Unmanned aerial vehicle / ship asynchronous cooperative adaptive control method based on extreme learning machine

Abstract

The invention discloses an unmanned aerial vehicle / ship asynchronous cooperative adaptive control method based on an extreme learning machine, and the method comprises the steps: building a vector information consistent nonlinear mathematical model of an unmanned aerial vehicle / ship cooperative system, and further obtaining the relation between a control instruction and a force and a torque; obtaining a position / attitude error vector of the system based on the reference position / attitude vector; designing a virtual control law by using a backstepping method in combination with a conditional filter; defining a speed error vector according to the virtual control law, performing fitting through a continuous function separation technology and an extreme learning machine, and combining a Lyapunov stability theory to obtain an intermediate control law of a position loop / attitude loop; and through an intermediate control law of a position loop / attitude loop, a control instruction is solved so as to realize accurate and stable closed-loop control of the unmanned aerial vehicle / ship cooperative system. The method is oriented to different working conditions and navigation scenes, and accurate and rapid reference signal tracking of the aircraft-ship cooperative system can be ensured.

Description

Technical Field

[0001] This invention relates to the intersection of intelligent unmanned systems and marine engineering technology, and in particular to an asynchronous cooperative adaptive control method for unmanned aerial vehicles / ships based on extreme learning machines. Background Technology

[0002] The path tracking control of unmanned maritime devices consists of three subsystems: guidance, control, and navigation. The guidance system can automatically construct a heading reference signal based on the positional relationship between the ship's current attitude and the desired path. The control system can achieve effective convergence by stabilizing the error between the current attitude and the heading reference signal. The navigation system can transmit the position and attitude information of the controlled object to the guidance and control systems through sensors. In the control design, robust adaptive control algorithms based on radial basis function neural networks (RBFNNs) or fuzzy logic systems are generally used to solve the path tracking control problem of the cooperative body under conditions of unknown model parameters and marine environmental disturbances.

[0003] In existing research, to achieve accurate tracking of the reference path by a ship-machine cooperative system under uncertain model parameters and complex marine environments, a path tracking control strategy for the cooperative system is usually designed by integrating dynamic surface control with RBFNNs or fuzzy logic systems. Dynamic surface control is essentially a first-order filter, which effectively avoids the increased computational complexity caused by repeated differentiation in the traditional backstepping method. The introduction of RBFNNs or fuzzy logic systems is to estimate / compensate for uncertainties and external disturbances in the cooperative system in real time, so as to ensure the performance of the control system even when an accurate mathematical model is lacking.

[0004] However, existing ship-machine cooperative tracking control strategies have the following drawbacks: the time constant in existing dynamic surface control systems that address computational "complexity explosion" is fixed. A larger time constant can significantly enhance robustness to external disturbances and unmodeled dynamics, but the response speed is slow; a smaller time constant can improve the response speed, but it is susceptible to noise and model errors; existing RBFNNs or fuzzy logic systems rely on engineering experience or offline training to obtain Gaussian function parameters or membership function parameters. This strong parameter dependence not only limits their adaptive capabilities but may also lead to a decrease in control performance under unknown operating conditions due to the limited parameter coverage; while designing the adaptive law to simultaneously update position / shape parameters online can improve coverage and adaptive capabilities, it significantly increases the approximation speed and computational complexity. Summary of the Invention

[0005] This invention provides an asynchronous cooperative adaptive control method for UAVs / ships based on extreme learning machines, which overcomes the problem that existing systems cannot dynamically adjust the time constant and require relying on engineering experience or offline training to obtain parameters.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows: An asynchronous cooperative adaptive control method for unmanned aerial vehicles (UAVs) / unmanned vessels (USVs) based on extreme learning machines includes: S1. Establish a vector information consistency nonlinear mathematical model for the UAV / ship collaborative system; further obtain the relationship between control commands and forces and torques; S2. Obtain the reference position / attitude vector, and combine it with the vector information consistent nonlinear mathematical model of the UAV / ship cooperative system to obtain the position / attitude error vector; S3. Based on the position / attitude error vector, obtain the virtual control law through backstepping, and obtain the filtered virtual control law based on the conditional filter. S4. Define the velocity error vector based on the filtered virtual control law, and obtain the derivative of the velocity error vector by combining the vector information consistency nonlinear mathematical model of the UAV / ship cooperative system; based on the derivative of the velocity error vector, fit it using continuous function separation technology and extreme learning machine, and obtain the intermediate control law of the position loop / attitude loop by combining Lyapunov stability theory. S5. By using the intermediate control law of the position / attitude loop and combining the relationship between control commands and force and torque, control commands for the UAV / ship are obtained to achieve closed-loop control of the UAV / ship collaborative system.

