enemy aircraft. An occupancy comprehensive value determination module determines the occupancy comprehensive advantage value and occupancy comprehensive threat value of our aircraft relative to the enemy aircraft based on this situation analysis model, and a radar comprehensive value determination module determines the radar comprehensive advantage value and radar comprehensive threat value of the airborne radar of our aircraft relative to the airborne radar of the enemy aircraft. Thus, an anti-aircraft decision determination module makes an optimal anti-aircraft decision on the next maneuver sequence feature of the enemy aircraft according to the occupancy comprehensive advantage value, occupancy comprehensive threat value, radar comprehensive advantage value and radar comprehensive threat value, which can improve the ability of autonomous intelligent decision-making in a complex battlefield environment and effectively improve the efficiency and accuracy of autonomous intelligent decision-making.

[0141] In one embodiment, the construction of the tactical behavior sample set includes:

[0142] Obtain the movement trajectory information of the unmanned aircraft at historical moments, and determine the basic actions of the unmanned aircraft according to the movement trajectory information;

[0143] Form maneuver actions from the basic actions, and form tactical behaviors based on the maneuver actions of at least two consecutive time series; wherein, the tactical behaviors represent tactical intentions;

[0144] Establish a continuous maneuver sequence of the unmanned aircraft with typical tactical behaviors based on a preset aircraft dynamics model, a preset flight envelope and the tactical intention to form a tactical behavior library, and use the tactical behavior library as the tactical behavior sample set;

[0145] Among them, the maneuver sequence features of the typical tactical behaviors at least include: no tactical maneuver, dive sharp turn, spiral escape maneuver, small-radius loop, snake maneuver, bell maneuver, Herbst maneuver, Immelmann maneuver, cobra maneuver, vertical snake maneuver and zoom sharp turn; each maneuver sequence feature corresponds to corresponding maneuver parameter values; the maneuver parameter values at least include: altitude value range, speed value range, pitch angle value, yaw angle value, roll angle value.

[0146] In one embodiment, the behavior prediction model is a recurrent neural network model; the recurrent neural network model is based on the MobileNet-V1 network architecture; the convolutional layer of the recurrent neural network model consists of a DWCONV+(1×1CONV) to form a depthwise separable convolution structure; the activation function of the recurrent neural network model is RELU;

[0147] Correspondingly, the data acquisition module 1110 includes:

[0148] A sample processing unit is used to perform zero-padding operations on the time series of each sample in the tactical behavior sample set, so as to make the time series lengths of all the samples consistent to obtain processed samples; wherein, the tactical behavior sample set includes maneuver sequence features of typical tactical behaviors and maneuver parameter values corresponding to each of the maneuver sequence features.

[0149] A sample enhancement unit is used to perform data enhancement on the processed samples to obtain data-enhanced samples.

[0150] A target sample set determination unit is used to insert a level flight action after the action of each sample for the data-enhanced samples, so as to initialize the next sample action and obtain a target sample set.

[0151] A model training unit is used to input the target sample set into a behavior prediction model to optimize model parameters until the model parameters reach the optimal or a predefined loss function reaches the minimum, so as to obtain a trained tactical behavior prediction model.

[0152] In one embodiment, the action parameter information at least includes altitude, speed, pitch angle, yaw angle, roll angle, longitude and latitude; correspondingly, the enemy aircraft maneuver sequence prediction module 1120 includes:

[0153] A prediction unit is used to perform data preprocessing on the action parameter information and input the preprocessed action parameter information into the tactical behavior prediction model to predict the next maneuver sequence feature of the enemy aircraft.

[0154] In one embodiment, the situation analysis establishment module 1130 includes:

[0155] A position determination unit is used to detect the enemy aircraft using the on-board radar of our aircraft to determine the spatial position of the enemy aircraft relative to our aircraft.

[0156] An attitude angle determination unit is used to determine the attitude angle of the enemy aircraft relative to the incident beam of the on-board radar of our aircraft at each moment according to the spatial position; wherein, the attitude angle includes: azimuth angle and pitch angle.

