A game-guided decision control method for unmanned aerial vehicle

Abstract

This invention discloses a game-theoretic guidance-based UAV decision control method, belonging to the field of UAV decision control technology. The method includes: acquiring information on individuals participating in attack and defense; performing target matching on these individuals based on preset rules; performing game-theoretic guidance calculations based on the target matching results; converting the game-theoretic guidance calculation results into desired attitude and throttle, and sending them to the flight control stack; the inner loop of the flight control stack uses PID control to convert the desired attitude and throttle into motor speed and servo control quantities, controlling the actions in simulation or on a physical UAV. This invention supports decision control solutions and inputs at the cluster level, and is not limited to individual control. It can achieve optimal control under high-maneuverability pursuit and escape conditions. This invention is aimed at highly realistic UAV nodes and physical UAV platforms, and compared to current differential game theory technology and other highly intelligent modern flight control technologies, it has higher feasibility and technological maturity.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) decision control technology, and more specifically to a game-theoretic guidance-based UAV decision control method. Background Technology

[0002] The control problem of aircraft is essentially a trajectory tracking problem. Current solutions to the aircraft control problem can be categorized into two types. One type of approach decomposes the problem into a guidance and control problem, and further separates it into an external guidance loop and an internal control loop, referred to as the outer loop and inner loop, respectively. The other type of approach integrates both the inner and outer loop functions, such as rolling time domain, differential smoothing, and neural networks. However, due to its versatility, reliability, and interpretability, this type of approach has not yet been widely adopted.

[0003] Due to the emergence of guided intercept munitions and the need for maneuvering pursuit in aerospace, existing technologies use the principles of game theory, variational methods, and control theory to solve problems involving dynamic conflicts between two or more agents. By employing differential dynamic programming, differential game theory can break free from the discrete-time constraints of traditional game theory and solve for real-time, dynamic optimal equilibrium strategies.

[0004] Existing flight control algorithms for real aircraft can solve the tracking problems of waypoints and tracks well, but they cannot meet the optimal requirements for tracking and controlling highly intelligent and maneuverable targets. Meanwhile, differential game theory methods, which are designed for optimal decision-making in pursuit and escape, have long remained at the level of numerical analysis and calculation. The aircraft mass model is far from the actual model, and most research scenarios are simplified to a two-dimensional plane, which is also far from practical application.

[0005] Therefore, how to propose a game-theoretic guided UAV decision control method, applicable to real-world scenarios, and improve the optimality of UAV decision control results in pursuit scenarios is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] In view of this, the present invention provides a game-theoretic guided UAV decision control method, applicable to target pursuit, two-agent combat, and swarm attack and defense; it also supports centralized and distributed flight decision control. To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A game-theoretic guided UAV decision-making and control method includes:

[0008] Obtain information on individuals participating in the attack and defense, and perform target matching on these individuals based on preset rules;

[0009] Perform game-guided calculations based on the target matching results;

[0010] The game-theoretic guidance calculation results are converted into desired attitude and throttle and then sent to the flight control stack.

[0011] The inner loop of the flight control stack uses PID controllers to convert the desired attitude and throttle into motor speed and servo control values, thereby controlling the actions in the simulation or the actual device.

[0012] Optionally, obtaining the individual information participating in the attack and defense includes: the position and Euler angle information of all individuals in the pursuing group and the escaping group.

[0013] Optionally, the individual needs to check its own state before sending a message to other individuals, and determine the scale of the cluster game based on the number of drones sending and receiving messages in different clusters.

[0014] Optionally, the game-guided calculation based on the target matching result includes: outputting the target matching result to the UAV, each UAV performing game-guided calculation based on its own and the state information of the matching object, or obtaining the acceleration of each UAV through centralized calculation and then sending instructions to each UAV to perform target matching and decision-making.

[0015] Optionally, the target matching is based on an auction mechanism, whereby the pursuing group determines the ownership of each target drone through a bidding and auction cycle involving individuals within the escape group.

[0016] Optionally, the value of the escaping swarm of drones can be determined using a value function based on the three-dimensional Dubins path length.

