Unmanned aerial vehicle adaptive visual servo control method based on second-order cone constraint

Abstract

The invention belongs to the technical field of unmanned aerial vehicle visual servo control, and particularly discloses an unmanned aerial vehicle self-adaptive visual servo control method based on second-order cone constraint, and the method comprises the steps: building a four-rotor kinetic model, and building an inertial coordinate system and a vehicle body coordinate system; a virtual image plane thought is introduced, a visual servo model in a virtual image plane is deduced, and an image point dynamic model which does not change along with the posture is obtained; defining image moment characteristics and deducing a decoupling image moment characteristic dynamic model; the difference between the image features and the expected image features is used as input, the target is restrained to be within the view range of the camera through a second-order cone rule, and the restrained features are converged to the expected position with the shortest path; and finally, selecting a reference servo gain value according to the size of the image feature error, and adjusting the adaptive servo gain according to the size of the image feature error. The method can enable the control system to converge rapidly and stably, and effectively guarantees that the image features are always in the field of view.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle (UAV) visual servo control, and in particular to an adaptive UAV visual servo control method based on second-order cone constraints. Background Art

[0002] As a core method for autonomous navigation and target tracking in unmanned aerial vehicles (UAVs), visual servo control technology generates control commands by processing image information in real time. In recent years, it has shown broad application prospects in fields such as agricultural inspection, disaster relief, and industrial inspection. Due to the strong coupling and under-actuation characteristics of quadrotor UAVs, a single control output can affect multiple degrees of freedom, increasing the difficulty of control system design. Image-based visual servoing (IBVS) directly uses image feature errors on the image plane to derive the control input for controlling the robot. It does not require the restoration of three-dimensional position information, is robust to camera parameters, and can handle some constraints of under-actuated systems. However, when the UAV undergoes drastic attitude changes (such as non-zero pitch and roll angles), image features are prone to nonlinear distortion, resulting in feature tracking failure or a sudden drop in convergence speed.

[0003] To overcome the impact of attitude disturbances on IBVS, existing technologies attempt to introduce inertial measurement unit (IMU) data or pose estimators to dynamically compensate for image features. For example, a virtual camera coordinate system is constructed through Euler angle decoupling to partially eliminate the influence of attitude angles. However, this method only designs compensation strategies for a single degree of freedom (such as yaw angle) and lacks systematic modeling of the coupling effects of pitch and roll angles, making it difficult to completely eliminate virtual perspective projection errors. In addition, existing methods often use the coordinates of a single feature point as feedback, resulting in low-dimensional feature information and weak anti-occlusion capabilities. In complex scenarios, control divergence is easily caused by feature loss. Existing methods still fail to effectively address the problem of nonlinear distortion of image features.

[0004] The existing UAV visual servo control technology still has the following key problems: 1) There is a lack of effective compensation mechanism for image feature distortion caused by changes in the pitch and roll angles of the UAV; 2) The dynamic characteristics of the image moment features are not deeply coupled with the UAV kinematic and dynamic models; 3) The control law design does not systematically integrate fast convergence and robustness optimization under multiple constraints.

[0005] To address the above defects, it is urgent to propose a new visual servo control architecture and method that integrates virtual perspective compensation, high-dimensional feature dynamics modeling and nonlinear optimization constraints to improve the control performance in complex disturbance scenarios. Summary of the Invention

[0006] In order to solve the problems existing in the prior art, the present invention provides a UAV adaptive visual servo control method based on second-order cone constraints, which solves the problems mentioned in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solution: a method for adaptive visual servo control of a UAV based on second-order cone constraints, comprising the following steps: S1. Build a quadrotor dynamics model and establish the inertial coordinate system and the quadrotor's own body coordinate system; S2. The idea of ​​virtual image plane is introduced, and the visual servoing model in the virtual image plane is derived to obtain the image point dynamics model that does not change with the posture; S3. Define image moment features and derive a decoupled image moment feature dynamics model; S4, taking the difference between the image feature S and the desired image feature S* as input, servo-controls the drone, and constrains the target to be within the camera's field of view using the second-order cone rule. The constraint feature is to converge to the desired position along the shortest path. S5. Finally, a reference servo gain value is selected according to the size of the image feature error, and the adaptive servo gain is adjusted according to the size of the image feature error.

