59 results about "Air combat" patented technology

The invention provides an air combat strategy generation method based on lightweight sequence modeling imitation learning. According to the training method, an air combat decision model based on rules can be quickly migrated into a parameter model in a neural network form. The method comprises the following steps: firstly, constructing a rule-based air combat decision model, and generating high-quality training data through adversarial simulation; and then, the training data is converted into a state-action sequence pair, and the state-action sequence pair is input into a lightweight sequence modeling network based on a Transform architecture for learning and training. According to the method, decision knowledge of a rule model is converted into a neural network model through an imitation learning method, and efficient utilization and flexible generalization of expert knowledge in air combat decision model construction are realized; the model designed by the method can accurately reproduce the decision behavior of the rule model through a small amount of training, and compared with a reinforcement learning method, the strategy construction speed is higher and the strategy performance is more stable.

The invention relates to the technical field of unmanned aerial vehicle control, in particular to an unmanned aerial vehicle path planning method based on deep reinforcement learning. The method comprises the following steps: constructing an unmanned aerial vehicle-missile confrontation environment model for providing reward calculation, termination judgment and a visual interface so as to provide training data and evaluate the path planning performance of a reinforcement learning strategy; an Actor-Critic deep neural network model based on strategy gradient is constructed, the unmanned aerial vehicle state information is taken as input, the maneuvering acceleration is taken as output, and adaptive learning of the unmanned aerial vehicle maneuvering strategy is realized; in the reinforcement learning process, mechanisms such as dominant function normalization, generalized dominant estimation and gradient cutting are introduced, and the training stability and convergence speed are improved; by adopting a layered reward mechanism and an improved PPO algorithm, the method effectively solves the problems of difficulty in training convergence, poor strategy stability and low energy utilization rate of a traditional deep reinforcement learning method in a complex air combat environment, realizes autonomous decision making and optimal avoidance path planning of the unmanned aerial vehicle under a high confrontation condition, and improves the unmanned aerial vehicle experience. And the avoidance success rate and the training efficiency are obviously improved.

The invention discloses an aircraft cluster multi-task scheduling system based on a cooperative game and a working method thereof. The system comprises a simulation environment module, a hierarchical strategy network module, a centralized value network module and a training and execution module. The simulation environment module is used for constructing a multi-agent air combat confrontation environment and generating states, rewards and interaction data required by training; the hierarchical strategy network comprises a shared space-time representation encoder f theta (.), a high-layer strategy network pi H (aHz) and a low-layer strategy network pi L (aLz; aH); the centralized value network is used for receiving global state information in a training stage, estimating the overall return of our agent cluster, calculating a dominant function, and realizing the optimization of the network by minimizing the value loss; and the training and execution module optimizes a centralized value network parameter and hierarchical strategy network parameters theta H and theta L by using a global state St in a centralized training stage, and outputs an air combat decision ai according to local observation independent decisions of each agent in a distributed execution stage.

The invention relates to the field of optical imaging and computer vision, discloses a large-view-field binocular image combination dynamic parameter adjusting system, and aims to solve the problems that a traditional head-up display is limited in view field and monochromatic display and information response are lagged. The system comprises a left and right eye image acquisition and preprocessing unit, a target positioning and trajectory prediction unit, a binocular parallax compensation and registration unit, a dynamic fusion display control unit, a color enhancement and information superposition unit, an eye movement tracking feedback unit and a parameter adaptive adjustment unit. Through binocular wide-angle acquisition, three-dimensional target prediction, sub-pixel-level image registration, panoramic stitching fusion, semantic color coding and dynamic parameter adjustment of eye movement driving, 120-degree * 60-degree distortionless panoramic display, multi-source information intelligent layered superposition and visual focus area adaptive enhancement are realized. According to the method, the air combat situation sensing efficiency and the weapon response speed are improved, and the method has high real-time performance (delay of lt, 80ms), strong environmental adaptability and engineering integration.