[0017] Optionally, the game-theoretic guidance calculation includes: calculating the three-axis acceleration of each UAV based on the pairing formed by each UAV after target matching, and the required information for calculation includes the position, velocity and attitude of the UAV.

[0018] Optionally, the step of converting the game-guided calculation results into the desired attitude and throttle includes:

[0019] The acceleration obtained from the game-guided calculation is transformed from the fixed coordinate system to the carrier coordinate system according to the coordinate system transformation method; the desired roll angle is generated from the y-axis acceleration in the carrier coordinate system; the desired velocity and desired altitude are generated from the x-axis and z-axis acceleration in the carrier coordinate system, and then converted into the desired pitch angle and desired throttle after being calculated by the Total Energy Control System (TECS).

[0020] Optionally, the desired pitch angle, desired roll angle, and desired throttle are sent to the inner loop of the flight control stack via external program communication. The external program communicates with the flight control stack and each UAV flight control stack via ROS2. Each aircraft's flight control stack and the external program are different communication nodes, and each UAV belongs to a different domain.

[0021] Optionally, the target matching is applied to the pursuit and escape of individuals among up to three clusters. For the pursuit and escape of individual drones, no target matching process is required. For the attack and defense between two individuals, the target matching step is skipped directly.

[0022] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a game-theoretic guidance-based UAV decision control method, which has the following beneficial effects:

[0023] Compared to the PX4 autopilot, this invention supports cluster-level decision control solutions and inputs, and is not limited to individual control. It achieves optimal control under high-maneuverability pursuit and escape conditions, is more intelligent than existing traditional UAV control technologies, and offers greater reliability and interpretability compared to neural network-based UAV control technologies. This invention targets highly realistic UAV nodes and physical UAV platforms, and compared to current differential game theory techniques and other highly intelligent modern flight control technologies, it possesses higher feasibility and technological maturity. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0025] Figure 1 This is a schematic diagram of the framework of a game-theoretic guidance-based UAV decision control method provided by the present invention.

[0026] Figure 2 This is a schematic diagram of the coordinate system of the UAV carrier provided by the present invention.

[0027] Figure 3 This is a schematic diagram of the three-party cluster attack and defense pairing provided by the present invention.

[0028] Figure 4 This is a schematic diagram of the Dubins path trajectory under low elevation differences provided by the present invention.

[0029] Figure 5 This is a schematic diagram of the Dubins path trajectory under medium height difference provided by the present invention.

[0030] Figure 6 This is a schematic diagram of the Dubins path trajectory under high elevation differences provided by the present invention.

[0031] Figure 7 The game-theoretic guidance control loop diagram provided by this invention.

[0032] Figure 8 The schematic diagram of the outer loop controller based on game-theoretic guidance provided by this invention.

[0033] Figure 9 This is a schematic diagram illustrating the working principle of FAST RTPS provided by the present invention.

[0034] Figure 10 A schematic diagram illustrating the centralized decision-making and control principle provided by this invention.

[0035] Figure 11 This is a schematic diagram illustrating the distributed decision control principle provided by the present invention. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] This invention discloses a game-theoretic guidance-based decision-making and control method for unmanned aerial vehicles (UAVs), such as... Figure 1 As shown, it includes:

[0038] Obtain information on individuals participating in the attack and defense, and perform target matching on these individuals based on preset rules;

[0039] Perform game-guided calculations based on the target matching results;

[0040] The game-theoretic guidance calculation results are converted into desired attitude and throttle and then sent to the flight control stack.

[0041] The inner loop of the flight control stack uses PID controllers to convert the desired attitude and throttle into motor speed and servo control values, thereby controlling the actions in the simulation or the actual device.

[0042] Specifically, a game-theoretic guided UAV decision-making and control method includes:

[0043] In the complete mission flow, each UAV, after switching to offboard mode, makes independent decisions based on the decision-making methodology, with the onboard computer running the decision-making program and inputting the decisions into the autopilot for actual control. Each decision-making and control cycle includes the following processes:

[0044] First, it is assumed that all drones can communicate with each other. Before sending messages to other individuals, each drone will check its own status to ensure that it is in good condition to participate in the attack and defense. The scale of the cluster game is determined based on the number of drones sending and receiving messages in different clusters.