[0008] Preferably, in step S1, a quadrotor dynamics model is constructed to establish an inertial coordinate system and a quadrotor body coordinate system, specifically including the following: S11. Build a quadrotor dynamics model. The rotational speeds of the four rotors are: By increasing the speed of the specified rotor, the quadrotor will generate a corresponding torque , the lift of the quadrotor and torque It is obtained by the following expression: ; in, is the lift coefficient, is the drag coefficient, is the distance between the position of each rotor center and the center of mass of the quadrotor drone; S12. Establish an inertial coordinate system according to the right-hand rule: ,in is the origin of the inertial coordinate system, Pointing north, Pointing east, Pointing vertically downward to the center of the earth; the body coordinate system is established as: , is the origin of the body coordinate system, Pointing to the front of the body, Point to the right, vertically downward; S13. The roll angle, pitch angle and yaw angle of the quadrotor drone are: , the rotation matrix The rotation matrix is ​​also composed of these three Euler angles. It shows the degree of change of the body coordinate system relative to the inertial coordinate system. It is based on the rotation around the coordinate axis. Rotate in sequence Obtained by angle transformation; its specific form is: ; in: .

[0009] Preferably, in step S2, the idea of ​​a virtual image plane is introduced, a visual servoing model in the virtual image plane is derived, and an image point dynamics model that does not change with the posture is obtained, which specifically includes the following: S21. Assume a stationary point The coordinates in the inertial coordinate system are , whose coordinates in the camera system are By defining a virtual camera, the roll and pitch angles of the virtual camera are always zero. At the same time, a dynamic rotation matrix based on the attitude angle change rate is introduced to generate a compensation matrix according to the real-time attitude angle of the drone, and the actual image features are projected onto the dynamically adjusted virtual image plane. S22. Define the coordinates of point P in the virtual camera coordinate system , and its coordinate relationship with point P in the inertial coordinate system is: ,in are the coordinates of point p, represents the virtual image plane, is the coordinate of the camera in the virtual camera coordinate system, is the rotation matrix in the yaw direction, yes Transpose the matrix. By using the perspective projection theorem, the coordinates of point P on the virtual image plane are: ; represents the coordinates of point P on the virtual image plane, is the focal length of the camera; S23, deriving the coordinates of step 22, we can obtain: ; in Yes The derivative with respect to time, is the derivative of point P with respect to time; Written in matrix form: ; in is the yaw angle The derivative of The relationship between the rate of change of a point on the virtual image plane and the quadrotor speed is obtained, namely the image point dynamics model.

[0010] Preferably, in step S3, defining image moment features and deriving a decoupled image moment feature dynamics model specifically includes the following: S31. Define three image moment features related to the projection coordinates of a specified point on the virtual image plane for controlling the three-dimensional translation motion of the quadrotor: ; Image features The coordinates on the virtual image plane are , , composed of functions; where: ; Indicates the selection of the visual target point for research. ,and , is the relative distance between the quadrotor drone and the target, yes expected value; express Any number within the value range, represents the virtual image plane; In the virtual image plane there are: ; in, Indicates the depth of the feature point relative to the virtual camera, yes expected value; S32. Derivative the image moment feature and combine it with the visual servo model obtained in step S2 to obtain a dynamic model of the image moment feature: ; The decoupling between image motion and the pitch and roll motion of the drone is achieved; sk ( ) is an antisymmetric matrix, is the linear velocity of the virtual camera in the virtual camera coordinate system; ; yes The partial derivative of is the derivative of the yaw angle; S33. Define the image moment feature for controlling the yaw angle: ; Its derivative is: , therefore, the image features are expanded into: , the dynamic model of image moment features is expressed as: .