[0157] An RCS value determination unit is used to determine the RCS value of the enemy aircraft according to the attitude angle; wherein, the RCS value represents the RCS value when the enemy aircraft is exposed to the on-board radar of our aircraft at each moment.

[0158] A situation model construction unit is used to analyze the spatial situation according to the RCS value of the enemy aircraft to construct a situation analysis model.

[0159] In one embodiment, the situation analysis model is expressed by the formula: ; where, is represented as an angle situation analysis model, is represented as a speed situation analysis model, is represented as a distance situation analysis model, is represented as an altitude situation analysis model; 、 、 and respectively represent the weight coefficients corresponding to the angle situation analysis model, speed situation analysis model, distance situation analysis model, and altitude situation analysis model;

[0160] Among them, the angle situation analysis model includes: azimuth angle advantage and approach angle advantage , which are respectively expressed by the formulas: ; In the formula, represents the azimuth angle advantage; represents the approach angle advantage; 、 respectively represent the weights of the azimuth angle and approach angle, ; Among them,

[0161] , ; In the formula, represents the azimuth angle advantage; represents the approach angle advantage; represents the approach angle, represents the azimuth angle, represents the maximum search azimuth angle, represents the maximum off-axis launch angle, represents the non-escape cone angle, where, ;

[0162] The speed situation analysis model is expressed as: when , ; when , the speed situation analysis model is expressed as: ; In the formula, represents the speed of our aircraft; represents the speed of the enemy aircraft; represents the optimal air combat speed of our aircraft;

[0163] The distance situation analysis model is expressed as: , in the formula, represents the distance advantage; represents the distance between our aircraft and the enemy aircraft; represents the maximum detection distance of the radar, represents the maximum attack distance, Represents the maximum non - escapable distance, Represents the minimum non - escapable distance, Represents the minimum attack distance, where, ;

[0164] The height situation analysis model is expressed by the formula: ; In the formula, Represents height advantage; Represents the height of our aircraft; Represents the height of the enemy aircraft; Represents the optimal occupancy height of our aircraft.

[0165] In one embodiment, the occupancy comprehensive value determination module 1140 includes:

[0166] The occupancy threat value determination unit is used to inversely solve the first variable weight coefficients corresponding to each first index in the situation analysis model according to the preset CRITIC method, and determine the sum of the products of each first variable weight coefficient and the corresponding first index, so as to obtain the comprehensive occupancy threat value of our aircraft relative to the enemy aircraft; where, the first index includes: height, angle, distance and speed; where, the first variable weight coefficient is the occupancy threat coefficient;

[0167] The occupancy advantage value determination unit is used to positively solve the second variable weight coefficients corresponding to each first index in the situation analysis model according to the preset CRITIC method, and determine the sum of the products of each second variable weight coefficient and the corresponding first index, so as to obtain the comprehensive occupancy advantage value of our aircraft relative to the enemy aircraft; where, the second variable weight coefficient is the occupancy advantage coefficient.

[0168] In one embodiment, the radar comprehensive value determination module 1150 includes:

[0169] The signal - to - noise ratio determination unit is used to determine the signal - to - noise ratio according to the power received by the radar antenna of our aircraft's airborne radar and the generated noise power; where, the signal - to - noise ratio characterizes the radar monitoring ability;

[0170] The maximum detection distance determination unit is used to obtain the RCS value of the enemy aircraft, and determine the maximum detection distance of our aircraft's airborne radar according to the signal - to - noise ratio and the RCS value of the enemy aircraft;

[0171] A weight coefficient determination unit is configured to use the maximum detection range, the signal-to-noise ratio, and the RCS value as second indicators, and based on a preset digital combat simulator (DCS) software, simulate and calculate the first empirical value results and the second empirical value results respectively corresponding to each of the second indicators, use the first empirical value results as the corresponding first weight coefficients respectively, and use the second empirical value results as the corresponding second weight coefficients respectively; wherein, the first weight coefficient is the radar threat coefficient corresponding to each of the second indicators; the second weight coefficient is the radar advantage coefficient corresponding to each of the second indicators;

[0172] A radar comprehensive value determination unit is configured to determine the sum of the first products of each of the first weight coefficients and the corresponding second indicators to obtain the radar comprehensive advantage value of the airborne radar of our own aircraft relative to the airborne radar of the enemy aircraft, and determine the sum of the second products of each of the second weight coefficients and the corresponding second indicators to obtain the radar comprehensive threat value of the airborne radar of our own aircraft relative to the airborne radar of the enemy aircraft.