[0045] Secondly, each drone performs target matching between two groups of pursuit clusters: the defense and attack cluster, and the attack and target cluster. The required information includes the positions and Eulerian angles of all individual drones in both the pursuing and escaping clusters. Target matching is crucial for decoupling cluster decision-making. The matching process is based on an auction mechanism, where the pursuing cluster determines the ownership of each target drone through a bidding and auction cycle involving individuals within the escaping cluster. The value of the escaping drone is determined using a value function based on the three-dimensional Dubins path length.

[0046] Then, based on the pairing formed after target matching, the three-axis acceleration of each UAV is calculated by differential game guidance. The required information for the calculation includes the position, velocity and attitude of the three UAVs.

[0047] Finally, for the calculated acceleration, it is first transformed from the fixed coordinate system to the carrier coordinate system according to the coordinate system transformation method, such as... Figure 2 As shown, the desired roll angle is generated from the y-axis acceleration in the carrier coordinate system; the desired velocity and desired altitude are generated from the x-axis and z-axis accelerations in the carrier coordinate system. These are then calculated by the Total Energy Control System (TECS) and converted into the desired pitch angle and desired throttle. The desired pitch angle, roll angle, and throttle are sent to the flight control inner loop via external program communication. The communication process between the external program and the flight control system, as well as between the flight control systems of each aircraft, is handled by ROS2.

[0048] It should be noted that the target matching method can be used to achieve the pursuit and escape of individuals among up to three clusters. The target matching and decision-making process can be carried out in a centralized manner to calculate the acceleration of each machine and then send instructions to each machine, or it can be carried out in a distributed structure, where each machine obtains the information required for decision-making and calculates the acceleration independently.

[0049] Specifically, a game-theoretic guided UAV decision-making and control method is proposed for pursuit and escape between UAVs and another entity, and can also be extended to offense and defense between up to three individuals in a cluster. It comprises three main modules: target matching, game-theoretic guided decision-making, and flight control. As a distributed decision-making method, the framework is identical for all UAVs.

[0050] The steps are as follows:

[0051] S1: After the program runs, the drone switches from normal flight mode to offboard external control mode. This mode switching process is automatic as the program runs. After switching, the number and faction of individuals participating in the attack and defense are first determined, as well as the number of message sending and receiving nodes. For individual drone pursuit, no target matching process is required. If the number of attack and defense targets is greater than one, or the scale has reached the level of attack and defense between clusters, then a target matching process is first performed for attack and defense pairing. This pairing function can be extended to attack and defense between three clusters, such as... Figure 3As shown, there is theoretically no upper limit to the number of applicable individuals. For attack and defense between two individuals, the target matching step is skipped entirely.

[0052] S2: After target matching, each aircraft receives the matching result and performs game-theoretic guidance calculations based on its own and the matched target's state information. The game-theoretic guidance calculation outputs the three-axis acceleration of each aircraft in a fixed coordinate system, which is then post-processed into the desired attitude and throttle before being sent to the flight control system.

[0053] S3: The flight control stack is mainly divided into an outer loop and an inner loop. The game result is sent from the outer loop to the inner loop, where it is converted into motor speed and servo control inputs via PID control. This is used to control the actions of the simulation or the actual aircraft.

[0054] S4: The entire communication process uses the ROS2 framework. Each aircraft's flight control and game guidance program is a different communication node, and each aircraft belongs to a different domain to avoid interference between internal aircraft messages.