[0011] Preferably, in step S4, the following is specifically included: S41, define the expected image features as , using image moment features to construct image error , taking the derivative of the above formula, we get ,in is the image Jacobian matrix of the image features, is the motion state of the virtual camera; from step 3, we can know that: ; in, is the depth of the feature point relative to the virtual camera, yes expected value; is the image feature; is the yaw angle The derivative of S42, construct on 2D image plane is the image feature at the next moment To the current image features and expected image features Distance of connection line: ; because , so the above formula can be written as: ; make , by deducing ;in is the identity matrix of the corresponding dimension; yes The transpose of Control rate based on image feedback: , is the sampling interval; if ,but exist and On the connection line; S43. Constrain the target to be within the camera's field of view through the second-order cone rule, and the constraint feature is to converge to the desired position by the shortest path.

[0012] Preferably, in step S43, the second-order cone rule constraint conditions include three constraints: error convergence constraint, distance constraint, and physical constraint; specifically, they are as follows: 1) To ensure that the error converges to 0 at an exponential rate at the next moment, the error convergence constraint is: ,in is the convergence coefficient; 2) Define distance constraints to ensure tends to 0, and the distance constraint is: ,in: is a projection matrix, is an auxiliary variable used to represent The upper bound of 3) Define the physical constraints of the control input: ,in and are the lower and upper bounds of the control input.

[0013] Preferably, in step S5, the following is specifically included: From step 4, we can see that the derivative of the error with respect to time is: ; make ,So: ; in, is the image Jacobian matrix of the image features, is the motion state of the virtual camera; is the generalized inverse of the Jacobian matrix, is the servo gain, defining a reference value , set the servo gain that is adaptively adjusted according to the size of the image feature error: ; Among them, tanh is the hyperbolic tangent function, is a real number greater than 0; yes The transpose of When the control process just starts, the image error between the starting position and the target position in the image is is the largest, and the output of tanh is close to 0. At this time, the servo gain Close to the benchmark value , the system will be close to the benchmark value The control gain reduces image errors and does not cause image features to exceed the camera's FOV while ensuring stability. when When it approaches 0, the output of tanh is close to the maximum value of tanh() function, 1. At this time, the servo gain Close to maximum gain , the system will be close to the maximum gain The control gain can quickly adjust the error, thus achieving fast convergence.

[0014] The present invention provides the following beneficial effects: The adaptive visual servoing control method for unmanned aerial vehicles (UAVs) based on second-order cone constraints enables rapid and stable control system convergence in complex environments while effectively ensuring that image features are not lost. This method directly generates control instructions based on image features, without requiring three-dimensional pose information or relying on prior models. This method enables rapid and stable control system convergence while effectively ensuring that image features remain within the field of view, improving the stability and efficiency of the control system. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 Schematic diagram of the process flow of the adaptive visual servo control method for a UAV based on second-order cone constraints in an embodiment of the present invention; Figure 2 Schematic diagrams of two coordinate systems in an embodiment of the present invention; Figure 3 Schematic diagram of a virtual camera plane in an embodiment of the present invention; Figure 4 The projection diagrams of points in different coordinate systems according to the embodiment of the present invention are as follows; Figure 5 Schematic diagram of the adaptive visual servo control principle of a UAV based on second-order cone constraints in an embodiment of the present invention; Figure 6 A schematic diagram of the trajectory of image features generated by a simulation experiment according to an embodiment of the present invention; Figure 7 Schematic diagram of the changing trajectory of image point coordinates in a real camera according to an embodiment of the present invention; Figure 8 Schematic diagram of the motion trajectory of a drone in three-dimensional space in an embodiment of the present invention. DETAILED DESCRIPTION

