A Composite Anti-Interference Control Method for the End of a Flying Manipulator Against Multi-Source Interference

Abstract

The present invention provides a method for composite anti-interference control of the end of a flying robotic arm against multi-source interference, aiming to solve the influence of multi-source interference on the flying robotic arm when performing tasks. First, the kinematic equation of the end of the flying robotic arm actuator in the world coordinate system is constructed, and then through quantitative analysis, it is divided into the controllable part of the unmanned aerial vehicle and the controllable part of the robotic arm. Secondly, a joint angle planner is designed according to the kinematic equation to suppress the base floating interference. After that, a multi-source interference model of the end control system of the actuator is established, and a neural network with the state as the input is designed to estimate the dynamic model. Finally, a nonlinear disturbance observer is designed to compensate for the estimation residual of the neural network. By reasonably selecting control parameters, the high-precision performance of the end of the actuator can be effectively guaranteed. Based on the composite hierarchical anti-interference architecture, the present invention can significantly improve the accuracy of the end of the actuator of the flying robotic arm system under multi-source interference and can be used for aerial operation tasks such as emergency rescue, target capture, and power tower maintenance.

Description

Technical Field

[0001] The present invention belongs to the field of flight robot control, and particularly relates to a composite anti-interference control method for the end of a flight manipulator against multi-source interference, which is applicable to the flight manipulator control system for performing high-precision active operation tasks. Background Art

[0002] In recent years, with the continuous improvement and development of electronic technology and material technology, a flight manipulator composed of a multi-rotor unmanned aerial vehicle (UAV) and a manipulator, which can physically interact with the environment, has become a very popular field, as Figure 2 shown. The ability of the flight manipulator to physically interact with the environment has extended its application to various operation tasks. When physically interacting with the environment, the accuracy of the end of the flight manipulator actuator determines its specific application fields. In various operation tasks, there are some special scenarios with very strict requirements for the accuracy of the end of the flight manipulator actuator. For example, bridge detection, pipeline maintenance, emergency rescue, and non-cooperative target capture in narrow spaces, etc. Therefore, ensuring the high precision of the end of the flight manipulator actuator is a key problem that urgently needs to be solved. However, compared with traditional fixed-base manipulators, the flight manipulator has multi-source interferences during the task execution process, including base floating interference, model uncertainty, and unknown external disturbances, which seriously affect the end execution accuracy. At the kinematic level, on the one hand, affected by wind interference and coupling interference, there is a certain control error in the UAV base, which will continuously jitter near the desired position. On the other hand, due to the multi-link characteristics of the manipulator, the floating interference of the UAV base transmitted to the end of the actuator will have an error amplification effect, seriously deteriorating the accuracy of the end of the actuator. At the dynamic level, the manipulator in the flight manipulator is a multi-link structure, and the center of mass is constantly changing. It is very difficult to accurately establish its dynamic model, so there is a certain model uncertainty. At the same time, when the flight manipulator performs high-precision operation tasks, the manipulator will be affected by the downwash airflow interference caused by the rotation of the UAV propellers, which will also have a certain impact on the accuracy of the end of the actuator.

[0003] Therefore, in order to ensure the high precision of the end of the flight manipulator actuator under multi-source interference and complete precise operation tasks, the composite anti-interference control algorithm of the flight manipulator actuator must solve the above-mentioned multi-source interferences of base floating interference, model uncertainty interference, and unknown external disturbances during the design process.

[0004] Chinese invention patent CN201810010602.3 proposes an arm - equipped drone with a parallel manipulator. A parallel manipulator is provided below the rotorcraft to improve the control accuracy of the end - effector. However, the proposed device has problems such as a small working space and the actuator can only work below the rotorcraft. The rotor - flight grasping manipulator proposed in Chinese invention patent CN202123017099.0 also has problems such as a small working space and cannot complete active operation tasks. Chinese invention patent CN202110993793.1 proposes a control scheme that combines a disturbance observer and an H - infinity controller to handle the multi - source disturbances suffered by the flight manipulator. However, the prerequisite is to obtain the manipulator dynamics model, and it is very difficult to obtain the dynamics model of a multi - degree - of - freedom manipulator. Chinese invention patent CN202010801707.8 designs a disturbance observer to estimate the amplitude of the base floating disturbance, but it requires that the floating disturbance of the base must be a periodic disturbance. Chinese invention patent CN202110994016.9 proposes a control algorithm that combines an anti - saturation controller and an external force estimator, but does not handle the floating disturbance of the drone base, resulting in a large control error at the end of the flight manipulator and unable to meet the requirements of high - precision operation tasks. Chinese invention patent CN201810094313.6 proposes a rotor - flight manipulator system and algorithm based on dynamic center - of - gravity compensation, but there are two problems: (1) It does not consider the kinematic disturbance caused by the floating of the rotorcraft drone to the end - effector of the manipulator actuator; (2) It does not consider the dynamic coupling disturbance caused by the movement of the manipulator to the drone base.