[0055] In a specific implementation, the target matching steps are as follows:

[0056] The matching principle between pursuit drones is to select pairs that maximize the pairing value between the i-th pursuer and the j-th escapee. ij The sum is maximized. The optimal allocation function is as follows:

[0057]

[0058] A value function based on a 3D Dubins path is adopted as the optimal allocation function for the aerial pursuit problem. Generating the 3D Dubins path requires first calculating the 2D Dubins path corresponding to the ground projections of the aircraft's starting and ending points' positions and velocity directions, and also requires knowing the maximum pitch angle. The 3D Dubins path has three different cases, depending on the altitude difference |z| between the starting and ending points. e -z s | The two-dimensional Dubins path length L of the ground projection car and maximum pitch angle θ max The limitations are that these three situations are defined as low altitude difference, medium altitude difference, and high altitude difference, such as... Figure 4 , Figure 5 and Figure 6 As shown.

[0059] Before assessing the situation, it is necessary to determine the minimum turning radius of the drone. For fixed-wing drones, under ideal conditions of constant speed, no autopilot lag, and sufficiently fast aircraft angle adjustments, the relationship between yaw and roll angles is given by the coordinated turn conditions:

[0060]

[0061] Where ψ is the yaw angle and φ is the roll angle. Due to the geometric relationship between the turning radius and the yaw angle:

[0062] R×ψ=V

[0063] The minimum turning radius of a drone is given by the following formula:

[0064]

[0065] Where V is the aircraft's fixed flight speed, and g is the acceleration due to gravity. The three-dimensional Dubins path length L under three altitude differences. air It is given by the following formula:

[0066]

[0067] Among them, z s and z e θ represents the heights of the starting and ending points, respectively. * The optimal pitch angle at low elevation differences satisfies:

[0068]

[0069] φ * The optimal central angle for the intermediate arc under moderate elevation differences satisfies:

[0070] [L car (φ * )tanθ max =|z e -z s |

[0071] R * The optimal turning radius under high elevation differences satisfies the following when the number of coils in the helix is ​​k:

[0072] [L car (R * )+2πkR * ]tanθ max =|z e -z s |

[0073] Substituting the path length into the following formula yields the value based on the 3D Dubins path:

[0074]

[0075] In a specific implementation, the principle of the game-theoretic guided decision-making module includes:

[0076] The essence of game theory is an optimization process. First, the Hamiltonian function H is constructed as follows:

[0077]

[0078] In the formula, and R represents the control acceleration vectors of the interceptor and the target, respectively. p and R e The acceleration control weight matrices for the interceptor and the target are respectively, λ represents the terminal condition, F and G are the state coefficient matrix and the input coefficient matrix, respectively, and y 12 It is a vector representing relative state (relative position and relative velocity). for These are the additional interference vectors for the interceptor and the target, respectively (assumed to be 0).

[0079] According to the necessary conditions for optimization, the partial derivatives of the above equation have the following properties:

[0080]

[0081]

[0082]

[0083] Substituting the Hamiltonian function, we get:

[0084]

[0085]

[0086] Moving the control input term to the left side of the equation yields the following form:

[0087]

[0088]

[0089] Since the desired outcome is to construct the control input as a function of the system state, λ can be set to the following form:

[0090] λ=Py 12 +Q

[0091] Where P is a 6×6 matrix, a solution to the matrix Riccati differential equations (MRDE); Q is a 6×1 vector, which will be shown later to be a solution to the vector Riccati differential equations (VRDE). Observing the form, it can be seen that matrix P is related to the game input, while Q contains additional acceleration. Substituting λ, we have:

[0092]

[0093]

[0094] From the necessary conditions and the Hamiltonian function, we can obtain:

[0095]

[0096] Substituting into the expression for λ, after expansion and algebraic simplification, we get:

[0097]

[0098] Since the above formula requires that for all y 12 Therefore, y is true. 12 The coefficient of and the right side of the equation must be equal to zero, that is:

[0099]

[0100]

[0101] Next, in order to apply the mathematical model to practice, the weight matrix structure is defined as follows:

[0102]

[0103]

[0104] And define matrix R:

[0105] R p =(R p ) -1 -(R e ) -1

[0106] Let the structure of matrix R be:

[0107]

[0108] The elements in the matrix have the following relationship:

[0109]

[0110] To obtain the applicable results, the following settings are made:

[0111] s 11 =s 22 =s 33 =s1

[0112] s 14 =s 25 =s 36 =s2

[0113] s 44 =s 55 =s 66 =s3

[0114]

[0115]

[0116]

[0117] The relevant parameters for game-theoretic guidance are s1, s2, s3, and r. p r e constitute.