[0007] Furthermore, the expression for the vector information consistency nonlinear mathematical model of the UAV / ship cooperative system is as follows: (1) In the formula, Represents the generalized position and attitude vectors of the system. ,in, Represents the ship's position and attitude state variables. These represent the ship's pitch, yaw, and bow angle, respectively. This represents the position and state variables of the drone. These represent the forward / backward displacement, left / right displacement, and vertical displacement of the drone, respectively. Let be the attitude state variables of the drone. These represent the pitch angle, roll angle, and yaw angle of the UAV, respectively. Represents the system's generalized velocity vector. ,in, Indicates the speed of the ship. These represent the ship's forward speed, drift speed, and bow angular velocity, respectively. This indicates the linear velocity of the drone. These represent the linear velocities of the UAV in the forward / backward, left / right, and vertical directions, respectively. This indicates the angular velocity of the drone. These represent the pitch rate, roll rate, and yaw rate of the UAV, respectively. Let be the quality matrix, where, It is the ship's added mass matrix. Both refer to the mass of the ship's appendages. and It is the mass and moment of inertia matrix of the drone. This represents the reciprocal of the drone's airframe mass. This indicates the mass of the drone's airframe. Both represent the moment of inertia of the drone; ,in d It is the diagonal diameter of the drone; It is a vector containing force and torque, i.e., the actual control input, where , These represent the ship's forward force and bow moment, respectively. , This represents the total lift generated by the four propellers of the drone. , These represent the pitch moment, roll moment, and yaw moment of the UAV, respectively. Let represent a rotation matrix, and I It is an identity matrix. For the transformation matrix of ships, For the transformation matrix related to the drone; It is a generalized nonlinear term; These are parameters used to describe external environmental disturbances; in, The expression is: (2) The expression is: (3) Generalized nonlinear terms The expressions for each component are as follows: (4) (5) (6) in, (7) In the formula, Represents the Coriolis centripetal force matrix; This represents the nonlinear resistance coefficient of a ship. Index indicating the damping coefficient; Indicates the design parameters of the drone; This represents the total rotor speed of the drone, and These are the rotational speeds corresponding to the four rotors of the drone.

[0008] Furthermore, based on the force and torque vectors and the corresponding rotational speeds of the four rotors of the UAV, combined with the X-shaped layout of the four rotors, the relationship between control commands and force and torque is obtained, expressed as: (8) In the formula, Represents the control commands of the collaborative system, and ,in, and These represent the ship's main engine speed and rudder angle, respectively. The gain matrix is ​​expressed as: (9) In the formula, This represents the actuator gain coefficient of the cooperative system.

[0009] Furthermore, the expression for the position / attitude error vector is: (10) In the formula, Represents the position / attitude error vector; Represents the reference position / attitude vector; Taking the derivative of equation (10), we obtain the derivative of the position / attitude error vector, which is expressed as: (11).

[0010] Furthermore, based on the position / attitude error vector, the virtual control law is obtained through backstepping, and its expression is: (12) In the formula, Represents a virtual control law; This represents the positive definite diagonal matrix in the virtual control law.

[0011] Furthermore, the expression for the conditional filter is: (13) In the formula, This represents the filtered virtual control law; This represents the derivative vector of the filtered virtual control law; This represents the initial virtual control law; This represents the initial filtered virtual control law; Let be a coefficient matrix, where The conditional filter coefficients are expressed as follows: (14) In the formula, Indicates the attenuation coefficient; Indicates avoidance Small positive numbers that exhibit singular values; Indicates runtime; express exist The conditional filter coefficients calculated at time, where, This represents a manually set time constant; Define the filtering error as: (15) In the formula, Indicates the filtering error; Combining equation (15) with equation (13), we obtain the derivative of the filtering error, expressed as: (16) In the formula, Description The matrix of continuous functions of derivatives, where, ,and Let be a known positive constant.