[0005] None of the above - mentioned methods can solve the problem that during the task execution of the flight manipulator, due to the influence of multi - source disturbances including base floating disturbance, model uncertainty disturbance, and unknown external disturbances, the accuracy of the end - effector deteriorates, resulting in the failure of precise operation tasks. Summary of the Invention

[0006] To overcome the defects of the existing methods, for a flight manipulator system composed of a multi - rotor drone and a multi - degree - of - freedom manipulator, a compound anti - disturbance control structure for the end - effector of the flight manipulator against multi - source disturbances is provided, which can ensure the high precision of the end - effector of the flight manipulator under multi - source disturbances to complete various precise operation tasks. The multi - source disturbances include base floating disturbance, model uncertainty disturbance, and unknown external disturbances.

[0007] To achieve the above - mentioned purpose, the technical solution adopted by the present invention is as follows:

[0008] First step, establish the kinematic equation of the end - effector pose of the flight manipulator in the world coordinate system, and through quantitative analysis, divide it into the drone - controllable part and the manipulator - controllable part:

[0009] According to the motion characteristics of the flying robotic arm, a kinematic model of the end - pose of the flying robotic arm actuator is established based on the homogeneous transformation principle, which is expressed as follows:

[0010]

[0011] In the formula, P e represents the position of the end of the flying robotic arm actuator in the world coordinate system, and R e is the rotation matrix of the end of the flying robotic arm actuator in the world coordinate system; P b represents the position of the centroid of the UAV base in the world coordinate system, and R b is the rotation matrix of the UAV centroid in the world coordinate system; represents the position of the end of the robotic arm actuator in the UAV coordinate system, represents the rotation matrix of the end of the robotic arm actuator in the UAV coordinate system; Differentiating the above formula, we can get:

[0012]

[0013] Among them, S(·) is the skew - symmetric matrix operator for performing the cross - product operation. Ω b and Ω e respectively represent the angular velocities of the UAV centroid and the end of the flying robotic arm actuator in the world coordinate system; and respectively define the linear velocity and angular velocity of the end of the actuator relative to the UAV centroid; According to the above two formulas, the expression of the end of the actuator in the world coordinate system can be obtained as:

[0014]

[0015]

[0016] Among them, represents the Jacobian matrix of the end of the actuator relative to the base; I 3×3 and 0 3×3 represent the 3×3 identity matrix and zero matrix; represents the angular velocity of the robotic arm joint angle. represents the linear velocity and angular velocity of the UAV centroid in the world coordinate system; Define the conversion matrix between the angular velocity of the end of the actuator and the differential of the Euler angles as T e , and the conversion matrix between the angular velocity of the UAV centroid and the differential of the Euler angles as T b , the expression of the end - pose of the flying robotic arm actuator in the world coordinate system can be obtained as:

[0017]

[0018] Among them, The differential of the position x of the end of the flight robotic arm actuator in the world coordinate system e . The differential of the position x of the center of mass of the UAV in the world coordinate system. According to the above formula, b is obtained as follows: is the controllable part of the UAV, is the controllable part of the robotic arm. Rewrite the above formula into the following form:

[0019]

[0020] where d represents the base floating disturbance caused by the movement of the UAV base. When the flight robotic arm is performing tasks, the UAV base is usually in a quasi-static motion, and its state changes are relatively small. Therefore, J b , T b and are all bounded values. Therefore, the base floating disturbance d should be a bounded disturbance. Let b1 represent the upper bound of the base floating disturbance, that is, ||d|| ≤ b1. ||·|| represents the 2-norm of the vector.

[0021] Second step, design an angle planner for the robotic arm joints according to the established kinematic equation to suppress the base floating disturbance:

[0022] First, define the desired position x d of the end of the flight robotic arm actuator in the world coordinate system. Then, the tracking error at the end is Δx = x e - x d . Design the potential energy function related to the tracking error of the actuator end as:

[0023]

[0024] where K p , σ and N are parameters to be designed; max{θ1, θ2} represents the larger value of θ1 and θ2. To suppress the kinematic disturbance caused by the UAV base floating, design the reference input signal of the robotic arm joint angle as:

[0025]

[0026] where Δε is the partial derivative of the potential energy function P(Δx) with respect to the error Δx, that is and represent the pseudo-inverse and transpose of the matrix T2 respectively. K d , α and a are parameters to be designed.