[0118] Let the remaining time T = t f -t, t f To determine the deadline, let t be the current time. According to the MRDE solution, we can obtain:

[0119]

[0120]

[0121]

[0122] The feedback gain matrices for the pursuers and the escapees are as follows:

[0123]

[0124]

[0125] The control input feedback function for both the pursuer and the escapee is:

[0126]

[0127]

[0128] The control principle of game-guided control is represented by a control loop block diagram. Substituting the acceleration expression into the kinematic equations, the implementation of game-guided control is as follows: Figure 7 As shown.

[0129] In a specific implementation, the flight control steps are as follows:

[0130] (1) The airspeed and altitude of a fixed-wing aircraft are simultaneously controlled via a Total Energy Control System (TECS). TECS provides a solution that reflects the problem based on energy rather than initial setpoints. The total energy of an aircraft is the sum of its kinetic and potential energy, and thrust, controlled by the throttle, increases the aircraft's total energy. A given total energy state can be achieved through any combination of potential and kinetic energy. That is, the total energy of an aircraft flying at low airspeed at high altitude is equivalent to that flying at high airspeed at low altitude. This situation is called specific energy balance, which is calculated based on the current altitude and the actual airspeed setpoint. The specific energy balance of the aircraft can be controlled by controlling the pitch angle. Increasing the pitch angle converts kinetic energy into potential energy, and decreasing the pitch angle does the opposite. Thus, by converting the initial airspeed and altitude setpoints into energy levels, where airspeed and altitude are coupled, while energy levels can be controlled independently, the control problem can be decoupled. The throttle is used to adjust the specific total energy of the aircraft, and the pitch angle is used to maintain a specific balance point between potential energy (altitude) and kinetic energy (airspeed). The game-theoretic guidance principle including the TECS module is as follows: Figure 8 As shown.

[0131] (2) The flight module mainly uses the attitude control of the inner loop of the PX4 flight control stack. The attitude controller works by using a cascaded loop method. The outer loop calculates the error between the attitude setpoint and the estimated value, and multiplies the error by a gain (proportional controller) to generate the angular rate setpoint. The inner loop calculates the angular rate error and uses a proportional-integral controller to generate a desired angular acceleration.

[0132] Then, based on the desired angular acceleration and prior system information, the angular offset of the actuators (ailerons, horizontal stabilizers, vertical stabilizers, etc.) is calculated through control allocation (also known as mixed control). Furthermore, since control surfaces are more efficient at high speeds and less efficient at low speeds, the controller, adjusted according to cruise speed, uses airspeed measurements for scaling.

[0133] Without an airspeed sensor, gain adjustment for fixed-wing attitude control would be disabled, and airspeed feedback would not be available in the total energy control system. However, to minimize lateral acceleration caused by sideslip, the yaw controller uses steering coordination constraints to generate a yaw rate setpoint.

[0134] Feedforward gain is used to compensate for aerodynamic damping. The two main torque components around the fuselage axis come from: the control surfaces (ailerons, horizontal stabilizer, vertical stabilizer, driving the fuselage rotation) and aerodynamic damping (proportional to the fuselage angular rate, preventing the fuselage from rotating). To maintain a constant angular rate, feedforward can be used in the angular rate loop to compensate for this aerodynamic damping.

[0135] The roll and pitch controllers have the same structure and assume that the longitudinal and lateral dynamics are sufficiently decoupled to operate independently. However, to minimize the lateral acceleration caused by aircraft sideslip, the yaw controller utilizes steering coordination constraints to generate a yaw rate setpoint. The turn coordination algorithm is based solely on coordinated turn geometry calculations, as shown in the following equation. Where... Here, yaw rate is the setpoint, g is the gravitational acceleration, and V is the yaw rate. T Let φ be the longitudinal velocity of the aircraft. sp and θ sp These are the desired roll angle and pitch angle, respectively. Yaw rate control helps counteract adverse yaw effects and can prevent roll by providing additional damping.

