Six-degree-of-freedom unmanned combat aircraft close combat method and system based on machine game

Abstract

The present application belongs to the technical field of close combat of unmanned combat aircraft, and discloses a six-degree-of-freedom unmanned combat aircraft close combat method and system based on machine game, which comprises an environment perception module, a main control module, a wireless communication module, a combat target identification module, a combat target tracking module, a machine game decision module, a shooting module, a fault prediction module and a display module. The combat target tracking module has high calculation speed and fast target tracking speed. Meanwhile, the fault prediction module fully considers the individual differences of different six-degree-of-freedom unmanned combat aircraft engines in terms of running environment, service life, factory setting and wear degree, trains a plurality of SOM models of the training six-degree-of-freedom unmanned combat aircraft engines respectively, so as to obtain the accurate SOM model training results and corresponding minimum quantization error MQE of the corresponding six-degree-of-freedom unmanned combat aircraft engines, and the fault prediction accuracy is high.

Description

Technical Field

[0001] This invention belongs to the field of close-range combat technology for unmanned combat aircraft, and particularly relates to a six-degree-of-freedom unmanned combat aircraft close-range combat method and system based on machine game theory. Background Technology

[0002] Unmanned Combat Aircraft (UCAV) is a novel aerial weapon system. Previously, unmanned combat aircraft primarily served as combat support equipment for tasks such as aerial reconnaissance, battlefield surveillance, and combat damage assessment. They have evolved into major combat equipment capable of suppressing enemy air defenses, attacking ground targets, and conducting air-to-air combat. Currently, their main functions are suppressing air defenses and conducting deep strikes. With the development of real-time battlefield information networks and artificial intelligence technologies, these technical challenges that have hindered the development of UCAVs have been resolved. Currently, UCAVs have become a hot topic in the development of unmanned aerial vehicle (UAV) technology; however, existing UCAVs in close-range combat systems often cannot accurately determine the attack priority of multiple hostile targets. This may lead to the fighter jet failing to prioritize attacks on the most critical or threatening targets during missions, thus reducing combat efficiency. Simultaneously, existing systems do not fully consider the individual differences in the operating environment and service life of different six-degree-of-freedom UCAV engines, which may result in lower accuracy in predicting machine failures, thus affecting the reliability and service life of the fighter jet.

[0003] Based on the above analysis, the problems and shortcomings of the existing technology are as follows:

[0004] (1) Existing close-range combat systems for unmanned combat aircraft cannot accurately detect and determine the attack priority of multiple hostile targets.

[0005] (2) The individual differences in the operating environment and service life of different six-degree-of-freedom unmanned combat aircraft engines were not taken into account, resulting in low accuracy of fault prediction results. Summary of the Invention

[0006] To address the problems existing in the prior art, this invention provides a six-degree-of-freedom unmanned combat aerial vehicle close-range combat method and system based on machine game theory.

[0007] This invention is implemented as follows: a six-degree-of-freedom unmanned combat aerial vehicle close-range combat system based on machine game theory, comprising:

[0008] Environmental perception module, main control module, wireless communication module, combat target recognition module, combat target tracking module, machine game decision-making module, shooting module, fault prediction module, and display module;

[0009] Environmental perception module: Connected to the main control module, it uses cameras, lidar and millimeter-wave radar to monitor combat targets and acquire environmental information;

[0010] Main control module: Connects to all modules and is used to control the normal operation of each module;

[0011] The wireless communication module, connected to the main control module, is used to connect to a wireless network via a wireless chip for wireless communication.

[0012] The combat target recognition module is connected to the main control module and the enhanced environmental perception module. It uses deep learning algorithms to identify combat targets, thereby improving the accuracy and efficiency of target recognition.

[0013] The combat target tracking module is connected to the main control module and the deep learning-optimized target recognition module, and is used to track combat targets;

[0014] The machine game decision-making module, connected to the main control module, is used to determine the priority of attack on combat targets;

[0015] The firing module, connected to the main control module, is used to fire at combat targets;

[0016] The fault prediction module, connected to the main control module, is used to predict generator faults in six-degree-of-freedom unmanned combat aircraft.

[0017] The display module, connected to the main control module, is used to display monitoring videos, combat targets, and prediction results.

[0018] Furthermore, the environmental perception module uses cameras, lidar, and millimeter-wave radar to perceive the environment, and the collected information is integrated and utilized by the main control module for target identification, tracking, and shooting.

[0019] The camera captures visible light to obtain color and texture information of the environment, and transforms it into useful information through image processing and computer vision algorithms;

[0020] The lidar emits a laser beam and measures the time it takes for the reflected light to travel to obtain the target distance. For a large number of laser beams, a high-precision three-dimensional point cloud can be obtained, providing accurate spatial information about the environment. The point cloud is in the form of P = {p_1, p_2, ..., p_n}, where each p_i = (x_i, y_i, z_i) represents a point in the point cloud.

[0021] Millimeter-wave radar detects targets by emitting high-frequency electromagnetic waves and receiving the reflected waves. It can work effectively in various weather and lighting conditions and is used for the detection and tracking of moving targets.