[0027] Third step, establish a multi-source interference model for the pose control system of the end of the flight robotic arm actuator, and design a neural network with the robotic arm state as the input to estimate the overall dynamic model of the robotic arm:

[0028] For the model uncertainty interference and unknown external disturbances of the flying robotic arm system, an overall dynamic model of the robotic arm system is established, and a neural network with the robotic arm state as the input is designed to estimate the overall dynamic model of the robotic arm. The dynamic model is expressed as follows:

[0029]

[0030] where the state vector q, respectively represent the rotational angles, angular velocities, and angular accelerations of the joints of the robotic arm; τ dis represents the unknown external disturbance received by the robotic arm system; M e (q) is the nominal part of the positive definite inertia matrix of the robotic arm system, is the nominal part containing the Coriolis force and centripetal force matrix, and G e (q) represents the nominal part of the gravity matrix; is the model uncertainty interference part caused by the movement of the robotic arm; τ represents the control input torque of the system; A neural network is used to estimate the overall dynamic model of the robotic arm: With a sufficient number of neurons, the neural network can estimate any non-linear continuous function, and the expression is as follows:

[0031]

[0032] where W is the unknown ideal weight matrix of the neurons in the neural network, represents the neural network activation function with the state vector q, as the input, represents the estimation of the overall dynamic model of the robotic arm containing model uncertainty and unknown external disturbances by the neural network; Use E d to represent the neural network estimation residual; According to the above formula, the input torque of the robotic arm is designed as:

[0033]

[0034] where K s , K v and K a are control parameter matrices to be designed; is the actual estimated output of the neural network; is the actual output of the non-linear disturbance observer; Combining the above two formulas, we can get:

[0035]

[0036] where, represents the estimation error between the ideal weight matrix and the actual weight matrix of the neurons; Tracking error of the manipulator joint angular velocity; Tracking error of the manipulator joint angular acceleration; Denote the estimation error of the nonlinear disturbance observer; the adaptive update rate of the designed neuron weight matrix is:

[0037]

[0038] where, k2 is the adaptive update rate parameter to be designed.

[0039] Step 4: Design a nonlinear disturbance observer to compensate for the estimation residual of the neural network:

[0040] According to the characteristics of the estimation residual, establish the following nonlinear disturbance observer:

[0041]

[0042] where, φ is the auxiliary variable of the nonlinear disturbance; L is the observation gain of the nonlinear disturbance observer; for the estimation error of the nonlinear disturbance observer, taking the differential gives:

[0043]

[0044] Then, by selecting the control parameters α, k2, γ1, γ2, γ3 and γ4, as well as the parameter matrices L, K s and K v to satisfy the following inequalities, to ensure that the errors Δx, and are ultimately uniformly bounded, so that the end of the flying manipulator actuator can accurately track the trajectory under the multi-source disturbances including the base floating disturbance, model uncertainty and unknown external disturbances:

[0045]

[0046] where, I 4×4 represents the 4×4 identity matrix.

[0047] The advantages of the present invention compared with the prior art are:

[0048] A compound anti-interference control structure for the end of a flying manipulator against multi-source interferences including base floating interference, model uncertainty, and unknown external disturbances involved in the present invention is mainly for a flying manipulator system composed of a multi-rotor UAV and a multi-degree-of-freedom manipulator, which can ensure the high-precision performance of the end of the actuator of the flying manipulator under multi-source interferences to complete various precise aerial operation tasks. Compared with traditional fixed-base manipulators, the flying manipulator has the significant advantages of high mobility and high flexibility. However, the multi-source interferences existing in the process of performing tasks will seriously deteriorate the accuracy of the end of the actuator of the flying manipulator, resulting in the failure of precise operation tasks. This method designs a compound anti-interference control method for a flying manipulator containing multi-source interferences such as base floating interference, model uncertainty, and unknown external disturbances. First, the kinematic equation of the end pose of the flying manipulator actuator in the world coordinate system is constructed, and then it is divided into a UAV controllable part and a manipulator controllable part through quantitative analysis. Secondly, according to the established kinematic equation, an angle planner for the manipulator joints is designed at the kinematic level to suppress the base floating interference. After that, a multi-source interference model of the end pose control system of the flying manipulator actuator is established, and a neural network with the manipulator state as the input is designed to estimate the overall dynamic model of the manipulator. Finally, a nonlinear disturbance observer is designed to compensate for the estimation residual of the neural network; at the dynamic level, by reasonably selecting the parameters of the neural network and the nonlinear disturbance observer, the end accuracy of the flying manipulator under multi-source interferences can be significantly improved, ensuring the precise completion of active operation tasks. Description of the Drawings