[0136]

[0137] In a specific implementation, the ROS2 communication steps are as follows:

[0138] The FAST RTPS bridge, also known as the micro RTPS bridge, adds a real-time publish-subscribe (RTPS) interface to the PX4 autopilot, enabling real-time exchange of uORB messages between various flight control internal components and (off-board) programs. It also allows for better integration with applications running and linked in ROS2, sharing sensor data, commands, and other aircraft information.

[0139] The micro RTPS bridge exchanges messages between flight controllers and between flight controllers and external programs, such as target matching and game-theoretic guidance programs, seamlessly translating between uORB and RTPS / DDS messages used by each system. Its main architectural components include... Figure 9 The client and agent processes are shown.

[0140] (1) Client

[0141] The client is a middleware daemon running on the flight controller. The client subscribes to uORB topics published by other PX4 autopilot components and sends any updates to the agent via UART or UDP ports. It also receives messages from the agent and publishes them as uORB messages to the PX4 autopilot.

[0142] (2) Agent

[0143] The agent runs as a background process on an external computer, i.e., outside the flight controller. This agent monitors uORB update messages from clients and publishes these messages via RTPS. It also subscribes to uORB RTPS / DDS messages from other DDS participant applications and forwards them to clients.

[0144] (3) Communication between agent / client

[0145] The agent and client connect via a serial link (UART) or UDP network. uORB messages are serialized using CDR (CDR serialization provides a common format for exchanging serial data between different platforms) before being sent.

[0146] The agent and any Fast DDS application connect via UDP and can reside on the same or different devices. In a typical configuration, they will reside on the same system, such as a development computer, a Linux companion computer, or a computing board, and connect to the client, either via Wi-Fi or USB.

[0147] This invention designs a game-theoretic guidance-based aircraft decision control method, applicable to target pursuit and escape, two-agent systems, and swarm attack and defense. The control decision method uses an auction-framework-based target matching method to scale up flight control. For single-aircraft decision control, a game-theoretic guidance method is used as the outer loop of the controller, and the TECS module is used to convert the desired acceleration into the desired attitude, which is then directly input into the inner loop of the flight control for PID control. The ROS2 framework is used as the communication means between the flight controllers of various UAVs and between the flight controller and other programs (target matching and game-theoretic guidance), and FAST RTPS is used as the message format conversion interface. It supports both centralized and distributed flight decision control, such as... Figure 10 , Figure 11 As shown.