[0173] In one embodiment, the confrontation decision-making determination module 1160 includes:

[0174] A comprehensive value determination unit is configured to determine a comprehensive advantage value based on the position occupancy comprehensive advantage value and the radar comprehensive advantage value, and determine a comprehensive threat value based on the position occupancy comprehensive threat value and the radar comprehensive threat value;

[0175] A first decision-making unit is configured to, when the comprehensive threat value is greater than or equal to a preset first threat threshold and the comprehensive advantage value is less than a preset first advantage threshold, make the confrontation decision a first type of decision;

[0176] A second decision-making unit is configured to, when the comprehensive threat value is greater than or equal to a preset second threat threshold, less than the preset first threat threshold, and the comprehensive advantage value is less than a preset second advantage threshold, make the confrontation decision a second type of decision;

[0177] A third decision-making unit is configured to, when the comprehensive threat value is less than a preset third threat threshold and the comprehensive advantage value is less than a preset first advantage threshold, make the confrontation decision a third type of decision;

[0178] An optimal decision-making selection unit is configured to select an optimal confrontation decision from the first type of decision, or the second type of decision, or the third type of decision;

[0179] Wherein, the preset first threat threshold is greater than the preset second threat threshold; the preset second threat threshold is greater than the preset third threat threshold; the preset first advantage threshold is greater than the preset second advantage threshold;

[0180] Among them, the first type of decision is an escape decision; the second type of decision is a defense decision; the third type of decision is an attack decision.

[0181] The drone intelligent autonomous decision-making device provided by the embodiment of the present invention can execute the drone intelligent autonomous decision-making method provided by any embodiment of the present invention, and has function modules and beneficial effects corresponding to the execution of the method.

[0182] In one embodiment, Figure 12 FIG. is a schematic structural diagram of an electronic device provided by an embodiment of the present invention. The electronic device 10 is intended to represent various forms of digital computers, such as, for example, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as, for example, personal digital processors, cellular telephones, smart telephones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0183] As Figure 12 shown, the electronic device 10 includes at least one processor 11, and a memory communicatively connected to the at least one processor 11, such as ROM 12, RAM 13, etc. Among them, the memory stores a computer program executable by the at least one processor. The processor 11 may execute various appropriate actions and processes according to the computer program stored in the ROM 12 or the computer program loaded from the storage unit 18 into the RAM 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 may also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. The I / O interface 15 is also connected to the bus 14.

[0184] A plurality of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a magnetic disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0185] The processor 11 may be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as the intelligent autonomous decision-making method for the drone.

[0186] In some embodiments, the intelligent autonomous decision-making method for the drone may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed onto the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the intelligent autonomous decision-making method for the drone described above may be executed. Alternatively, in other embodiments, the processor 11 may be configured to execute the intelligent autonomous decision-making method for the drone in any other suitable manner (e.g., by means of firmware).

[0187] Various embodiments of the systems and techniques described above in this document may be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-chip (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include: being implemented in one or more computer programs, which may be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a special-purpose or general-purpose programmable processor, and may receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0188] The computer program for implementing the method of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to the processor of a general-purpose computer, a special-purpose computer, or other programmable intelligent autonomous decision-making devices for the drone, such that when the computer program is executed by the processor, the functions / operations specified in the flowchart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, executed partially on the machine as an independent software package and partially on a remote machine, or executed entirely on a remote machine or server.

[0189] In the context of the present invention, a computer-readable storage medium can be a tangible medium that can contain or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. The computer-readable storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, the computer-readable storage medium can be a machine-readable signal medium. More specific examples of the machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0190] To provide for interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the electronic device. Other kinds of devices can also be used to provide for interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0191] The systems and techniques described herein can be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0192] A computing system may include a client and a server. The client and the server are generally far from each other and usually interact via a communication network. The client-server relationship is created by computer programs running on respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system, and solves the defects of difficult management and weak business scalability existing in traditional physical hosts and VPS services.