[0022] The main control module fuses the information acquired by these sensors, typically using a Kalman filter as the sensor fusion technology, to gain a comprehensive and accurate understanding of the environment, thereby optimizing identification, tracking, and firing tasks.

[0023] Furthermore, a close-range combat method for six-DOF unmanned combat aircraft based on machine game theory includes the following steps:

[0024] Step 1: Monitor the combat target using video surveillance via the environmental perception module;

[0025] Step two: The main control module uses a wireless chip to connect to the wireless network for wireless communication via the wireless communication module;

[0026] Step 3: Identify the combat target using the combat target identification module; track the combat target using the combat target tracking module;

[0027] Step four: Using the machine game module, a priority list of current combat targets is generated based on the current situation of the unmanned combat aircraft.

[0028] Step 5: Based on the attack priority list, fire at the combat targets using the firing module; predict generator failures of the six-degree-of-freedom unmanned combat aircraft using the fault prediction module;

[0029] Step six: Display the monitoring video, combat target, and prediction results through the display module.

[0030] Furthermore, the tracking method of the combat target tracking module is as follows:

[0031] (1) Configure the camera parameters of the six-degree-of-freedom unmanned combat aircraft, continuously capture at least two images of the combat target using the camera on the six-degree-of-freedom unmanned combat aircraft, and perform noise reduction processing on the target images; and mark the ground coordinates, heading angle and time of acquisition of the combat target image of the six-degree-of-freedom unmanned combat aircraft in each image; and perform coordinate transformation on each combat target image so that they are in the same image coordinate system;

[0032] (2) Filter multiple fighting target images, then extract the foreground image including the target image from the filtered fighting target image, extract feature points from the foreground image, extract the feature points of each region in the previous foreground image and match them with the feature points of the template of the tracked target, update the template of the tracked target using the successfully matched region, and match the feature points of the updated template of the tracked target with the feature points of each region in the current foreground image.

[0033] (3) Based on the combat target situation generated in (1) and (2), and combined with the terrain obstacle module information obtained by the main control module, a list of targets can be formed, and the current UAV game situation is further modeled based on the extended game model.

[0034] (4) Determine the speed and direction of the tracked target based on the positional relationship of the feature points in the regions of the two successfully matched foreground images; adjust the speed and heading of the six-degree-of-freedom unmanned combat aircraft based on the speed and direction of the tracked target, thereby tracking the target.

[0035] Furthermore, based on a method combining Monte Carlo Tree Search (MCTS) and Deep Reinforcement Learning (MAT), future state projections and benefit assessments for different strike priority combinations are generated; based on the benefit assessment scores of the strike list combinations, a strike priority list is generated.

[0036] Furthermore, the affine transformation is used to transform each image to the same image coordinate system.

[0037] Furthermore, the fault prediction module uses the following prediction method:

[0038] 1) Configure the engine parameters of the six-degree-of-freedom unmanned combat aircraft, monitor the engine of the six-degree-of-freedom unmanned combat aircraft through monitoring equipment, obtain the engine operating parameters of the six-degree-of-freedom unmanned combat aircraft, and obtain the full life cycle data of multiple training six-degree-of-freedom unmanned combat aircraft engines; the full life cycle data is divided into normal data and abnormal data according to the usage time; the abnormal data includes fault warning data within the warning interval of the full life cycle;

[0039] 2) Train the corresponding SOM model of the six-degree-of-freedom unmanned combat aircraft engine using normal data from each of the training six-degree-of-freedom unmanned combat aircraft engines, and obtain the minimum quantization error (MQE) of the fault warning data in the six-degree-of-freedom unmanned combat aircraft engine; determine the minimum quantization error (MQE) range corresponding to the warning interval based on the average value of the minimum quantization error (MQE) of the fault warning data in the multiple training six-degree-of-freedom unmanned combat aircraft engines.

[0040] 3) Test the early warning accuracy of the minimum quantization error (MQE) range using multiple test six-degree-of-freedom unmanned combat aircraft engines respectively. Take the proportion of test six-degree-of-freedom unmanned combat aircraft engines whose accuracy exceeds the accuracy threshold as the test pass rate. If the test pass rate exceeds the first test pass rate threshold, then perform fault prediction of the six-degree-of-freedom unmanned combat aircraft engine to be predicted based on the minimum quantization error (MQE) range.

[0041] Furthermore, each of the tested six-DOF unmanned combat aircraft engines tests the accuracy of the Minimum Quantization Error (MQE) range by performing the following steps:

[0042] The SOM model of the test six-degree-of-freedom unmanned combat aircraft engine is trained using normal data from the test six-degree-of-freedom unmanned combat aircraft engine, and the minimum quantization error (MQE) of the abnormal data in the test six-degree-of-freedom unmanned combat aircraft engine is obtained.

[0043] The minimum quantization error (MQE) of the abnormal data is obtained, and the abnormal data range within the range of the minimum quantization error (MQE) is obtained.