The invention relates to the technical field of artificial intelligence, and provides an unmanned aerial vehicle air combat aircraft action decision-making method based on near-end strategy optimization and a gating circulation unit, and the method comprises the steps: constructing a three-dimensional unmanned aerial vehicle air combat simulation environment; an Actor-Critic neural network architecture based on a GRU is constructed, and iterative training is performed on the neural network through a large amount of self-adversarial and learning. According to the scheme provided by the invention, the time sequence feature perception capability is enhanced, and the intelligent agent can better capture and understand the time sequence dependency relationship of the flight state in the air combat environment by introducing the GRU network, such as inferring the target intention or predicting the future situation from the continuous state change, so that a more prospective decision is made. And the method adapts to a complex dynamic environment, wherein the unmanned aerial vehicle air combat environment has high dynamic and uncertainty. According to the method, the unmanned aerial vehicle can gradually adapt to the complex environment and learn an effective confrontation strategy through an online interaction and learning mechanism of reinforcement learning and the processing capability of the GRU on the dynamic sequence.

This invention discloses a distributed MIMO radar resource allocation method for integrated reconnaissance, interception, and communication, belonging to the technical field of radar resource allocation. It aims to solve the performance balance problem under multi-task coordination in distributed MIMO radar air combat scenarios. An integrated reconnaissance, interception, and communication adversarial scenario is constructed, using the lower bound of position estimation error (SPEB), missile detection probability, and inter-node communication capacity as performance evaluation indicators for reconnaissance, interception, and communication, respectively. Then, with the goal of minimizing the total system transmit power, a model is constructed by combining relevant performance constraints and node task rules. Finally, the model is transformed into a mixed-integer second-order cone programming problem, and the non-convex terms are convexified using the McCormick envelope method. A numerical solution is obtained based on the CVX solver in MATLAB, achieving joint optimization of node task allocation and power allocation. This invention reduces the total transmit power while satisfying three types of performance requirements, achieving efficient waveform resource scheduling and improving the practicality and robustness of radar countermeasures.

The present application relates to a kind of reinforcement learning model training methods for unmanned aerial vehicle air combat decision-making, including several training rounds, in each training round, including: (1) setting network architecture and network parameters;(2) obtain input data, and input to the reinforcement learning model of current training round, obtain output data;(3) according to the decision data output by reinforcement learning model, the reward function of current training round is calculated, reward function is obtained by the superposition of basic reward and predicted gain reward, wherein, the predicted gain reward is determined by decision difference, the decision difference is the difference between the decision data output by the reinforcement learning model and the pre-determined large language model for the input data;(4) according to the reward function of current training round, adjust the network parameters of reinforcement learning model, obtain the initial network parameters of next training round;(5) return (1) execute next training round until reach the preset stop condition.

The application discloses a beyond-visual-range air combat decision-making method based on dynamic fire field and artificial potential field fusion, and relates to the field of aerospace. The method comprises the following steps: establishing a beyond-visual-range air combat simulation environment, modeling enemy and friendly camps in the air combat; grouping the enemy camp in real time; constructing a time-varying killing performance model of a missile to determine a three-view dynamic fire field of an aircraft-missile-target so as to establish an artificial potential field of combat groups, and determining the potential field gradient generated by each combat group in the grouping result; improving a particle updating formula in a particle swarm optimization algorithm by using the potential field gradient generated by the combat groups; solving a target function to determine a maneuvering decision of a friendly aircraft. The improved particle swarm optimization algorithm is obtained by improving the particle updating formula in the particle swarm optimization algorithm by using the potential field gradient generated by all the combat groups in the grouping result. The application can improve the coping ability, autonomous decision-making ability and coping ability in a complex battlefield environment of the beyond-visual-range air combat.

The application discloses an aircraft air combat deduction review iteration system, comprising a scenario editing module, a constraint configuration module and a simulation deduction module, wherein the scenario editing module comprises a quick scenario unit and a self-defined scenario unit, the quick scenario unit is used for quickly loading a saved combat scenario, the self-defined scenario unit comprises an environment scenario and a combat unit scenario, the constraint configuration module comprises one-key configuration constraints and self-defined configuration constraints, a set of scenario constraint schemes are generated through the scenario editing module and the constraint configuration module, the simulation deduction module comprises a batch simulation submodule and a data recording submodule, the batch simulation submodule performs simulation deduction based on the configured scenario constraint schemes, human intervention factors can be added at any time during the simulation deduction, a large amount of simulation data is generated for analysis, and the data recording submodule is used for recording data during the simulation deduction of the batch simulation submodule and reviewing the simulation deduction according to needs.