[0049] Figure 1 It is a design flow chart of a compound anti-interference control structure for the end of a flying manipulator against multi-source interferences according to the present invention;

[0050] Figure 2 It is a structural diagram of the invented flying manipulator system. Detailed Embodiments

[0051] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0052] Taking a general flying manipulator system composed of a multi-rotor UAV and a multi-degree-of-freedom manipulator as an example to illustrate the specific implementation of the system and method, when the flying manipulator performs precise operation tasks, it has high requirements for the accuracy of the end of the actuator;

[0053] As Figure 1 shown, the specific implementation steps of the present invention are as follows:

[0054] First, establish the kinematic equation of the end pose of the flight manipulator actuator in the world coordinate system, and divide it into the controllable part of the UAV and the controllable part of the manipulator through quantitative analysis:

[0055] According to the motion characteristics of the flight manipulator, establish a kinematic model of the end pose of the flight manipulator actuator based on the homogeneous transformation principle, which is expressed as follows:

[0056]

[0057] In the formula, P e represents the position of the end of the flight manipulator actuator in the world coordinate system, and R e is the rotation matrix of the end of the flight manipulator actuator in the world coordinate system; P b represents the position of the centroid of the UAV base in the world coordinate system, and R b is the rotation matrix of the UAV centroid in the world coordinate system; represents the position of the end of the manipulator actuator in the UAV coordinate system, represents the rotation matrix of the end of the manipulator actuator in the UAV coordinate system; Differentiating the above formula, we can get:

[0058]

[0059] Among them, S(·) is the skew-symmetric matrix operator for performing the cross product operation. Ω b and Ω e respectively represent the angular velocities of the UAV centroid and the end of the flight manipulator actuator in the world coordinate system; and respectively define the linear velocity and angular velocity of the end of the actuator relative to the UAV centroid; According to the above two formulas, the expression of the end of the actuator in the world coordinate system can be obtained as:

[0060]

[0061]

[0062] Among them, represents the Jacobian matrix of the end of the actuator relative to the base; I 3×3 and 0 3×3 represent the 3×3 identity matrix and zero matrix; represents the angular velocity of the manipulator joint angle. represents the linear velocity and angular velocity of the UAV centroid in the world; Define the conversion matrix between the angular velocity of the end of the actuator and the differential of the Euler angles as T e , and the conversion matrix between the angular velocity of the UAV centroid and the differential of the Euler angles as T b , the expression of the end pose of the flight manipulator actuator in the world coordinate system can be obtained as:

[0063]

[0064] Among them, is the differential of the position x of the end of the flight robotic arm actuator in the world coordinate system e . is the differential of the position x of the center of mass of the UAV in the world coordinate system b . According to the above formula, it can be obtained that is the controllable part of the UAV, is the controllable part of the robotic arm. Rewrite the above formula into the following form:

[0065]

[0066] Among them, d represents the base floating disturbance caused by the movement of the UAV base. When the flight robotic arm is performing tasks, the UAV base is usually in a quasi-static motion, and its state changes are relatively small. Therefore J b , T b and are all bounded values. Therefore, the base floating disturbance d should be a bounded disturbance. Let b1 represent the upper bound of the base floating disturbance, that is, ||d|| ≤ b1. ||·|| represents the 2-norm of the vector.

[0067] Second step, design an angle planner for the robotic arm joints according to the established kinematic equation to suppress the base floating disturbance:

[0068] First, define the desired position x d of the end of the flight robotic arm actuator in the world coordinate system, then the tracking error of the end is Δx = x e - x d . Design a potential energy function related to the tracking error of the actuator end as:

[0069]

[0070] Among them, K p , σ and N are parameters to be designed; max{θ1, θ2} represents the larger value of θ1 and θ2. To suppress the kinematic disturbance caused by the floating of the UAV base, design the reference input signal of the robotic arm joint angle as:

[0071]

[0072] Among them, Δε is the partial derivative of the potential energy function P(Δx) with respect to the error Δx, that is and represent the pseudo-inverse and transpose of the matrix T2 respectively. K d , α and a are parameters to be designed.