[0012] Furthermore, the specific steps for obtaining the intermediate control law and adaptive gain law of the position / attitude loop include: S41. Define the speed error vector according to the filtered virtual control law, and its expression is: (17) In the formula, This is the velocity error vector; S42. Taking the derivative of equation (17) and combining it with the vector information consistency nonlinear mathematical model of the UAV / ship cooperative system, the derivative of the velocity error vector is obtained, and its expression is: (18) S43. Using extreme learning machine and continuous function separation techniques, the generalized nonlinear term in equation (18) is processed. To perform the fitting, the expression is: (19) In the formula, and This represents the intermediate design vector, and ; This represents the hidden layer output matrix; express The real number space; Indicates the number of hidden layer nodes; This represents the output weight matrix; express The real number space; This represents the fitting error of the Extreme Learning Machine; S44. Substituting equation (19) into equation (18), we obtain the following expression: (20) In the formula, This represents the intermediate design vector, and ; make To output the Euclidean norm of each element in the hidden layer's output matrix, To obtain the Euclidean norm of each element in the output weight matrix, we get The upper bound estimate is expressed as: (twenty one) In the formula, They represent and D A vector formed by the unknown upper bounds of each component; express and The maximum value of each element in each row; ,in, Represents the identity matrix; S45. Based on equations (20) and (21), and combined with Lyapunov stability theory, the intermediate control law of the position / attitude loop is obtained, and its expression is: (twenty two) in, (twenty three) In the formula, This represents the intermediate control law of the position / attitude loop; This represents the adaptive gain law; Represents the positive definite diagonal matrix in the intermediate control law of the position / attitude loop; Represents positive numbers; Let represent the equivalent velocity error matrix, and ,in, .

[0013] Furthermore, by using the intermediate control law of the position / attitude loop and combining it with the relationship between control commands and forces and torques, the specific steps to obtain the control commands for the UAV / ship include: S51. Define the intermediate control law for the position / attitude loop as follows: (twenty four) In the formula, ,in, These represent the equivalent components of the ship's forward and backward motion, left and right motion, and bow moment, respectively. ,in, These represent the equivalent components of the UAV's forward and backward motion, left and right motion, and vertical motion, respectively. ,in, These represent the equivalent pitch moment, roll moment, and yaw moment of the UAV, respectively. S52. Based on the intermediate control law and the actual control input, the relationship between the intermediate control law and the actual control input is obtained, expressed as: (25); S53. Based on the relationship between the intermediate control law and the actual control input, and combined with the relationship between the control command and force and torque, the inverse solution of the relationship between the control command and force and torque is obtained, and the expression is: (26) The control commands of the cooperative system obtained from equation (26) Receive ship main engine speed control command and rudder angle control commands and the control commands for the four rotors of the drone. .

[0014] Beneficial effects: This invention provides an asynchronous cooperative adaptive control method for UAVs / ships based on extreme learning machines. By designing a conditional filter, it effectively reduces computational complexity and solves the problem that existing dynamic surface control methods struggle to balance stability and speed at different stages of system operation. Furthermore, it adaptively adjusts the conditional coefficients according to the system state, enhancing robustness and suppressing overshoot in the initial stage, and improving response speed in the steady-state stage, thus achieving global performance optimization from transient to steady state. By combining extreme learning machines with continuous function separation techniques, this design can effectively estimate model uncertainties and external time-varying disturbances online, avoiding the cumbersome offline parameter training process in traditional methods (such as RBFNNs or fuzzy logic systems). This design, with its fixed random feature structure, quickly and easily improves the system's online adaptability, making it suitable for real-time control applications with limited computing and storage resources. Attached Figure Description