The application discloses a kind of real-time solving method of tactical control distance in air combat simulation environment, and is obtained by generating and training real-time solving model, and the model training process includes: constructing the three-dimensional motion model of beyond-visual-range air combat and air-to-air missile and its constraint model and setting simulation limit condition to obtain offline solving model;With each group of sample motion data, the maximum initial search range of the boundary of tactical control distance is used to obtain the corresponding tactical control distance value by enemy missile and aircraft motion simulation, and then a sample database is constructed;Based on sparse auto-encoding network, an initial real-time solving model is constructed, and a target real-time solving model is obtained by network training using the sample database.The corresponding tactical control distance value is obtained by inputting each group of motion data to be tested into the target real-time solving model, and then the boundary of tactical control distance can be obtained.The application first proposes the concept of tactical control distance and provides theoretical guidance for air combat decision through quantitative model, with high solving accuracy and strong real-time performance.

The application discloses a near-distance air combat pilot task performance prediction method, which predicts the task performance of a near-distance air combat pilot through a task performance prediction model, and the construction method of the task performance prediction model comprises the following steps: acquiring physiological state data, aircraft control data and air combat situation data of the pilot; extracting physiological state deep features, aircraft control deep features and air combat situation deep features from the physiological state data, the aircraft control data and the air combat situation data respectively; inputting the physiological state deep features, the aircraft control deep features and the air combat situation deep features into a cross-modal gated attention fusion module to fuse the physiological state deep features, the aircraft control deep features and the air combat situation deep features, so as to obtain cross-modal fusion features; and inputting the cross-modal fusion features into a full-connection predictor module to output a predicted pilot task performance grade, thereby realizing the prediction of the pilot task performance.

The application discloses a dynamic dimension air combat situation information fusion method based on a Set Transformer, belongs to the field of air combat situation information fusion, and comprises the following steps: fixed dimension and dynamic dimension air combat situation information categories are respectively defined and divided, dynamic dimension information categories in the air combat situation information are respectively expressed as set data, a Set Transformer neural network structure is defined, set data is respectively input into the Set Transformer to obtain set features with a fixed number and dimension, fixed dimension information in the air combat situation information is spliced with the set features with the fixed number and dimension output by the Set Transformer, and the spliced set features are taken as the input of subsequent machine learning technology; the air combat situation information fusion method provided by the application avoids the defect that a traditional feedforward neural network can only process fixed dimension input, realizes lossless and efficient processing of dynamic dimension input, and facilitates expansion of an upper algorithm between any number of formation battles.

The invention discloses a visual collaborative guidance simulation method and related equipment, and relates to the technical field of unmanned aerial vehicle collaboration, and the method comprises the steps: obtaining initial simulation data corresponding to a plurality of aircrafts in response to a collaborative guidance simulation instruction, and carrying out the visual processing of the initial simulation data based on a preset data preprocessing strategy, the data preprocessing strategy is constructed on the basis of a three-dimensional visualization platform, and the three-dimensional visualization platform is used for displaying the flight state of the three-dimensional visualization aircraft and analyzing the target simulation data on the basis of the target simulation data so as to carry out collaborative guidance simulation. According to the method, a data bridge from a simulation framework to a three-dimensional visualization platform is established, the simulation data is converted into the three-dimensional visualization scene in real time, three-dimensional real-time visualization of a complex air combat scene is achieved, and therefore the analysis efficiency and credibility of a simulation result are improved.

The application provides a simplified particle swarm weapon target distribution method with priority, and belongs to the field of resource distribution. First, weapon and target information of a combat unit is acquired, and priority of the combat unit to a strike target is calculated. Then, a distribution method is selected according to the priority, if the priority is less than a threshold value, a simplified particle swarm method is adopted, if the priority is greater than the threshold value, weapon distribution is directly performed according to threat degrees of the strike targets in sequence. The application determines the priority of the strike target, so that the strike target with higher threat is preferentially distributed with the weapon, and the threat degree of the combat unit is reduced. The application adopts the simplified particle swarm algorithm, and the convergence time of the algorithm can be shortened, which is of great significance for efficient strike of targets in cooperative air combat.

The invention discloses a dual-computer situation assessment and target allocation method, which performs calculation through an angle evaluation function and a distance evaluation function, adopts a PSO (particle swarm optimization) algorithm to perform weight optimization, accurately performs situation assessment, and performs target allocation more reasonably through analysis of a dual-computer situation assessment matrix and optimization of tactical decisions. The method improves the accuracy and efficiency of decision making, and is suitable for tactical decision support in modern air combats. The method is used for solving the problem of situation assessment and target allocation in the over-the-horizon dual-machine formation air confrontation process, and the key step is to make a situation assessment function and target allocation.