[0193] It should be understood that various forms of the processes shown above can be used, steps can be reordered, added or deleted. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and no limitations are imposed herein.

[0194] The above specific embodiments do not constitute a limitation on the protection scope of the present invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An intelligent autonomous decision-making method for an unmanned aerial vehicle, characterized in that, The method includes: Obtaining a pre-constructed set of tactical behavior samples, and using the set of tactical behavior samples to train a behavior prediction model to obtain a trained tactical behavior prediction model; Obtaining the action parameter information of an enemy aircraft in a preset time series, and determining the next maneuver sequence feature of the enemy aircraft according to the action parameter information and the tactical behavior prediction model; Constructing a situation analysis model based on the spatial positions of our aircraft and the enemy aircraft; wherein, the situation analysis model includes: an angle situation analysis model, a distance situation analysis model, a speed situation analysis model, and an altitude situation analysis model; Determining the comprehensive occupancy advantage value and the comprehensive occupancy threat value of our aircraft relative to the enemy aircraft based on the situation analysis model; Determining the comprehensive radar advantage value and the comprehensive radar threat value of the airborne radar of our aircraft relative to the airborne radar of the enemy aircraft; Making an optimal countermeasure decision on the next maneuver sequence feature of the enemy aircraft according to the comprehensive occupancy advantage value, the comprehensive occupancy threat value, the comprehensive radar advantage value, and the comprehensive radar threat value; 2. The method according to claim 1, wherein The construction of the set of tactical behavior samples includes: Obtaining the motion trajectory information of a drone at a historical moment, and determining the basic actions of the drone according to the motion trajectory information; Forming maneuver actions from the basic actions, and forming tactical behaviors based on the maneuver actions in at least two consecutive time series; wherein, the tactical behaviors represent tactical intentions; Establishing a continuous maneuver sequence of a drone for typical tactical behaviors based on a preset aircraft dynamics model, a preset flight envelope, and the tactical intention to form a tactical behavior library, and using the tactical behavior library as the set of tactical behavior samples; Wherein, the maneuver sequence features of the typical tactical behaviors at least include: no tactical maneuver, diving sharp turn, circling escape maneuver, small-radius loop, snake maneuver, bell maneuver, Herbst maneuver, Immelmann maneuver, cobra maneuver, vertical snake maneuver, and climbing sharp turn; each maneuver sequence feature corresponds to a corresponding maneuver parameter value; the maneuver parameter values at least include: altitude value range, speed value range, pitch angle value, yaw angle value, roll angle value.

3. The method according to claim 1, characterized in that, The behavior prediction model is a recurrent neural network model; the recurrent neural network model is based on the MobileNet-V1 network architecture; the convolutional layer of the recurrent neural network model consists of a DWCONV+(1×1CONV) to form a depthwise separable convolution structure; the activation function of the recurrent neural network model is RELU; Correspondingly, the using the set of tactical behavior samples to train a behavior prediction model to obtain a trained tactical behavior prediction model includes: Performing zero-padding operations on the time series of each sample in the set of tactical behavior samples to make the time series lengths of the samples consistent to obtain processed samples; wherein, the set of tactical behavior samples includes the maneuver sequence features of typical tactical behaviors and the maneuver parameter values corresponding to the maneuver sequence features; Performing data augmentation on the processed samples to obtain data-augmented samples; For the samples after data augmentation, a level flight action is inserted after the action of each sample to initialize the next sample action, thereby obtaining a target sample set; The target sample set is input into the behavior prediction model to optimize the model parameters until the model parameters reach the optimal or a predefined loss function reaches the minimum, thereby obtaining a trained tactical behavior prediction model.

4. The method according to claim 1, wherein The action parameter information at least includes altitude, speed, pitch angle, yaw angle, roll angle, longitude and latitude; Correspondingly, determining the next maneuver sequence feature of the enemy aircraft according to the action parameter information and the tactical behavior prediction model includes: Performing data preprocessing on the action parameter information, and inputting the preprocessed action parameter information into the tactical behavior prediction model to predict the next maneuver sequence feature of the enemy aircraft.