[0044] The accuracy rate is obtained based on the ratio of the intersection of the abnormal data interval and the warning interval to the duration of the warning interval.

[0045] Furthermore, the method also includes: if the test pass rate is between the second test pass rate threshold and the first test pass rate threshold, then adjust the SOM model parameters of the test six-degree-of-freedom unmanned combat aircraft engine with an accuracy rate lower than the accuracy rate threshold or the proportion of normal data in the test six-degree-of-freedom unmanned combat aircraft engine, and retrain the SOM model of the test six-degree-of-freedom unmanned combat aircraft engine, and reacquire the accuracy rate and test pass rate;

[0046] Wherein, the second test pass rate threshold is lower than the first test pass rate threshold.

[0047] Furthermore, each of the above includes: if the test pass rate is lower than the second test pass rate threshold, then adjust one or more of the following parameters: the SOM model parameters of the training six-degree-of-freedom unmanned combat aircraft engine, the proportion of normal data in the full life cycle data, and the range of the warning interval, and redetermine the minimum quantization error (MQE) range, and reacquire the accuracy and test pass rate.

[0048] Furthermore, adjusting the SOM model parameters for training and testing the six-degree-of-freedom unmanned combat aircraft engine includes adjusting one or more parameters in the SOM model, such as the learning rate initialization parameter, the initial winning neighborhood, and the initial neuron weights.

[0049] Furthermore, the fault prediction of the six-degree-of-freedom unmanned combat aircraft engine based on the minimum quantization error (MQE) range includes:

[0050] For each normal data point collected for a six-degree-of-freedom unmanned combat aircraft engine to be predicted, the collected normal data is used to train the SOM model of the six-degree-of-freedom unmanned combat aircraft engine to be predicted.

[0051] For each abnormal data point collected from a six-DOF unmanned combat aircraft engine to be predicted, the minimum quantization error (MQE) of the abnormal data in the six-DOF unmanned combat aircraft engine to be predicted is obtained.

[0052] If the minimum quantization error (MQE) of the abnormal data in the six-degree-of-freedom unmanned combat aircraft engine to be predicted is within the range of the minimum quantization error (MQE), then the fault prediction of the six-degree-of-freedom unmanned combat aircraft engine to be predicted is performed.

[0053] Based on the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solution to be protected by this invention are as follows:

[0054] This invention brings a series of advantages and positive effects to the close-range combat system of unmanned combat aircraft through an enhanced environmental perception module and a target recognition and tracking module optimized by deep learning:

[0055] 1) By integrating multiple sensors such as cameras, lidar, infrared sensors, and millimeter-wave radar, UAVs can acquire more comprehensive and accurate environmental information, including but not limited to the target's position, size, shape, speed, and direction, as well as environmental information such as lighting and temperature. This enhanced environmental awareness enables UAVs to conduct effective operations in various complex environments.

[0056] 2) Deep learning-optimized target recognition and tracking modules can significantly improve the accuracy of UAV target recognition and tracking, reducing the possibility of misjudgments and missed detections. This is crucial for UAVs to quickly and accurately identify and track targets in battlefield environments.

[0057] 3) Unmanned combat aircraft can more accurately lock onto and engage targets, thereby improving combat effectiveness. This not only enhances the battlefield survivability of drones but also reduces the risk of friendly fire or harm to non-combat personnel.

[0058] 4) This invention can adapt to more types and more complex battlefield environments, including but not limited to low-light environments, fog and haze environments, and high-speed moving targets. This gives drones greater operational space and possibilities.

[0059] 5) Through deep learning technology, drones can improve their perception, understanding and decision-making abilities regarding the environment through continuous training and learning, thereby enhancing their ability to complete tasks independently without human command.

[0060] This invention features a high-speed target tracking module for close-range combat, enabling rapid target selection and accurate engagement. Based on an extended game theory model, it utilizes a combination of tree search and reinforcement learning to quickly prioritize multiple targets within a swarm configuration. This approach enhances the collaborative capabilities of UAV swarms in close-quarters combat, scientifically and rationally allocating target selection among multiple UAVs facing multiple adversaries, significantly improving the combat effectiveness of the UAV swarm. Simultaneously, the fault prediction module fully considers the individual differences in operating environment, lifespan, factory settings, and wear levels of various six-degree-of-freedom (6DOF) UAV engines. Multiple SOM (System-of-Mean-Operations) models for different 6DOF UAV engines are trained to obtain accurate SOM model training results and corresponding minimum quantization error (MQE) for each engine. The MQE range corresponding to the warning interval can be determined by calculating the average MQE of the fault warning data from multiple trained 6DOF UAV engines. This method maximizes the integration of individual differences among different 6DOF UAV engines and can be used for fault warning of the same type of 6DOF UAV engine, resulting in high fault prediction accuracy.