According to the collaborative maneuvering decision-making method in the air combat game confrontation of the multiple unmanned aerial vehicles disclosed by the invention, the dynamic response of the enemy is fully considered in the decision-making process by introducing the enemy behavior prediction model, so that the robustness and flexibility of decision-making are improved. Different from traditional multi-unmanned aerial vehicle game research, the method is characterized in that an enemy strategy is explicitly modeled, and online reasoning is carried out by adopting deep reinforcement learning. Specifically, the behavior of an enemy is predicted by establishing an enemy strategy modeling device, and perception and intention information is integrated in combination with a cognitive map collaborative network, so that the global visual field of decision making is improved. In addition, situation awareness information is processed by introducing a multi-head self-attention mechanism, prediction and evaluation of enemy response are optimized, and the adaptability of the decision making system is further enhanced. And in combination with centralized training and a distributed execution architecture, the problem of information asymmetry in collaborative decision making is solved, so that the unmanned aerial vehicle can make efficient decisions based on incomplete information in the execution stage.

The application discloses a kind of dual-machine formation 2v2 cooperative air combat auxiliary decision generation method, which realizes the efficiency promotion of dual-machine cooperative combat by the dual optimization of tactics and maneuver, and the dynamic analysis of air battlefield situation is realized by cooperative tactical decision-making, and the tactical formulation is carried out according to the situation relationship of enemy and our side, and then the combat target is selected;According to the tactics and the distribution of the combat target tactical role;Based on the target maneuver action library, the target prediction state is calculated, and then the final target prediction state is selected;Then the final target prediction state and the current state of our aircraft are input, the control amount of our aircraft in continuous domain is optimized by improved EDA algorithm, and the optimal flight trajectory of our aircraft is obtained.

This invention provides an autonomous decision-making method for UAVs in line-of-sight (LAS) two-to-one combat based on improved local reinforcement optimization, relating to the field of UAV technology. The method includes the following steps: Step S1, constructing an LAS air combat decision-making model; Step S2, determining preliminary maneuver decisions using an improved matrix decision-making method based on real-time information; Step S3, further optimizing the preliminary maneuver decisions using an improved local reinforcement optimization algorithm based on information exchange strategies to generate final maneuver decisions; Step S4, executing the final maneuver decisions and updating the UAV status; Step S5, repeating steps S2 to S4 until the maximum number of iterations is reached, at which point the iteration ends. This invention, employing the above-mentioned autonomous decision-making method for UAVs in LAS two-to-one combat based on improved local reinforcement optimization, improves the decision-making efficiency, collaborative combat capability, and overall combat effectiveness of UAVs in LAS two-to-one cooperative air combat.

The application belongs to the technical field of unmanned aerial vehicle air combat decision, and particularly relates to an unmanned aerial vehicle air combat decision method and system, wherein the unmanned aerial vehicle air combat decision method comprises the following steps: abstracting an air combat unmanned aerial vehicle situation, and constructing an unmanned aerial vehicle air combat situation graph; dividing the unmanned aerial vehicle air combat situation graph into multiple unmanned aerial vehicle air combat relationship graphs; performing feature extraction on each unmanned aerial vehicle air combat relationship graph, fusing features extracted from each unmanned aerial vehicle air combat relationship graph to obtain unmanned aerial vehicle air combat fusion features; and generating a friendly unmanned aerial vehicle air combat strategy based on the unmanned aerial vehicle air combat fusion features.

The invention belongs to the technical field of unmanned aerial vehicles, and discloses a self-adaptive unmanned aerial vehicle cooperative formation task decision-making method based on a knowledge base and a dynamic Bayesian network, and the method comprises the steps: firstly, analyzing the features of an unmanned aerial vehicle beyond visual range air combat task, dividing the execution of the air combat task into an overall task layer and a task cooperation layer, establishing a task-oriented unmanned aerial vehicle autonomous decision framework; secondly, according to the air combat process of the unmanned aerial vehicle, air entities possibly encountered by the unmanned aerial vehicle formation in the air combat process are analyzed, the combat experience of a pilot is described by a generative rule, and a rule base is established; and finally, analyzing factors influencing the collaborative style during task execution, and establishing a dynamic Bayesian network decision model. A simulation result shows that when the air combat situation changes, along with the change of the state of the air entity, the proposed method can quickly make a decision, and the decision result conforms to the air combat experience of a pilot.