5. The method according to claim 1, wherein Constructing a situation analysis model based on the spatial occupancy of the enemy aircraft and our aircraft includes: Using the airborne radar of our aircraft to detect the enemy aircraft to determine the spatial position of the enemy aircraft relative to our aircraft; Determining the attitude angle of the enemy aircraft relative to the incident beam of the airborne radar of our aircraft at each moment according to the spatial position; wherein, the attitude angle includes: azimuth angle and pitch angle; Determining the radar cross section (RCS) value of the enemy aircraft according to the attitude angle; wherein, the RCS value represents the RCS value when the enemy aircraft is exposed to the airborne radar of our aircraft at each moment; Analyzing the spatial situation according to the RCS value of the enemy aircraft to construct a situation analysis model.

6. The method according to any one of claims 1 or 5, characterized in that The situation analysis model is expressed by the formula as follows: ; where represents the angle situation analysis model, represents the speed situation analysis model, represents the distance situation analysis model, represents the altitude situation analysis model; , , and respectively represent the weight coefficients corresponding to the angle situation analysis model, the speed situation analysis model, the distance situation analysis model and the altitude situation analysis model; Among them, the angle situation analysis model includes: azimuth angle advantage and approach angle advantage , which are respectively expressed by the formulas: ; In the formula, represents the azimuth angle advantage; represents the approach angle advantage; , respectively represent the weights of the azimuth angle and the approach angle, ; Among them, , ; In the formula, represents the azimuth angle advantage; represents the entry angle advantage; represents the entry angle, represents the azimuth angle, represents the maximum search azimuth angle, represents the maximum off-axis launch angle, represents the non-escape cone angle, where ; The speed situation analysis model is expressed as: when then ; when the speed situation analysis model is expressed as: ; where represents the speed of our aircraft; represents the speed of the enemy aircraft; represents the optimal air combat speed of our aircraft; The distance situation analysis model is expressed as: , where represents the distance advantage, represents the distance between our aircraft and the enemy aircraft; represents the maximum detection distance of the radar, represents the maximum attack distance, represents the maximum non-escape distance, represents the minimum non-escape distance, represents the minimum attack distance, where ; The height situation analysis model is expressed by the formula: ; In the formula, represents the height advantage; represents the height of our aircraft; represents the height of the enemy aircraft; represents the optimal occupancy height of our aircraft.

7. The method according to claim 1, wherein Determining the overall occupancy advantage value and overall occupancy threat value of our aircraft relative to the enemy aircraft based on the situation analysis model includes: Performing reverse solution on the first variable weight coefficients corresponding to the first indicators in the situation analysis model according to the preset CRITIC method, and determining the sum of the products of each first variable weight coefficient and the corresponding first indicator, so as to obtain the overall occupancy threat value of our aircraft relative to the enemy aircraft; wherein, the first indicators include: altitude, angle, distance and speed; wherein, the first variable weight coefficient is the occupancy threat coefficient; Performing forward solution on the second variable weight coefficients corresponding to the first indicators in the situation analysis model according to the preset CRITIC method, and determining the sum of the products of each second variable weight coefficient and the corresponding first indicator, so as to obtain the overall occupancy advantage value of our aircraft relative to the enemy aircraft; wherein, the second variable weight coefficient is the occupancy advantage coefficient.