[0061] The machine game decision-making module in this invention utilizes techniques such as Monte Carlo tree search and deep reinforcement learning to generate different combinations of attack priorities based on the situational state of the unmanned combat aircraft, and performs future state projection and benefit evaluation to optimize target selection. Considering the individual differences in the engines of different unmanned combat aircraft, this module combines the results of the fault prediction module to achieve adaptive decision-making, avoid fault occurrence, and improve stability and reliability. Through real-time identification, tracking, and attack decision-making, combined with comprehensive analysis of information from multiple modules, this module can accurately detect and determine multiple hostile targets and make corresponding decisions in a timely manner. Attached Figure Description

[0062] Figure 1 This is a flowchart of a close-range combat method for a six-degree-of-freedom unmanned combat aircraft based on machine game theory, provided in an embodiment of the present invention.

[0063] Figure 2 This is a structural block diagram of a six-degree-of-freedom unmanned combat aerial vehicle close-range combat system based on machine game theory provided in an embodiment of the present invention;

[0064] Figure 3 This is a flowchart of the combat target tracking module tracking method provided in an embodiment of the present invention;

[0065] Figure 4 This is a flowchart of the fault prediction module prediction method provided in an embodiment of the present invention.

[0066] Figure 2The components are: 1. Environmental perception module; 2. Main control module; 3. Wireless communication module; 4. Combat target recognition module; 5. Combat target tracking module; 6. Shooting module; 7. Fault prediction module; 8. Display module. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0068] like Figure 1 As shown, the six-degree-of-freedom unmanned combat aerial vehicle close-range combat method based on machine game theory provided by this invention includes the following steps:

[0069] S101 uses an environmental perception module to monitor combat targets via video.

[0070] S102, the main control module uses a wireless chip to connect to a wireless network for wireless communication via a wireless communication module;

[0071] S103 identifies combat targets through a combat target recognition module and tracks combat targets through a combat target tracking module;

[0072] S104, through the machine game module, forms a list of attack priorities for the current combat targets based on the current situation of the unmanned combat aircraft;

[0073] S105, based on the strike priority list, fires at combat targets via the firing module; and predicts generator failures in the six-degree-of-freedom unmanned combat aircraft via the fault prediction module.

[0074] S106 displays monitoring video, combat targets, and prediction results through a display module.

[0075] like Figure 2 As shown, the six-degree-of-freedom unmanned combat aerial vehicle close-range combat system based on machine game theory provided in this embodiment of the invention includes:

[0076] Modules include: 1. Environmental perception module; 2. Main control module; 3. Wireless communication module; 4. Melee target recognition module; 5. Melee target tracking module; 6. Machine game decision-making module; 7. Shooting module; 8. Fault prediction module; and 9. Display module.

[0077] Environmental perception module 1, connected to main control module 2, is used to monitor the combat target via camera.

[0078] The main control module 2 is connected to the environmental perception module 1, wireless communication module 3, combat target recognition module 4, combat target tracking module 5, machine game decision-making module 6, shooting module 7, fault prediction module 8, and display module 9, and is used to control the normal operation of each module.

[0079] The wireless communication module 3 is connected to the main control module 2 and is used to connect to a wireless network via a wireless chip for wireless communication.

[0080] The combat target identification module 4 is connected to the main control module 2 and is used to identify combat targets;

[0081] The combat target tracking module 5 is connected to the main control module 2 and is used to track combat targets;

[0082] Machine game decision-making module 6, connected to the main control module, is used to determine the priority of attack on combat targets;

[0083] The firing module 7, connected to the main control module 2, is used to fire at combat targets;

[0084] Fault prediction module 8, connected to main control module 2, is used to predict generator faults in six-degree-of-freedom unmanned combat aircraft;

[0085] Display module 9, connected to main control module 2, is used to display monitoring video, combat targets, and prediction results.

[0086] like Figure 3 As shown, the tracking method of the combat target tracking module 5 provided by the present invention is as follows:

[0087] S201 configures the camera parameters of a six-degree-of-freedom unmanned combat aircraft, continuously captures at least two images of the target during combat using the onboard camera of the six-degree-of-freedom unmanned combat aircraft, performs noise reduction processing on the target images, and marks the ground coordinates, heading angle, and time of acquisition of the target image of the six-degree-of-freedom unmanned combat aircraft in each image; and performs coordinate transformation on each target image to make them all in the same image coordinate system.

[0088] S202, filter multiple fighting target images, then extract the foreground image including the target image from the filtered fighting target image, extract feature points from the foreground image, extract the feature points of each region in the previous foreground image and match them with the feature points of the template of the tracked target, update the template of the tracked target using the successfully matched regions, and match the feature points of the updated template of the tracked target with the feature points of each region in the current foreground image.

[0089] S203, based on the combat target situation generated by (1)(2), combined with the terrain obstacle module information obtained by the main control module, a list of targets to be attacked is formed, and the current UAV game situation is further modeled based on the extended game model;

[0090] S204: Determine the speed and direction of movement of the tracked target based on the positional relationship of feature points in the regions of the two successfully matched foreground images; adjust the speed and heading of the six-degree-of-freedom unmanned combat aircraft based on the speed and direction of movement of the tracked target, thereby tracking the target.