The invention relates to the technical field of air combat intelligent aid decision making, in particular to an aid decision making recommendation model adaptive method and system based on a Bayesian neural network, and is particularly suitable for man-machine collaborative decision making in a 2V2 air combat game scene. By designing a self-adaptive framework and an incremental learning algorithm, the model can be dynamically updated based on real-time data in the use process of a user, so that the recommendation accuracy and robustness are improved. The method comprises the following steps: constructing an incremental learning algorithm of a Bayesian neural network, analyzing the time complexity and the space complexity of the incremental learning algorithm, evaluating the real-time performance of the incremental learning algorithm in an online reasoning process, and researching the robustness performance under the condition of data uncertainty. The method can be widely applied to strategy selection in an intelligent recommendation system, an auxiliary decision-making platform and a multi-agent game.

This invention provides a training platform and method for air combat command and decision-making intelligent agents, comprising: a support layer: providing general hardware and software support for modeling, training, demonstration, and verification of command and decision-making intelligent agents, as well as system confrontation modeling and simulation; a resource layer: building a combat simulation model library based on a general simulation platform to quickly build a learning environment to meet different air combat command and decision-making intelligent agent training needs, and building an intelligent agent model library based on a general reinforcement learning platform to meet different command and decision-making intelligent agent training needs; and an application layer: decomposing actual command and decision-making problems according to the application needs of combat command and decision-making intelligence, and specifically integrating intelligent agent training application systems to carry out parallel training of command and decision-making intelligent agents and related technology verification applications. This invention flexibly and quickly constructs intelligent agents and their learning environments, and, guided by military business rules, combines multi-agent parallel training with large scenario samples to improve the training efficiency and generalization and expansion capabilities of command and decision-making intelligent agent models.

A kind of air combat real-time situation comment generation method and system based on deep learning.The present application relates to the field of air combat situation comment generation, in particular to a kind of air combat real-time situation comment generation method and system.The method mainly includes comment model acquisition step and situation comment generation step, through real-time air combat structured situation data, cooperate with comment text sequence generation network model, corresponding artificial comment content can be generated extremely fast, its generation speed and accuracy will far exceed ground artificial comment.It can make ground command personnel more accurately master the real-time situation of this air combat, to make the most timely and correct command decision.

The invention discloses a multi-machine multi-missile collaborative air combat decision-making method based on SM-HAPPO deep reinforcement learning, and relates to the technical field of air combat intelligent decision-making. Starting from a multi-aircraft multi-missile collaborative air combat decision-making method, a dynamic model of an aircraft, a missile guidance model and an air combat situation model are established, and simulation of a complex environment is achieved. For collaborative decision-making of isomorphic or heterogeneous aircrafts, a PPO algorithm is expanded to the field of heterogeneous agents, a GRU-based state memory module is introduced, and an SM-HAPPO algorithm suitable for deep reinforcement learning of the heterogeneous agents is provided. In the strategy network, a loss function with a mask mechanism and a strategy entropy regular term is introduced, and the stability and exploratory performance of the algorithm are improved. Finally, a simulation experiment of aircraft collaborative air combat decision making is completed in a python environment, and it is proved that compared with a traditional intelligent algorithm, the collaborative decision making capacity and the air combat winning rate of the unmanned aerial vehicle can be further improved.

The application provides an air combat intelligent decision-making method and device based on course reinforcement learning, equipment and medium, and relates to the field of reinforcement learning. It comprises the following steps: determining corresponding battle advantage indexes based on multiple sets of initial state data; storing the initial state data in a difficulty-guided state pool after sorting based on the battle advantage indexes; updating the initial state data in the progressive sliding buffer in an increasing difficulty manner from the difficulty-guided state pool; dynamically formulating intermediate task targets until the final task target is formulated; sampling the current initial state data from the progressive sliding buffer and inputting the current initial state data into the policy network of the SAC model to obtain the current action; training the SAC model based on at least the intermediate task target, the current action and the current initial state data until the final task target is reached to obtain the trained SAC model for making intelligent decisions, thereby improving the decision-making ability and adaptability of the aircraft in complex environments.