[0015] 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 some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart of the asynchronous cooperative adaptive control method for UAVs / ships based on extreme learning machines according to the present invention; Figure 2 This is a diagram illustrating the variables of the UAV / ship collaborative system in an embodiment of the present invention; Figure 3 This is a block diagram of the UAV / ship cooperative system tracking and control method in an embodiment of the present invention; Figure 4 This is a graph showing the parameter variation trend of the conditional filter in an embodiment of the present invention; Figure 5 This is a three-dimensional view of the reference / tracking path of the UAV / ship collaborative system for inspecting offshore wind turbine pods in an embodiment of the present invention; Figure 6 This is a top view of the reference / tracking path of the UAV / ship collaborative system for inspecting offshore wind turbine pods in an embodiment of the present invention; Figure 7 This is a diagram showing the position and heading error of the UAV / ship cooperative system in an embodiment of the present invention; Figure 8 This is a control command diagram of the UAV / ship collaborative system in an embodiment of the present invention; Figure 9 This is a graph showing the forward speed variation of the UAV / ship cooperative system in an embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0018] This embodiment provides an asynchronous cooperative adaptive control method for unmanned aerial vehicles (UAVs) / ships based on extreme learning machines, such as... Figure 1 As shown, it includes: S1. Establish a vector information consistency nonlinear mathematical model for the UAV / ship collaborative system; further obtain the relationship between control commands and forces and torques; S2. Obtain the reference position / attitude vector, and combine it with the vector information consistent nonlinear mathematical model of the UAV / ship cooperative system to obtain the position / attitude error vector; S3. Based on the position / attitude error vector, obtain the virtual control law through backstepping, and obtain the filtered virtual control law based on the conditional filter. S4. Define the velocity error vector based on the filtered virtual control law, and obtain the derivative of the velocity error vector by combining the vector information consistency nonlinear mathematical model of the UAV / ship cooperative system; based on the derivative of the velocity error vector, fit it using continuous function separation technology and extreme learning machine, and obtain the intermediate control law of the position loop / attitude loop by combining Lyapunov stability theory. S5. By using the intermediate control law of the position / attitude loop and combining the relationship between control commands and force and torque, control commands for the UAV / ship are obtained to achieve closed-loop control of the UAV / ship collaborative system.

[0019] Preferred, according to Figure 2 The variables of the UAV / ship cooperative system are shown. A nonlinear mathematical model for the vector information consistency of the UAV / ship cooperative system is established; in the figure, Using the geodetic coordinate system, and These are the attached coordinate systems for the ship and the drone, respectively; The expression for the vector information consistency nonlinear mathematical model of the UAV / ship cooperative system is as follows: (1) In the formula, Represents the generalized position and attitude vectors of the system. ,in, Represents the ship's position and attitude state variables. These represent the ship's pitch, yaw, and bow angle, respectively. This represents the position and state variables of the drone. These represent the forward / backward displacement, left / right displacement, and vertical displacement of the drone, respectively. Let be the attitude state variables of the drone. These represent the pitch angle, roll angle, and yaw angle of the UAV, respectively. Represents the system's generalized velocity vector. ,in, Indicates the speed of the ship. These represent the ship's forward speed, drift speed, and bow angular velocity, respectively. This indicates the linear velocity of the drone. These represent the linear velocities of the UAV in the forward / backward, left / right, and vertical directions, respectively. This indicates the angular velocity of the drone. These represent the pitch rate, roll rate, and yaw rate of the UAV, respectively. Let be the quality matrix, where, It is the ship's added mass matrix. Both refer to the mass of the ship's appendages. and It is the mass and moment of inertia matrix of the drone. This represents the reciprocal of the drone's airframe mass. This indicates the mass of the drone's airframe. These represent the moment of inertia of the drone; ,in d It is the diagonal diameter of the drone; It is a vector containing force and torque, i.e., the actual control input, where , These represent the ship's forward force and bow moment, respectively. , This represents the total lift generated by the four propellers of the drone. , These represent the pitch moment, roll moment, and yaw moment of the UAV, respectively. Let represent a rotation matrix, and I It is an identity matrix. For the transformation matrix of ships, For the transformation matrix related to the drone; It is a generalized nonlinear term; These are parameters used to describe external environmental disturbances; in, The expression is: (2) The expression is: (3) Generalized nonlinear terms The expressions for each component are as follows: (4) (5) (6) in, (7) In the formula, Represents the Coriolis centripetal force matrix; This represents the nonlinear resistance coefficient of a ship. Index indicating the damping coefficient; Indicates the design parameters of the drone; This represents the total rotor speed of the drone, and These are the rotational speeds corresponding to the four rotors of the drone.

[0020] Preferably, based on the force and torque vectors and the corresponding rotational speeds of the four rotors of the UAV, combined with the X-shaped layout of the four rotors of the UAV, the relationship between the control commands and the force and torque is obtained, expressed as: (8) In the formula, Represents the control commands of the collaborative system, and ,in, and These represent the ship's main engine speed and rudder angle, respectively. The gain matrix is ​​expressed as: (9) In the formula, This represents the actuator gain coefficient of the cooperative system.