[0091] Furthermore, based on a method combining Monte Carlo Tree Search (MCTS) and Deep Reinforcement Learning (MAT), future state projections and benefit assessments for different strike priority combinations are generated; based on the benefit assessment scores of the strike list combinations, a strike priority list is generated.

[0092] The present invention provides a method for transforming each image to the same image coordinate system using affine transformation.

[0093] like Figure 4 As shown, the fault prediction module prediction method provided by this invention is as follows:

[0094] S301 configures the parameters of the six-degree-of-freedom unmanned combat aircraft engine, monitors the engine through monitoring equipment, obtains the engine's operating parameters, and acquires the full life cycle data of multiple training six-degree-of-freedom unmanned combat aircraft engines; the full life cycle data is divided into normal data and abnormal data according to usage duration; the abnormal data includes fault warning data within the warning interval of the full life cycle.

[0095] S302, using normal data from each of the trained six-degree-of-freedom unmanned combat aircraft engines, train the corresponding SOM model of the six-degree-of-freedom unmanned combat aircraft engine, and obtain the minimum quantization error (MQE) of the fault warning data in the six-degree-of-freedom unmanned combat aircraft engine; based on the average value of the minimum quantization error (MQE) of the fault warning data in the multiple trained six-degree-of-freedom unmanned combat aircraft engines, determine the minimum quantization error (MQE) range corresponding to the warning interval.

[0096] S303, using multiple test six-degree-of-freedom unmanned combat aircraft engines, respectively test the early warning accuracy within the minimum quantization error (MQE) range, and take the proportion of test six-degree-of-freedom unmanned combat aircraft engines whose accuracy exceeds the accuracy threshold as the test pass rate. If the test pass rate exceeds the first test pass rate threshold, then based on the minimum quantization error (MQE) range, perform fault prediction on the six-degree-of-freedom unmanned combat aircraft engine to be predicted.

[0097] The present invention provides a test for the accuracy of each six-DOF unmanned combat aircraft engine within the Minimum Quantization Error (MQE) range by performing the following steps:

[0098] The SOM model of the test six-degree-of-freedom unmanned combat aircraft engine is trained using normal data from the test six-degree-of-freedom unmanned combat aircraft engine, and the minimum quantization error (MQE) of the abnormal data in the test six-degree-of-freedom unmanned combat aircraft engine is obtained.

[0099] The minimum quantization error (MQE) of the abnormal data is obtained, and the abnormal data range within the range of the minimum quantization error (MQE) is obtained.

[0100] The accuracy rate is obtained based on the ratio of the intersection of the abnormal data interval and the warning interval to the duration of the warning interval.

[0101] The method provided by the present invention further includes: if the test pass rate is between the second test pass rate threshold and the first test pass rate threshold, then adjust the SOM model parameters of the test six-degree-of-freedom unmanned combat aircraft engine with an accuracy rate lower than the accuracy rate threshold or the proportion of normal data in the test six-degree-of-freedom unmanned combat aircraft engine, and retrain the SOM model of the test six-degree-of-freedom unmanned combat aircraft engine, and reacquire the accuracy rate and test pass rate;

[0102] Wherein, the second test pass rate threshold is lower than the first test pass rate threshold.

[0103] The present invention also includes: if the test pass rate is lower than the second test pass rate threshold, adjusting one or more of the following parameters: the SOM model parameters of the training six-degree-of-freedom unmanned combat aircraft engine, the proportion of normal data in the full life cycle data, and the range of the warning interval, and redetermining the minimum quantization error (MQE) range, and re-acquiring the accuracy and test pass rate.

[0104] The present invention provides for adjusting the SOM model parameters of the training and testing six-degree-of-freedom unmanned combat aircraft engine, including adjusting one or more parameters in the SOM model such as the learning rate initialization parameter, the initial winning neighborhood, and the initial value of the neuron weight.

[0105] The present invention provides a method for predicting the faults of a six-degree-of-freedom unmanned combat aircraft engine based on the Minimum Quantization Error (MQE) range, including:

[0106] For each normal data point collected for a six-degree-of-freedom unmanned combat aircraft engine to be predicted, the collected normal data is used to train the SOM model of the six-degree-of-freedom unmanned combat aircraft engine to be predicted.

[0107] For each abnormal data point collected from a six-DOF unmanned combat aircraft engine to be predicted, the minimum quantization error (MQE) of the abnormal data in the six-DOF unmanned combat aircraft engine to be predicted is obtained.

[0108] If the minimum quantization error (MQE) of the abnormal data in the six-degree-of-freedom unmanned combat aircraft engine to be predicted is within the range of the minimum quantization error (MQE), then the fault prediction of the six-degree-of-freedom unmanned combat aircraft engine to be predicted is performed.