[0016] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0017] The present invention provides a technical solution: a method for adaptive visual servo control of UAV based on second-order cone constraints, such as Figure 1 As shown, the following steps are included: S1. Build the quadrotor dynamics model and establish the inertial coordinate system and the quadrotor's own body coordinate system. The details include the following: S11. Build a quadrotor dynamics model. The rotational speeds of the four rotors are: By increasing the speed of the specified rotor, the quadrotor will generate a corresponding torque , the lift of the quadrotor and torque It is obtained by the following expression: ; in, is the lift coefficient, is the drag coefficient, is the distance between the position of each rotor center and the center of mass of the quadrotor drone; S12. Establish an inertial coordinate system according to the right-hand rule: ,in is the origin of the inertial coordinate system, Pointing north, Pointing east, Pointing vertically downward to the center of the earth; the body coordinate system is established as: , is the origin of the body coordinate system, Pointing to the front of the body, Point to the right, Vertically downward, such as Figure 2 shown.

[0018] S13. The roll angle, pitch angle and yaw angle of the quadrotor drone are: , the rotation matrix The rotation matrix is ​​also composed of these three Euler angles. It shows the degree of change of the body coordinate system relative to the inertial coordinate system. It is based on the rotation around the coordinate axis. Rotate in sequence Obtained by angle transformation; its specific form is: ; in: .

[0019] S2. Introducing the concept of a virtual image plane, we derive a visual servoing model in the virtual image plane and obtain a dynamic model of image points that does not change with posture. The details are as follows: S21. Assume a stationary point The coordinates in the inertial coordinate system are , whose coordinates in the camera system are Since the under-actuated characteristics of the quadcopter will bring about complex image feature motion, the virtual camera is defined so that the roll angle and pitch angle of the virtual camera are always zero; at the same time, a dynamic rotation matrix based on the attitude angle change rate is introduced, and the real-time attitude angle of the drone (such as pitch angle) is adjusted. , roll angle ) generates a compensation matrix, such as Figure 3As shown, the actual image features are projected onto a dynamically adjusted virtual image plane to further reduce nonlinear distortion.

[0020] S22. Define the coordinates of point P in the virtual camera coordinate system , and its coordinate relationship with point P in the inertial coordinate system is: ,in are the coordinates of point p, represents the virtual image plane, is the coordinate of the camera in the virtual camera coordinate system, is the rotation matrix in the yaw direction, yes Transpose the matrix. By using the perspective projection theorem, the coordinates of point P on the virtual image plane are: ; represents the coordinates of point P on the virtual image plane, is the focal length of the camera; S23, deriving the coordinates of step 22, we can obtain: ; in Yes The derivative with respect to time, is the derivative of point P with respect to time; Written in matrix form: ; in is the yaw angle The derivative of The relationship between the rate of change of a point on the virtual image plane and the quadrotor speed is obtained, namely the image point dynamics model.

[0021] S3. Define image moment features and derive a decoupled image moment feature dynamics model; specifically, the following: S31. Define three image moment features related to the projection coordinates of a specified point on the virtual image plane for controlling the three-dimensional translation motion of the quadrotor: ; Image features The coordinates on the virtual image plane are , , composed of functions; where: ; Indicates the selection of the visual target point for research. ,and , is the relative distance between the quadrotor drone and the target, yes expected value; express Any number within the value range, represents the virtual image plane; In the virtual image plane there are: ; in, Indicates the depth of the feature point relative to the virtual camera, yes , represents the depth corresponding to the desired relative position; S32. Derivative the image moment feature and combine it with the visual servo model obtained in step S2 to obtain a dynamic model of the image moment feature: ; As can be seen from the formula, the rate of change of the image features is independent of the rate of change of the pitch and roll angles of the drone, achieving the decoupling between the image motion and the pitch and roll motion of the drone; in, sk ( ) is an antisymmetric matrix, is the linear velocity of the virtual camera in the virtual camera coordinate system; ; yes The partial derivative of is the derivative of the yaw angle; The virtual camera simplifies the controller design by decoupling attitude and position control, and can better describe the motion characteristics of the drone when combined with the dynamic model.