[0021] In a specific embodiment, the design process of the UAV / ship cooperative system tracking method is as follows: Figure 3 As shown, the corresponding steps are S2 to S5.

[0022] Preferably, the expression for the position / attitude error vector is: (10) In the formula, Represents the position / attitude error vector; Represents the reference position / attitude vector; Taking the derivative of equation (10), we obtain the derivative of the position / attitude error vector, which is expressed as: (11).

[0023] Preferably, based on the position / attitude error vector, the virtual control law is obtained through backstepping, and its expression is: (12) In the formula, Represents a virtual control law; This represents a positive definite diagonal matrix.

[0024] Preferably, a conditional filter is designed to address the increased computational complexity caused by virtual control laws; The expression for the conditional filter is: (13) In the formula, This represents the filtered virtual control law; This represents the derivative vector of the filtered virtual control law; This represents the initial virtual control law; This represents the initial filtered virtual control law; Let be a coefficient matrix, where The conditional filter coefficients are expressed as follows: (14) In the formula, Indicates the attenuation coefficient; Indicates avoidance Small positive numbers that exhibit singular values; Indicates runtime; express exist The conditional filter coefficients calculated at time, where, This represents a manually set time constant; Define the filtering error as: (15) In the formula, Indicates the filtering error; Combining equation (15) with equation (13), we obtain the derivative of the filtering error, expressed as: (16) In the formula, Description The matrix of continuous functions of derivatives, where, ,and Let be a known positive constant.

[0025] Preferably, the specific steps for obtaining the intermediate control law and adaptive gain law of the position / attitude loop include: S41. Define the speed error vector according to the filtered virtual control law, and its expression is: (17) In the formula, This is the velocity error vector; S42. Taking the derivative of equation (17) and combining it with the vector information consistency nonlinear mathematical model of the UAV / ship cooperative system, the derivative of the velocity error vector is obtained, and its expression is: (18) S43. Using extreme learning machine and continuous function separation techniques, the generalized nonlinear term in equation (18) is processed. To perform the fitting, the expression is: (19) In the formula, and This represents the intermediate design vector, and ; This represents the hidden layer output matrix; express The real number space; Indicates the number of hidden layer nodes; This represents the output weight matrix; express The real number space; This represents the fitting error of the Extreme Learning Machine; S44. Substituting equation (19) into equation (18), we obtain the following expression: (20) In the formula, This represents the intermediate design vector, and ; make To output the Euclidean norm of each element in the hidden layer's output matrix, To obtain the Euclidean norm of each element in the output weight matrix, we get The upper bound estimate is expressed as: (twenty one) In the formula, They represent andD A vector formed by the unknown upper bounds of each component; express and The maximum value of each element in each row; ,in, Represents the identity matrix; S45. Based on equations (20) and (21), and combined with Lyapunov stability theory, the intermediate control law of the position / attitude loop is obtained, and its expression is: (twenty two) in, (twenty three) In the formula, This represents the intermediate control law of the position / attitude loop; This represents the adaptive gain law; Represents a positive definite diagonal matrix; Represents positive numbers; Let represent the equivalent velocity error matrix, and ,in, .

[0026] Preferably, the specific steps for obtaining the control commands for the UAV / ship by using the intermediate control law of the position / attitude loop and combining the relationship between control commands and forces and torques include: S51. Define the intermediate control law for the position / attitude loop as follows: (twenty four) In the formula, ,in, These represent the equivalent components of the ship's forward and backward motion, left and right motion, and bow moment, respectively. ,in, These represent the equivalent components of the UAV's forward and backward motion, left and right motion, and vertical motion, respectively. ,in, These represent the equivalent pitch moment, roll moment, and yaw moment of the UAV, respectively. S52. Based on the intermediate control law and the actual control input, the relationship between the intermediate control law and the actual control input is obtained, expressed as: (25); S53. Based on the relationship between the intermediate control law and the actual control input, and combined with the relationship between the control command and force and torque, the inverse solution of the relationship between the control command and force and torque is obtained, and the expression is: (26) The control commands of the cooperative system obtained from equation (26) Receive ship main engine speed control command and rudder angle control commands and the control commands for the four rotors of the drone. .