[0109] The application embodiment of this invention features a fast target tracking module with high computational speed, fast target tracking speed, and accurate target selection. Based on an extended game theory model, it uses a combination of tree search and reinforcement learning to quickly prioritize multiple targets in a group setting. This scheme enhances the coordination capabilities of drone swarms in close combat, scientifically and rationally allocating target selection among multiple drones when facing multiple hostile targets, thus greatly improving the combat capabilities of the drone swarm. Simultaneously, the fault prediction module fully considers the individual differences in operating environment, service life, factory settings, and wear levels of different six-degree-of-freedom (6DOF) unmanned combat aerial vehicle (UCAV) engines. Multiple SOM (System of Quantization) models for these engines are trained separately to obtain accurate SOM model training results and corresponding minimum quantization error (MQE) for each engine. The MQE range corresponding to the warning interval can be determined by calculating the average MQE of the fault warning data from multiple trained 6DOF engines. This MQE range, determined in this way, can maximize the integration of individual differences among different 6DOF engines and can be used for fault warning of the same type of 6DOF engine; the fault prediction accuracy is high.

[0110] The machine game decision-making module in this invention utilizes techniques such as Monte Carlo tree search and deep reinforcement learning to generate different combinations of attack priorities based on the situational state of the unmanned combat aircraft, and performs future state projection and benefit evaluation to optimize target selection. Considering the individual differences in the engines of different unmanned combat aircraft, this module combines the results of the fault prediction module to achieve adaptive decision-making, avoid fault occurrence, and improve stability and reliability. Through real-time identification, tracking, and attack decision-making, combined with comprehensive analysis of information from multiple modules, this module can accurately detect and determine multiple hostile targets and make corresponding decisions in a timely manner.

[0111] The embodiments of the present invention can also adopt the following solutions:

[0112] The working principle includes data acquisition by the environmental perception module and data processing by the deep learning-optimized target recognition and tracking module. The specific working principle is as follows:

[0113] (1) Data acquisition of the environmental perception module: After various sensors (such as cameras, lidar, infrared sensors and millimeter-wave radar) acquire environmental data, these analog or digital signals need to be converted and processed. For example, cameras acquire image data, which needs to be preprocessed (denoising, enhancement, etc.); lidar acquires distance information, forming a point cloud data, which needs to be cleaned (removing invalid points, removing noise, etc.); infrared sensors and millimeter-wave radar acquire temperature and speed information, which need to be converted into appropriate data formats.

[0114] (2) Sensor data fusion: The main control module fuses the data acquired by each sensor. Typically, sensor fusion techniques such as Kalman filter and particle filter are used to effectively fuse data acquired by different sensors, or even the same sensor at different time points, to extract useful environmental and target feature information.

[0115] (3) Data processing of the target recognition and tracking module optimized by deep learning: This module uses deep learning algorithms to further process the data fused from the sensors. Common deep learning models include convolutional neural networks (CNNs) for image processing and target recognition, and recurrent neural networks (RNNs) or long short-term memory networks (LSTMs) for processing time series data, such as the motion trajectory of a target.

[0116] In target recognition, image features are first extracted using a convolutional neural network, and then the features are classified using a fully connected layer to obtain the target recognition result. In target tracking, RNNs or LSTMs can be used to process historical trajectory data and predict the target's future motion state.

[0117] It should be noted that embodiments of the present invention can be implemented in hardware, software, or a combination of both. The hardware portion can be implemented using dedicated logic; the software portion can be stored in memory and executed by a suitable instruction execution system, such as a microprocessor or dedicated-design hardware. Those skilled in the art will understand that the above-described devices and methods can be implemented using computer-executable instructions and / or included in processor control code, for example, such code provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The devices and modules of the present invention can be implemented by hardware circuitry such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field-programmable gate arrays, programmable logic devices, etc., or by software executed by various types of processors, or by a combination of the above-described hardware circuitry and software, such as firmware.