[0022] S33. Define the image moment feature for controlling the yaw angle: ; Its derivative is: , therefore, the image features are expanded into: , the dynamic model of image moment features is expressed as: .

[0023] S4. The difference between the image feature S and the desired image feature S* is used as input to perform servo control on the drone. The second-order cone rule is used to constrain the target to be within the camera's field of view. The constraint feature is to converge to the desired position by the shortest path. The specific steps are as follows: S41, define the expected image features as , using image moment features to construct image error ,if Is a constant, take the derivative of the above formula and get ,in is the image Jacobian matrix of the image features, is the motion state of the virtual camera; from step 3, we can know that: ; in, is the depth of the feature point relative to the virtual camera, yes expected value; is the image feature; is the yaw angle The derivative of S42, construct on 2D image plane is the image feature at the next moment To the current image features and expected image features Distance of connection line: ; because , so the above formula can be written as: ; make , by deducing ;in is the identity matrix of the corresponding dimension; yes The transpose of Control rate based on image feedback: , is the sampling interval; if ,but exist and On the connection line; S43. Use the second-order cone rule to constrain the target to be within the camera's field of view. The constraint feature is to converge to the desired position by the shortest path. The second-order cone rule constraint conditions include three constraints: error convergence constraint, distance constraint, and physical constraint; the details are as follows: 1) To ensure that the error converges to 0 at an exponential rate at the next moment, the error convergence constraint is: ,in is the convergence coefficient; 2) Define distance constraints to ensure tends to 0, and the distance constraint is: ,in: is a projection matrix, is an auxiliary variable used to represent The upper bound of 3) Define the physical constraints of the control input: ,in and are the lower and upper bounds of the control input.

[0024] S5. Finally, a reference servo gain value is selected based on the size of the image feature error, and the adaptive servo gain is adjusted based on the size of the image feature error. Specifically, the following are included: From step 4, we can see that the derivative of the error with respect to time is: ; We hope that the error is an exponentially decreasing function, so that it can converge in a finite time. ,So: ; in, is the image Jacobian matrix of the image features, is the motion state of the virtual camera; is the generalized inverse of the Jacobian matrix, is the servo gain, defining a reference value , set the servo gain that is adaptively adjusted according to the size of the image feature error: ; Among them, tanh is the hyperbolic tangent function, is a real number greater than 0; yes The transpose of When the control process just starts, the image error between the starting position and the target position in the image is is the largest, and the output of tanh is close to 0. At this time, the servo gain Close to the benchmark value , the system will be close to the benchmark value The control gain reduces image errors and does not cause image features to exceed the camera's FOV while ensuring stability. when When it approaches 0, the output of tanh is close to the maximum value of tanh() function, 1. At this time, the servo gain Close to maximum gain , the system will be close to the maximum gain The control gain can quickly adjust the error, thus achieving fast convergence.

[0025] Simulation experiment The principle of UAV adaptive visual servo control based on second-order cone constraints is as follows: Figure 5As shown in the figure, the adaptive visual servo control method for UAV with second-order cone constraint of the present invention is simulated on Matlab to show the effect of the designed controller. The goal of the visual servo experiment is to control the quadrotor from an initial position to a specified desired position by eliminating the deviation of image features. In the simulation, the initial position of the quadrotor is set to (-10, -10, 3) m, the initial Euler angle is (0, 0, 1) rad, and the initial image moment feature is Desired image moment features , .

[0026] According to the initial value and expected value of the image feature, we can generate the trajectory of the image feature according to the provided second-order cone constraint method as follows Figure 6 The changing trajectory of the image point coordinates in the real camera is shown as Figure 7 As shown, it can be seen that the image points do not exceed the camera plane and the trajectory is smooth. Figure 8 The motion trajectory of the drone in three-dimensional space is shown. Through simulation, it can be seen that the method of the present invention can enable the control system to converge quickly and stably in complex environments, while effectively ensuring that image features are not lost.

[0027] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or apparatus comprising the element.