[0027] In this embodiment, the control design of the ship's position loop and attitude loop is carried out simultaneously; the UAV needs to design the control of the position loop first, and then design the control of the attitude loop.

[0028] Specifically, through intermediate control law components The pitch and roll reference angles are calculated using the following expressions: (27) In the formula, This is the roll reference angle; This is the pitch reference angle; This is the reference heading angle for the drone.

[0029] In this embodiment, in order to verify the effectiveness of the cooperative path tracking control method proposed in this invention, the wind turbine pod inspection task of an offshore wind farm is designed as a simulation case, and numerical simulation verification is carried out based on the MATLAB platform. The wind farm is equipped with four identical wind turbines. Each wind turbine has a blade radius of 100 meters and a hub height of 150 meters. The east-west distance between adjacent wind turbines is 600 meters. The coordinates of the first wind turbine are (600m, 820m). The prevailing wind direction in the sea area where the wind farm is located is southerly, and the sea state is level 4. The external environmental disturbances include two types of disturbances: sea wind and sea waves, which are generated by the NORSOK wind spectrum and the JONSWAP wave spectrum, respectively. The sea wind / wave disturbances affect ships, while the sea wind disturbances only affect drones. The waypoints for the ships are set as follows: The waypoints for the drone are set as follows: Among them, point set It is based on the equidistant selection of coordinates of four wind turbines, point set The waypoints for the first wind turbine are set as {(400m, 840m), (600m, 940m), (800m, 840m)}; To ensure effective inspection of wind turbine nacelles, the drone's cruising altitude was set at 150 meters, and the ship's planned speed was set at 7 knots (approximately 3.60 meters per second). The position and heading reference signals of each drone / ship in the drone / ship collaborative system were generated through the set waypoints. Configure parameters and select the attenuation coefficient of the conditional filter. Design coefficient ,time The parameter changes in the first 200 seconds are as follows Figure 4 As shown; the number of hidden layer nodes in the Extreme Learning Machine is set to 25; a USV with a length of 38 meters and a UAV with a diagonal diameter of 0.8 meters are selected as simulation test objects; The initial state is set, with the USV at position (-20m, 0m) and all linear and angular velocities at 0. The UAV is docked on the ship's docking platform.

[0030] Based on the aforementioned simulation conditions, the path tracking control results of the ship-machine cooperative system for inspecting offshore wind turbine pods using conditional filters and extreme learning machines are as follows: Figures 5 to 9 As shown; according to Figure 5 and Figure 6 The UAV / ship cooperative system effectively tracked the desired path within 1064 seconds, significantly suppressed the overshoot phenomenon in the initial stage, and effectively compensated for model uncertainty and external disturbances. Figure 7 The study demonstrates the changes in position and heading errors of UAVs and ships. The tracking errors of ships in the x and y axes are complementary due to the influence of sea wind / wave disturbances, while the pose error fluctuations of UAVs are smaller than those of ships. Figure 8 This illustrates the changes in control commands for a drone / ship collaborative system. The first two graphs correspond to the main engine speed and rudder angle of a ship, respectively, while the third graph corresponds to the average change in the four rotors of the drone. When the ship is traversing the prevailing wind direction, the rudder angle fluctuation is more significant; when the ship is parallel to the prevailing wind direction, the main engine speed fluctuation is more significant. Figure 9 The study shows the changing trends of the linear velocity of ships and drones in their forward direction. The ship's forward speed fluctuates slightly around 3.5 m / s when crossing the prevailing wind direction, while the drone's forward speed fluctuates around 3.8 m / s during operation.

[0031] The present invention has the following beneficial effects: The present invention provides an asynchronous cooperative adaptive control method for UAVs / ships based on extreme learning machines. Through the design of conditional filters, it solves the problem that existing dynamic surface control is difficult to balance stability and speed at different stages of system operation while effectively reducing computational complexity. Furthermore, it adaptively adjusts the conditional coefficients according to the system state, enhancing robustness and suppressing overshoot in the initial stage, and improving response speed in the steady-state stage, thus achieving global performance optimization from transient to steady state. By combining extreme learning machines with continuous function separation techniques, this design can effectively estimate model uncertainties and external time-varying disturbances online, avoiding the cumbersome offline parameter training process in traditional methods (such as RBFNNs or fuzzy logic systems). This design, with its fixed random feature structure, quickly and easily improves the system's online adaptability, making it suitable for real-time control applications with limited computing and storage resources.