[0148] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0149] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A game-theoretic-guided decision-making and control method for unmanned aerial vehicles (UAVs), characterized in that, include: Obtain information on individuals participating in the attack and defense, and perform target matching on these individuals based on preset rules; Perform game-guided calculations based on the target matching results; 3D Dubins path length under three height differences It is given by the following formula: in, and These are the heights of the starting point and the ending point, respectively. The optimal pitch angle at low elevation differences satisfies: ； The optimal central angle for the intermediate arc under moderate elevation differences satisfies: ； The optimal turning radius under high elevation differences satisfies the following when the number of coils in the helix is ​​k: ； Substituting the path length into the following formula, we obtain the value based on the 3D Dubins path: ； The principles of game-theoretic guided decision-making include: The essence of game theory is an optimization process. First, the Hamiltonian function H is constructed as follows: ； In the formula, and These are the control acceleration vectors for the interceptor and the target, respectively. and These are the acceleration control weight matrices for the interceptor and the target, respectively. For terminal conditions, and These are the state coefficient matrix and the input coefficient matrix, respectively. It is a relative state vector. for These are the additional interference vectors for the interceptor and the target, respectively; According to the necessary conditions for optimization, the partial derivatives of the above equation have the following properties: ； ； ； Substituting the Hamiltonian function, we get: ； ； Moving the control input term to the left side of the equation yields the following form: ； ； Since the desired result is to construct the control inputs as a function of the system state, therefore... Set it to the following format: ； in, for The matrix is ​​a solution to the matrix Riccati differential equation; for Vector, which is subsequently indicated as a solution to the vector Riccati differential equation, matrix It is related to the game input, and It contains additional acceleration, Substituting the values: ； ； From the necessary conditions and the Hamiltonian function, we can obtain: ； Substitution The expression, after expansion and algebraic simplification, is obtained as follows: ； Because the above formula requires that for all Established, The coefficient of and the right side of the equation must be equal to zero, that is: ； ； To apply the mathematical model in practice, the weight matrix structure is defined as follows: And define the matrix : ； Let matrix The structure is as follows: ； The elements in the matrix have the following relationship: ； Make the following settings: ； ； ； ； ； ； The relevant parameters of game-theoretic guidance are derived from , , , , constitute; Record the remaining time , To solve for the deadline, According to MRDE, we can solve for: The feedback gain matrices for the pursuers and the escapees are as follows: ； ； The control input feedback functions for both the pursuer and the escapee are: ； ； Substitute the acceleration expression into the kinematic equations for game-theoretic guidance; The game-theoretic guidance calculation results are converted into desired attitude and throttle and then sent to the flight control stack. The process of converting the game-guided calculation results into the desired attitude and throttle includes: For the acceleration calculated by the game guidance, it is transformed from the fixed coordinate system to the carrier coordinate system according to the coordinate system transformation method; the desired roll angle is generated from the y-axis acceleration in the carrier coordinate system; the desired velocity and desired altitude are generated from the x-axis and z-axis acceleration in the carrier coordinate system, and then converted into the desired pitch angle and desired throttle after calculation by the total energy control system TECS. The inner loop of the flight control stack uses PID to convert the desired attitude and throttle into motor speed and servo control quantities to control the actions in the simulation or the actual device. The desired pitch angle, desired roll angle, and desired throttle are sent to the inner loop of the flight control stack via external program communication. The external program communicates with the flight control stack and each UAV flight control stack via ROS2. Each aircraft's flight control stack and the external program are different communication nodes, and each UAV belongs to a different domain.

2. The UAV decision-making and control method based on game-theoretic guidance according to claim 1, characterized in that, The information obtained about the individuals participating in the attack and defense includes the position and Euler angle information of all individuals in the pursuing and escaping groups.

3. The UAV decision-making and control method based on game-theoretic guidance according to claim 1, characterized in that, Before sending a message to other individuals, each individual needs to check its own status, and the scale of the cluster game is determined based on the number of drones sending and receiving messages in different clusters.

4. The UAV decision-making and control method based on game-theoretic guidance according to claim 1, characterized in that, The game-guided calculation based on the target matching result includes: outputting the target matching result to the UAV, and each UAV performing game-guided calculation based on its own and the state information of the matching object, or obtaining the acceleration of each UAV through centralized calculation and then sending instructions to each UAV to perform target matching and decision-making.

5. The UAV decision-making and control method based on game-theoretic guidance according to claim 1, characterized in that, The target matching is based on an auction mechanism. The pursuing group determines the ownership of each target drone through a bidding and auction cycle involving individuals within the escape group.

6. The game-theoretic guidance-based UAV decision-making and control method according to claim 5, characterized in that, The value of the escaping swarm of UAVs is determined using a value function based on the three-dimensional Dubins path length.

7. The UAV decision-making and control method based on game-theoretic guidance according to claim 1, characterized in that, The game-theoretic guidance calculation includes: calculating the three-axis acceleration of each UAV based on the pairing formed after each UAV is matched with the target using differential game-theoretic guidance, and the required information for the calculation includes the position, velocity and attitude of the UAV.

8. The UAV decision-making and control method based on game-theoretic guidance according to claim 1, characterized in that, The target matching is applied to the pursuit and escape of individuals among a maximum of three clusters. For the pursuit and escape of individual drones, no target matching process is required. For the attack and defense between two individuals, the target matching step is skipped directly.

Images