[0028] The terms used in the embodiments of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The singular forms "a", "an", "the" and "the" used in the embodiments of the present invention and the appended claims are also intended to include plural forms unless the context clearly indicates otherwise.

[0029] It should be understood that the term "and / or" as used herein is merely a description of the relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. Furthermore, the character " / " in this document generally indicates that the associated objects are in an "or" relationship.

[0030] The word "if," as used herein, may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to the determination" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)," depending on the context.

[0031] The references to "first" and "second" in the embodiments merely distinguish similar objects and do not represent a specific ordering of the objects. It is understood that the specific order or precedence of "first" and "second" can be interchanged where appropriate. It should be understood that the objects distinguished by "first" and "second" can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein.

[0032] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An adaptive visual servo control method for UAV based on second-order cone constraints, characterized in that: The steps include: S1. Build a quadrotor dynamics model and establish the inertial coordinate system and the quadrotor's own body coordinate system; S2. The idea of ​​virtual image plane is introduced, and the visual servoing model in the virtual image plane is derived to obtain the image point dynamics model that does not change with the posture; S3. Define image moment features and derive a decoupled image moment feature dynamics model; S4, taking the difference between the image feature S and the desired image feature S* as input, servo-controls the drone, and constrains the target to be within the camera's field of view using the second-order cone rule. The constraint feature is to converge to the desired position along the shortest path. S5. Finally, a reference servo gain value is selected according to the size of the image feature error, and the adaptive servo gain is adjusted according to the size of the image feature error.

2. The method for adaptive visual servo control of a UAV based on second-order cone constraints according to claim 1 is characterized in that: In step S1, the quadrotor dynamics model is constructed, and the inertial coordinate system and the quadrotor's own body coordinate system are established, specifically including the following: S11. Build a quadrotor dynamics model. The rotational speeds of the four rotors are: By increasing the speed of the specified rotor, the quadrotor will generate a corresponding torque , the lift of the quadrotor and torque It is obtained by the following expression: ； in, is the lift coefficient, is the drag coefficient, is the distance between the position of each rotor center and the center of mass of the quadrotor drone; S12. Establish an inertial coordinate system according to the right-hand rule: ,in is the origin of the inertial coordinate system, Pointing north, Pointing east, Pointing vertically downward to the center of the earth; the body coordinate system is established as: , is the origin of the body coordinate system, Pointing to the front of the body, Point to the right, vertically downward; S13. The roll angle, pitch angle and yaw angle of the quadrotor drone are: , the rotation matrix The rotation matrix is ​​also composed of these three Euler angles. It shows the degree of change of the body coordinate system relative to the inertial coordinate system. It is based on the rotation around the coordinate axis. Rotate in sequence Obtained by angle transformation; its specific form is: ； in: 。 3. The method for adaptive visual servo control of a UAV based on second-order cone constraints according to claim 1 is characterized in that: In step S2, the idea of ​​a virtual image plane is introduced, and a visual servoing model in the virtual image plane is derived to obtain an image point dynamics model that does not change with the posture. Specifically, it includes the following: S21. Assume a stationary point The coordinates in the inertial coordinate system are , whose coordinates in the camera system are By defining a virtual camera, the roll and pitch angles of the virtual camera are always zero. At the same time, a dynamic rotation matrix based on the attitude angle change rate is introduced to generate a compensation matrix according to the real-time attitude angle of the drone, and the actual image features are projected onto the dynamically adjusted virtual image plane. S22. Define the coordinates of point P in the virtual camera coordinate system , and its coordinate relationship with point P in the inertial coordinate system is: ,in, are the coordinates of point p, represents the virtual image plane, is the coordinate of the camera in the virtual camera coordinate system, is the rotation matrix in the yaw direction, yes Transpose the matrix. By using the perspective projection theorem, the coordinates of point P on the virtual image plane are: ； represents the coordinates of point P on the virtual image plane, is the focal length of the camera; S23, deriving the coordinates of step 22, we can obtain: ； in Yes The derivative with respect to time, is the derivative of point P with respect to time; Written in matrix form: ； in is the yaw angle The derivative of The relationship between the rate of change of a point on the virtual image plane and the quadrotor speed is obtained, namely the image point dynamics model.