[0032] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for asynchronous cooperative adaptive control of unmanned aerial vehicles / ships based on extreme learning machines, characterized in that, include: S1. Establish a vector information consistency nonlinear mathematical model for the UAV / ship collaborative system; further obtain the relationship between control commands and forces and torques; S2. Obtain the reference position / attitude vector, and combine it with the vector information consistent nonlinear mathematical model of the UAV / ship cooperative system to obtain the position / attitude error vector; S3. Based on the position / attitude error vector, obtain the virtual control law through backstepping, and obtain the filtered virtual control law based on the conditional filter. S4. Define the velocity error vector based on the filtered virtual control law, and obtain the derivative of the velocity error vector by combining the vector information consistency nonlinear mathematical model of the UAV / ship cooperative system; based on the derivative of the velocity error vector, fit it using continuous function separation technology and extreme learning machine, and obtain the intermediate control law of the position loop / attitude loop by combining Lyapunov stability theory. S5. By using the intermediate control law of the position / attitude loop and combining the relationship between control commands and force and torque, control commands for the UAV / ship are obtained to achieve closed-loop control of the UAV / ship collaborative system.

2. The asynchronous cooperative adaptive control method for UAVs / ships based on extreme learning machines according to claim 1, characterized in that, The expression for the vector information consistency nonlinear mathematical model of the UAV / ship cooperative system is as follows: （1） In the formula, Represents the generalized position and attitude vectors of the system. ,in, Represents the ship's position and attitude state variables. These represent the ship's pitch, yaw, and bow angle, respectively. This represents the position and state variables of the drone. These represent the forward / backward displacement, left / right displacement, and vertical displacement of the drone, respectively. Let be the attitude state variables of the drone. These represent the pitch angle, roll angle, and yaw angle of the UAV, respectively. Represents the system's generalized velocity vector. ,in, Indicates the speed of the ship. These represent the ship's forward speed, drift speed, and bow angular velocity, respectively. This indicates the linear velocity of the drone. These represent the linear velocities of the UAV in the forward / backward, left / right, and vertical directions, respectively. This indicates the angular velocity of the drone. These represent the pitch rate, roll rate, and yaw rate of the UAV, respectively. Let be the quality matrix, where, It is the ship's added mass matrix. Both refer to the mass of the ship's appendages. and It is the mass and moment of inertia matrix of the drone. This represents the reciprocal of the drone's airframe mass. This indicates the mass of the drone's airframe. Both represent the moment of inertia of the drone; ,in d It is the diagonal diameter of the drone; It is a vector containing force and torque, i.e., the actual control input, where , These represent the ship's forward force and bow moment, respectively. , This represents the total lift generated by the four propellers of the drone. , These represent the pitch moment, roll moment, and yaw moment of the UAV, respectively. Let represent a rotation matrix, and I It is an identity matrix. For the transformation matrix of ships, For the transformation matrix related to the drone; It is a generalized nonlinear term; These are parameters used to describe external environmental disturbances; in, The expression is: （2） The expression is: （3） Generalized nonlinear terms The expressions for each component are as follows: （4） （5） （6） in, （7） In the formula, Represents the Coriolis centripetal force matrix; This represents the nonlinear resistance coefficient of a ship. Index indicating the damping coefficient; Indicates the design parameters of the drone; This represents the total rotor speed of the drone, and These are the rotational speeds corresponding to the four rotors of the drone.

3. The asynchronous cooperative adaptive control method for UAVs / ships based on extreme learning machines according to claim 2, characterized in that, Based on the force and torque vectors and the corresponding rotational speeds of the four rotors of the UAV, combined with the X-shaped layout of the four rotors, the relationship between control commands and force and torque is obtained, expressed as: （8） In the formula, Represents the control commands of the collaborative system, and ,in, and These represent the ship's main engine speed and rudder angle, respectively. The gain matrix is ​​expressed as: （9） In the formula, This represents the actuator gain coefficient of the cooperative system.