[0118] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions, and improvements made by those skilled in the art within the scope of the technology disclosed in the present invention, and within the spirit and principles of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A six-degree-of-freedom unmanned combat aerial vehicle close-range combat system based on machine game theory, characterized in that, include: Environmental perception module, main control module, wireless communication module, combat target recognition module, combat target tracking module, machine game decision-making module, shooting module, fault prediction module, and display module; Environmental perception module: Connected to the main control module, it uses cameras, lidar and millimeter-wave radar to monitor combat targets and acquire environmental information; Main control module: Connects to all modules and is used to control the normal operation of each module; The wireless communication module, connected to the main control module, is used to connect to a wireless network via a wireless chip for wireless communication. The combat target recognition module is connected to the main control module and the enhanced environmental perception module. It uses deep learning algorithms to identify combat targets, thereby improving the accuracy and efficiency of target recognition. The combat target tracking module is connected to the main control module and the deep learning-optimized target recognition module, and is used to track combat targets; The machine game decision-making module, connected to the main control module, is used to determine the priority of attack on combat targets; The firing module, connected to the main control module, is used to fire at combat targets; The fault prediction module, connected to the main control module, is used to predict generator faults in six-degree-of-freedom unmanned combat aircraft. The display module, connected to the main control module, is used to display monitoring videos, combat targets, and prediction results. Based on a method combining Monte Carlo tree search and deep reinforcement learning, future state projections and benefit assessments are generated for different strike priority combinations; based on the benefit assessment scores of strike list combinations, a strike priority list is generated. Affine transformation is used to transform each image to the same image coordinate system; The fault prediction module uses the following prediction method: 1) Configure the engine parameters of the six-degree-of-freedom unmanned combat aircraft, monitor the engine of the six-degree-of-freedom unmanned combat aircraft through monitoring equipment, obtain the engine operating parameters of the six-degree-of-freedom unmanned combat aircraft, and obtain the full life cycle data of multiple training six-degree-of-freedom unmanned combat aircraft engines; the full life cycle data is divided into normal data and abnormal data according to the usage time; the abnormal data includes fault warning data within the warning interval of the full life cycle; 2) Train the corresponding SOM model of the six-degree-of-freedom unmanned combat aircraft engine using normal data from each of the training six-degree-of-freedom unmanned combat aircraft engines, and obtain the minimum quantization error (MQE) of the fault warning data in the six-degree-of-freedom unmanned combat aircraft engine; Based on the average value of the minimum quantization error (MQE) of the fault warning data in the multiple training six-degree-of-freedom unmanned combat aircraft engines, the range of minimum quantization error (MQE) corresponding to the warning interval is determined. 3) Test the early warning accuracy of the minimum quantization error (MQE) range using multiple test six-degree-of-freedom unmanned combat aircraft engines respectively. Take the proportion of test six-degree-of-freedom unmanned combat aircraft engines whose accuracy exceeds the accuracy threshold as the test pass rate. If the test pass rate exceeds the first test pass rate threshold, then perform fault prediction of the six-degree-of-freedom unmanned combat aircraft engine to be predicted based on the minimum quantization error (MQE) range.

2. The six-degree-of-freedom unmanned combat aerial vehicle close-range combat system based on machine game theory as described in claim 1, characterized in that, The tracking method of the combat target tracking module is as follows: (1) Configure the camera parameters of the six-degree-of-freedom unmanned combat aircraft, continuously capture at least two images of the combat target using the camera on the six-degree-of-freedom unmanned combat aircraft, and perform noise reduction processing on the target images; and mark the ground coordinates, heading angle and time of acquisition of the combat target image of the six-degree-of-freedom unmanned combat aircraft in each image; and perform coordinate transformation on each combat target image so that they are in the same image coordinate system; (2) Filter multiple fighting target images, then extract the foreground image including the target image from the filtered fighting target image, extract feature points from the foreground image, extract the feature points of each region in the previous foreground image and match them with the feature points of the template of the tracked target, update the template of the tracked target using the successfully matched region, and match the feature points of the updated template of the tracked target with the feature points of each region in the current foreground image. (3) Based on the combat target situation generated in (1) and (2), and combined with the terrain obstacle module information obtained by the main control module, a list of targets can be formed, and the current UAV game situation is further modeled based on the extended game model. (4) Determine the speed and direction of the tracked target based on the positional relationship of the feature points in the regions of the two successfully matched foreground images; adjust the speed and heading of the six-degree-of-freedom unmanned combat aircraft based on the speed and direction of the tracked target, thereby tracking the target.

3. The six-degree-of-freedom unmanned combat aerial vehicle close-range combat system based on machine game theory as described in claim 1, characterized in that, Each test six-DOF unmanned combat aircraft engine tests its accuracy within the Minimum Quantization Error (MQE) range by performing the following steps: The SOM model of the test six-degree-of-freedom unmanned combat aircraft engine is trained using normal data from the test six-degree-of-freedom unmanned combat aircraft engine, and the minimum quantization error (MQE) of the abnormal data in the test six-degree-of-freedom unmanned combat aircraft engine is obtained. The minimum quantization error (MQE) of the abnormal data is obtained, and the abnormal data range within the range of the minimum quantization error (MQE) is obtained. The accuracy rate is obtained based on the ratio of the intersection of the abnormal data interval and the warning interval to the duration of the warning interval; The method further includes: if the test pass rate is between the second test pass rate threshold and the first test pass rate threshold, then adjust the SOM model parameters of the test six-degree-of-freedom unmanned combat aircraft engine with an accuracy rate lower than the accuracy rate threshold or the proportion of normal data in the test six-degree-of-freedom unmanned combat aircraft engine, and retrain the SOM model of the test six-degree-of-freedom unmanned combat aircraft engine, and reacquire the accuracy rate and test pass rate. Wherein, the second test pass rate threshold is lower than the first test pass rate threshold.

4. The six-degree-of-freedom unmanned combat aerial vehicle close-range combat system based on machine game theory as described in claim 3, characterized in that, The method further includes: if the test pass rate is lower than the second test pass rate threshold, adjusting one or more of the following parameters: the SOM model parameters of the training six-degree-of-freedom unmanned combat aircraft engine, the proportion of normal data in the full life cycle data, and the range of the warning interval, and redetermining the minimum quantization error (MQE) range, and re-acquiring the accuracy and test pass rate.