4. The method for adaptive visual servo control of a UAV based on second-order cone constraints according to claim 1, characterized in that: In step S3, the image moment features are defined and the decoupled image moment feature dynamics model is derived, which specifically includes the following: S31. Define three image moment features related to the projection coordinates of a specified point on the virtual image plane for controlling the three-dimensional translation motion of the quadrotor: ； Image features The coordinates on the virtual image plane are , , composed of functions; where: ； Indicates the selection of the visual target point for research. ,and , is the relative distance between the quadrotor drone and the target, yes expected value; express Any number within the value range, represents the virtual image plane; In the virtual image plane there are: ； in, Indicates the depth of the feature point relative to the virtual camera, yes expected value; S32. Derivative the image moment feature and combine it with the visual servo model obtained in step S2 to obtain a dynamic model of the image moment feature: ； The decoupling between image motion and the pitch and roll motion of the drone is achieved; sk ( ) is an antisymmetric matrix, is the linear velocity of the virtual camera in the virtual camera coordinate system; ; yes The partial derivative of is the derivative of the yaw angle; S33. Define the image moment feature for controlling the yaw angle: ； Its derivative is: , therefore, the image features are expanded into: , the dynamic model of image moment features is expressed as: 。 5. The method for adaptive visual servo control of a UAV based on second-order cone constraints according to claim 1, characterized in that: In step S4, the specific steps include: S41, define the expected image features as , use image moment features to construct image error: , taking the derivative of the above formula, we get ,in is the image Jacobian matrix of the image features, is the motion state of the virtual camera; from step 3, we can know that: ； in, is the depth of the feature point relative to the virtual camera, yes expected value; is the image feature; is the yaw angle The derivative of S42, construct on 2D image plane is the image feature at the next moment To the current image features and expected image features Distance of connection line: ； because , so the above formula can be written as: ； make , by deducing ;in is the identity matrix of the corresponding dimension; yes The transpose of Control rate based on image feedback: , is the sampling interval; if ,but exist and On the connection line; S43. Constrain the target to be within the camera's field of view through the second-order cone rule, and the constraint feature is to converge to the desired position by the shortest path.

6. The method for adaptive visual servo control of a UAV based on second-order cone constraints according to claim 5, characterized in that: In step S43, the second-order cone rule constraint conditions include three constraints: error convergence constraint, distance constraint, and physical constraint; specifically, they are as follows: 1) To ensure that the error converges to 0 at an exponential rate at the next moment, the error convergence constraint is: ,in is the convergence coefficient; 2) Define distance constraints to ensure tends to 0, and the distance constraint is: ,in: is a projection matrix, is an auxiliary variable used to represent The upper bound of 3) Define the physical constraints of the control input: ,in and are the lower and upper bounds of the control input.

7. The method for adaptive visual servo control of a UAV based on second-order cone constraints according to claim 1, characterized in that: In step S5, the specific steps include: From step 4, we can see that the derivative of the error with respect to time is: ; make ,So: ; in, is the image Jacobian matrix of the image features, is the motion state of the virtual camera; is the generalized inverse of the Jacobian matrix, is the servo gain, defining a reference value , set the servo gain that is adaptively adjusted according to the size of the image feature error: ； Among them, tanh is the hyperbolic tangent function, is a real number greater than 0; yes The transpose of When the control process just starts, the image error between the starting position and the target position in the image is is the largest, and the output of tanh is close to 0. At this time, the servo gain Close to the benchmark value , the system will be close to the benchmark value The control gain reduces the image error; when When it approaches 0, the output of tanh is close to the maximum value of tanh() function, 1. At this time, the servo gain Close to maximum gain , the system will be close to the maximum gain The control gain quickly adjusts the error.

Images