This invention discloses a method for providing 3D target information prompts in virtual air combat scenarios. This method adjusts the prompting effect based on the distance between the user's target and non-user targets, reducing the workload of the virtual scenario, improving the display effect from the simulated perspective, and enhancing human-computer interaction in virtual air combat scenarios.

The invention discloses an air-space integrated quantum high-energy laser platform (YRCW stubble for short), and belongs to the technical field of aerospace, directional energy weapons and space resource development. The platform adopts a flying wing type fusion body aerodynamic layout, and the core is that advanced subsystems such as a thorium-based molten salt reactor energy system, a quantum high-energy laser transmitter and a quantum radar detection module are integrated; through cooperation of eight steps of energy activation, energy storage preparation, perception construction, system readiness, global maneuvering, target strike, communication guarantee and multi-domain adaptive adjustment, global maneuvering (the speed is more than 0-25 Mach, and the flight height is more than 0-80 thousand kilometers) from airport horizontal takeoff to near-earth orbit deployment is realized. The system has the remarkable advantages that nuclear energy, quantum technologies and directional energy weapons are deeply fused, and a comprehensive aerospace combat platform with strategic early warning, remote accurate strike, on-orbit service and space resource development capabilities is formed. The platform core subsystem can be independently expanded to land-based, ship-based, stratosphere air-based and space-based application scenes. Particularly, the invention provides an independent energy source and application scheme of the thorium-based molten salt reactor power generation module in remote mountainous areas, disaster rescue and lunar helium-3, water and ice resource development, and detailed technical and economic data are supplemented, so that the practical value and strategic significance of the technology are greatly expanded.

The invention relates to a complementary training method for improving the robustness of a reinforcement learning interception strategy, and the method comprises the steps: obtaining an average confrontation combat scene, describing the average confrontation combat scene through a first-order delay dynamics model, and defining a defense missile guidance process in the average confrontation combat scene as a Markov decision process; on the basis of a Markov decision process, a strategy is updated step by step by using a PPO algorithm based on a course learning mode, a basic interception strategy is obtained, full-scene performance evaluation is performed on the basic interception strategy, failure scenes with miss distance exceeding standards are screened out to construct a failure scene list, and mixed initial state distribution is established by using the failure scene list. And difference compensation training is carried out through a PPO algorithm to obtain an optimal interception strategy. Aiming at the problem that a specific scene is easy to fail in a complex air combat interception scene, the robustness and reliability of an interception strategy are improved by introducing a failure scene compensation training mechanism, and the task success rate in a diversified combat environment is ensured.

This invention discloses a dynamic task allocation method for UAV swarms based on coyote pack prey selection behavior: Step 1: Initialize UAV swarm air combat conditions; Step 2: Execute target selection tasks based on coyote pack behavior mechanisms; Step 3: Generate strike formations based on coyote hunting time window constraints; Step 4: UAV swarm executes strike / tracking tasks; Step 5: Update the air combat scenario map; Step 6: Reassign UAV swarm combat tasks. This invention addresses the problem of dynamic task allocation for tactical reconnaissance and strike integrated UAV swarms in air combat environments. Considering the dynamic real-time situation of the combat environment and the potential collaborative relationships of enemy swarms, it provides a dynamic task allocation method based on coyote prey selection behavior. This method prioritizes striking weaker targets, suppresses opportunities for enemy UAV situational shifts, reduces enemy swarm support and counter-encirclement, and effectively improves the combat effectiveness of UAV swarms.

This invention, belonging to the field of UAV technology, proposes a method for visual-to-visual (V2V) air combat maneuver decision-making in obstacle environments based on a mirror inference module. First, it constructs a kinematic and situational model of the UAV. Second, it models air combat game strategies using reinforcement learning and constructs state and action spaces. Third, it designs a proximal strategy optimization algorithm based on a curriculum-based self-game training framework and an attention mechanism. Fourth, considering factors such as angle, speed, and distance obstacle avoidance, it designs a comprehensive reward function to guide the UAV in obstacle avoidance and game-playing. Finally, it trains V2V air combat maneuver decisions in complex obstacle environments. This invention employs a curriculum-guided progressive experience accumulation and a self-game-driven adversarial strategy evolution approach, which improves training efficiency and enhances tactical diversity. The introduced attention mechanism flexibly adjusts the weights of the reward function, enabling the trained maneuver strategy to adapt to highly dynamic air combat scenarios.