5. The six-degree-of-freedom unmanned combat aerial vehicle close-range combat system based on machine game theory as described in claim 1, characterized in that, Adjusting the SOM model parameters for training and testing the six-degree-of-freedom unmanned combat aircraft engine includes adjusting one or more parameters in the SOM model, such as the learning rate initialization parameter, the initial winning neighborhood, and the initial value of the neuron weights.

6. The six-degree-of-freedom unmanned combat aerial vehicle close-range combat system based on machine game theory as described in claim 1, characterized in that, The method of predicting the faults of the six-degree-of-freedom unmanned combat aircraft engine based on the minimum quantization error (MQE) range includes: For each normal data point collected for a six-degree-of-freedom unmanned combat aircraft engine to be predicted, the collected normal data is used to train the SOM model of the six-degree-of-freedom unmanned combat aircraft engine to be predicted. For each abnormal data point collected from a six-DOF unmanned combat aircraft engine to be predicted, the minimum quantization error (MQE) of the abnormal data in the six-DOF unmanned combat aircraft engine to be predicted is obtained. If the minimum quantization error (MQE) of the abnormal data in the six-degree-of-freedom unmanned combat aircraft engine to be predicted is within the range of the minimum quantization error (MQE), then the fault prediction of the six-degree-of-freedom unmanned combat aircraft engine to be predicted is performed.

7. A method for a six-degree-of-freedom unmanned combat aerial vehicle close-range combat system based on machine game theory as described in claim 1, characterized in that, The environmental perception module uses cameras, lidar and millimeter-wave radar to perceive the environment. The collected information is integrated and used by the main control module for target identification, tracking and shooting. The camera captures visible light to obtain color and texture information of the environment, and transforms it into useful information through image processing and computer vision algorithms; The lidar emits a laser beam and measures the time it takes for the reflected light to travel to obtain the target distance. For a large number of laser beams, it acquires a high-precision three-dimensional point cloud, providing accurate spatial information about the environment. The point cloud is in the form of P = {p_1, p_2, ..., p_n}, where each p_i = (x_i, y_i, z_i) represents a point in the point cloud. Millimeter-wave radar detects targets by emitting high-frequency electromagnetic waves and receiving the reflected waves. It can work effectively in various weather and lighting conditions and is used for the detection and tracking of moving targets. The main control module fuses the information acquired by these sensors, typically using a Kalman filter as the sensor fusion technology, to gain a comprehensive and accurate understanding of the environment, thereby optimizing identification, tracking, and firing tasks. Based on a method combining Monte Carlo tree search and deep reinforcement learning, future state projections and benefit assessments are generated for different strike priority combinations; based on the benefit assessment scores of strike list combinations, a strike priority list is generated. Affine transformation is used to transform each image to the same image coordinate system; The fault prediction module uses the following prediction method: 1) Configure the engine parameters of the six-degree-of-freedom unmanned combat aircraft, monitor the engine of the six-degree-of-freedom unmanned combat aircraft through monitoring equipment, obtain the engine operating parameters of the six-degree-of-freedom unmanned combat aircraft, and obtain the full life cycle data of multiple training six-degree-of-freedom unmanned combat aircraft engines; the full life cycle data is divided into normal data and abnormal data according to the usage time; the abnormal data includes fault warning data within the warning interval of the full life cycle; 2) Train the corresponding SOM model of the six-degree-of-freedom unmanned combat aircraft engine using normal data from each of the training six-degree-of-freedom unmanned combat aircraft engines, and obtain the minimum quantization error (MQE) of the fault warning data in the six-degree-of-freedom unmanned combat aircraft engine; Based on the average value of the minimum quantization error (MQE) of the fault warning data in the multiple training six-degree-of-freedom unmanned combat aircraft engines, the range of minimum quantization error (MQE) corresponding to the warning interval is determined. 3) Test the early warning accuracy of the minimum quantization error (MQE) range using multiple test six-degree-of-freedom unmanned combat aircraft engines respectively. Take the proportion of test six-degree-of-freedom unmanned combat aircraft engines whose accuracy exceeds the accuracy threshold as the test pass rate. If the test pass rate exceeds the first test pass rate threshold, then perform fault prediction of the six-degree-of-freedom unmanned combat aircraft engine to be predicted based on the minimum quantization error (MQE) range.

8. The method as described in claim 7, characterized in that, The six-degree-of-freedom unmanned combat aerial vehicle close-range combat method based on machine game theory includes the following steps: Step 1: Use the environmental perception module and cameras to monitor the combat target. Step two: The main control module uses a wireless chip to connect to the wireless network for wireless communication via the wireless communication module; Step 3: Identify the combat target using the combat target identification module; track the combat target using the combat target tracking module; Step four: Using the machine game module, a priority list of current combat targets is generated based on the current situation of the unmanned combat aircraft. Step 5: Based on the attack priority list, fire at the combat targets using the firing module; predict generator failures of the six-degree-of-freedom unmanned combat aircraft using the fault prediction module; Step six: Display the monitoring video, combat target, and prediction results through the display module.

Images