4. The asynchronous cooperative adaptive control method for UAVs / ships based on extreme learning machines according to claim 1, characterized in that, The expression for the position / attitude error vector is: （10） In the formula, Represents the position / attitude error vector; Represents the reference position / attitude vector; Taking the derivative of equation (10), we obtain the derivative of the position / attitude error vector, which is expressed as: （11）。 5. The asynchronous cooperative adaptive control method for UAVs / ships based on extreme learning machines according to claim 4, characterized in that, Based on the position / attitude error vector, the virtual control law is obtained through backstepping, and its expression is: （12） In the formula, Represents a virtual control law; This represents the positive definite diagonal matrix in the virtual control law.

6. The asynchronous cooperative adaptive control method for UAVs / ships based on extreme learning machines according to claim 1, characterized in that, The expression for the conditional filter is: （13） In the formula, This represents the filtered virtual control law; This represents the derivative vector of the filtered virtual control law; This represents the initial virtual control law; This represents the initial filtered virtual control law; Let be a coefficient matrix, where The conditional filter coefficients are expressed as follows: （14） In the formula, Indicates the attenuation coefficient; Indicates avoidance Small positive numbers that exhibit singular values; Indicates runtime; express exist The conditional filter coefficients calculated at time, where, This represents a manually set time constant; Define the filtering error as: （15） In the formula, Indicates the filtering error; Combining equation (15) with equation (13), we obtain the derivative of the filtering error, expressed as: （16） In the formula, Description The matrix of continuous functions of derivatives, where, ,and Let be a known positive constant.

7. The asynchronous cooperative adaptive control method for UAVs / ships based on extreme learning machine according to claim 3, characterized in that, The specific steps to obtain the intermediate control law and adaptive gain law of the position / attitude loop include: S41. Define the speed error vector according to the filtered virtual control law, and its expression is: （17） In the formula, This is the velocity error vector; S42. Taking the derivative of equation (17) and combining it with the vector information consistency nonlinear mathematical model of the UAV / ship cooperative system, the derivative of the velocity error vector is obtained, and its expression is: （18） S43. Using extreme learning machine and continuous function separation techniques, the generalized nonlinear term in equation (18) is processed. To perform the fitting, the expression is: （19） In the formula, and This represents the intermediate design vector, and ; This represents the hidden layer output matrix; express The real number space; Indicates the number of hidden layer nodes; This represents the output weight matrix; express The real number space; This represents the fitting error of the Extreme Learning Machine; S44. Substituting equation (19) into equation (18), we obtain the following expression: （20） In the formula, This represents the intermediate design vector, and ; make To output the Euclidean norm of each element in the hidden layer's output matrix, To obtain the Euclidean norm of each element in the output weight matrix, we get The upper bound estimate is expressed as: （21） In the formula, They represent and D A vector formed by the unknown upper bounds of each component; express and The maximum value of each element in each row; ,in, Represents the identity matrix; S45. Based on equations (20) and (21), and combined with Lyapunov stability theory, the intermediate control law of the position / attitude loop is obtained, and its expression is: （22） in, （23） In the formula, This represents the intermediate control law of the position / attitude loop; This represents the adaptive gain law; Represents the positive definite diagonal matrix in the intermediate control law of the position / attitude loop; Represents positive numbers; Let represent the equivalent velocity error matrix, and ,in, .

8. The asynchronous cooperative adaptive control method for UAVs / ships based on extreme learning machines according to claim 7, characterized in that, The specific steps for obtaining control commands for UAVs / ships by using the intermediate control law of the position / attitude loop and combining the relationship between control commands and forces and torques include: S51. Define the intermediate control law for the position / attitude loop as follows: （24） In the formula, ,in, These represent the equivalent components of the ship's forward and backward motion, left and right motion, and bow moment, respectively. ,in, These represent the equivalent components of the UAV's forward and backward motion, left and right motion, and vertical motion, respectively. ,in, These represent the equivalent pitch moment, roll moment, and yaw moment of the UAV, respectively. S52. Based on the intermediate control law and the actual control input, the relationship between the intermediate control law and the actual control input is obtained, expressed as: （25）； S53. Based on the relationship between the intermediate control law and the actual control input, and combined with the relationship between the control command and force and torque, the inverse solution of the relationship between the control command and force and torque is obtained, and the expression is: （26） The control commands of the cooperative system obtained from equation (26) Receive ship main engine speed control command and rudder angle control commands and the control commands for the four rotors of the drone. .

Images