8. The method according to claim 1, wherein Determining the overall radar advantage value and overall radar threat value of the airborne radar of our aircraft relative to the airborne radar of the enemy aircraft includes: Determining the signal-to-noise ratio according to the power received by the radar antenna of the airborne radar of our aircraft and the generated noise power; wherein, the signal-to-noise ratio represents the radar monitoring ability; Obtaining the RCS value of the enemy aircraft, and determining the maximum detection distance of the airborne radar of our aircraft according to the signal-to-noise ratio and the RCS value of the enemy aircraft. Taking the maximum detection range, the signal-to-noise ratio, and the RCS value as the second indicators, and based on the preset digital combat simulator DCS software, simulating and calculating the first empirical value results and the second empirical value results corresponding to each of the second indicators respectively, taking the first empirical value results as the corresponding first weight coefficients respectively, and taking the second empirical value results as the corresponding second weight coefficients respectively; wherein, the first weight coefficient is the radar threat coefficient corresponding to each of the second indicators; the second weight coefficient is the radar advantage coefficient corresponding to each of the second indicators; Determining the sum of the first products of each of the first weight coefficients and the corresponding second indicators to obtain the radar comprehensive advantage value of the airborne radar of our aircraft relative to the airborne radar of the enemy aircraft, and determining the sum of the second products of each of the second weight coefficients and the corresponding second indicators to obtain the radar comprehensive threat value of the airborne radar of our aircraft relative to the airborne radar of the enemy aircraft.

9. The method according to claim 1, wherein The making of the optimal countermeasure decision for the next maneuver sequence feature of the enemy aircraft according to the occupancy comprehensive advantage value, the occupancy comprehensive threat value, the radar comprehensive advantage value, and the radar comprehensive threat value includes: Determining the comprehensive advantage value according to the occupancy comprehensive advantage value and the radar comprehensive advantage value, and determining the comprehensive threat value according to the occupancy comprehensive threat value and the radar comprehensive threat value; In the case where the comprehensive threat value is greater than or equal to the preset first threat threshold and the comprehensive advantage value is less than the preset first advantage threshold, the countermeasure decision is the first type of decision; In the case where the comprehensive threat value is greater than or equal to the preset second threat threshold, less than the preset first threat threshold, and the comprehensive advantage value is less than the preset second advantage threshold, the countermeasure decision is the second type of decision; In the case where the comprehensive threat value is less than the preset third threat threshold and the comprehensive advantage value is less than the preset first advantage threshold, the countermeasure decision is the third type of decision; Selecting the optimal countermeasure decision from the first type of decision, or the second type of decision, or the third type of decision; Wherein, the preset first threat threshold is greater than the preset second threat threshold; the preset second threat threshold is greater than the preset third threat threshold; the preset first advantage threshold is greater than the preset second advantage threshold; Wherein, the first type of decision is the escape decision; the second type of decision is the defense decision; the third type of decision is the attack decision.

10. An intelligent autonomous decision-making device for an unmanned aerial vehicle, characterized in that, The device includes: A data acquisition module, configured to acquire a pre-constructed tactical behavior sample set and use the tactical behavior sample set to train a behavior prediction model to obtain a trained tactical behavior prediction model; An enemy aircraft maneuver sequence prediction module, configured to acquire the action parameter information of the enemy aircraft in a preset time series and determine the next maneuver sequence feature of the enemy aircraft according to the action parameter information and the tactical behavior prediction model; A situation analysis establishment module, configured to construct a situation analysis model based on the spatial positions of our aircraft and the enemy aircraft; wherein, the situation analysis model includes: an angle situation analysis model, a distance situation analysis model, a speed situation analysis model, and an altitude situation analysis model; An occupancy comprehensive value determination module, configured to determine the occupancy comprehensive advantage value and the occupancy comprehensive threat value of our aircraft relative to the enemy aircraft based on the situation analysis model; A radar comprehensive value determination module, configured to determine the radar comprehensive advantage value and the radar comprehensive threat value of the airborne radar of our aircraft relative to the airborne radar of the enemy aircraft; An adversarial decision determination module, configured to make an optimal adversarial decision on the next maneuver sequence feature of the enemy aircraft according to the occupancy comprehensive advantage value, the occupancy comprehensive threat value, the radar comprehensive advantage value, and the radar comprehensive threat value; 11. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the unmanned aerial vehicle intelligent autonomous decision-making method according to any one of claims 1-9.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, and the computer instructions are used to implement the unmanned aerial vehicle intelligent autonomous decision-making method according to any one of claims 1-9 when executed by a processor.

13. A computer program product, characterized in that, The computer program product includes a computer program, and the computer program implements the unmanned aerial vehicle intelligent autonomous decision-making method according to any one of claims 1-9 when executed by a processor.

Images