[0073] In the third step, establish a multi-source interference model for the end pose control system of the flying robotic arm, and design a neural network with the robotic arm state as the input to estimate the overall dynamic model of the robotic arm:

[0074] For the model uncertainty interference and unknown external disturbances of the flying robotic arm system, establish the overall dynamic model of the robotic arm system, and design a neural network with the robotic arm state as the input to estimate the overall dynamic model of the robotic arm. The dynamic model is expressed as follows:

[0075]

[0076] where the state vector q, respectively represent the rotational angles, angular velocities, and angular accelerations of the joints of the robotic arm; τ dis represents the unknown external disturbance acting on the robotic arm system; M e (q) is the nominal part of the positive definite inertia matrix of the robotic arm system, is the nominal part including the Coriolis force and centripetal force matrix, and G e (q) represents the nominal part of the gravity matrix; is the model uncertainty interference part caused by the movement of the robotic arm; τ represents the control input torque of the system; use a neural network to estimate the overall dynamic model of the robotic arm: when there are enough neurons, the neural network can estimate any non-linear continuous function, and the expression is as follows:

[0077]

[0078] where W is the unknown ideal weight matrix of the neurons in the neural network, represents the neural network activation function with the state vector q, as the input, represents the estimation of the overall dynamic model of the robotic arm including model uncertainty and unknown external disturbances by the neural network; use E d to represent the neural network estimation residual; according to the above formula, design the input torque of the robotic arm as:

[0079]

[0080] where K s , K v and K a are control parameter matrices to be designed; is the actual estimated output of the neural network; is the actual output of the non-linear disturbance observer; combining the above two formulas, we can get:

[0081]

[0082] Among them, represents the estimation error between the ideal weight matrix and the actual weight matrix of the neuron; the tracking error of the angular velocity of the robotic arm joint; the tracking error of the angular acceleration of the robotic arm joint; represents the estimation error of the nonlinear disturbance observer; the adaptive update rate of the designed neuron weight matrix is:

[0083]

[0084] Among them, k2 is the adaptive update rate parameter to be designed.

[0085] Step 4, design a nonlinear disturbance observer to compensate for the estimation residual of the neural network:

[0086] According to the characteristics of the estimation residual, establish a nonlinear disturbance observer as follows:

[0087]

[0088] Among them, φ is the auxiliary variable of the nonlinear disturbance; L is the observation gain of the nonlinear disturbance observer; for the estimation error of the nonlinear disturbance observer, taking the derivative gives:

[0089]

[0090] Then, by selecting the control parameters α, k2, γ1, γ2, γ3, and γ4, as well as the parameter matrices L, K s and K v to satisfy the following inequalities to ensure that the errors Δx, and are ultimately uniformly bounded, so that the end of the actuator of the flying robotic arm can accurately track the trajectory under the multi-source disturbances including the base floating disturbance, model uncertainty, and unknown external disturbances:

[0091]

[0092] Among them, I 4×4 represents the 4×4 identity matrix.

[0093] The content not described in detail in the specification of the present invention belongs to the prior art well-known to those skilled in the art.

[0094] It is easy for those skilled in the art to understand that the above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A compound anti-interference control method for the end of a flying robotic arm against multi-source interference, characterized in that: Step 1: Construct the kinematic equation of the end - pose of the flight manipulator actuator in the world coordinate system, and then divide it into the controllable part of the UAV and the controllable part of the manipulator through quantitative analysis, including: According to the motion characteristics of the flight manipulator, establish a kinematic model of the end - pose of the flight manipulator actuator based on the homogeneous transformation principle, which is expressed as follows: ； Wherein, represents the position of the end of the flying robotic arm actuator in the world coordinate system, is the rotation matrix of the end of the flying robotic arm actuator in the world coordinate system; represents the position of the centroid of the UAV base in the world coordinate system, is the rotation matrix of the centroid of the UAV in the world coordinate system; represents the position of the end of the robotic arm actuator in the UAV coordinate system, represents the rotation matrix of the end of the robotic arm actuator in the UAV coordinate system; Differentiating the above equation gives: ； Among them, is the skew-symmetric matrix operator for performing the cross product operation, and respectively represent the angular velocities of the centroid of the UAV and the end of the flight manipulator actuator in the world coordinate system; and respectively define the linear velocity and angular velocity of the end of the actuator relative to the centroid of the UAV; according to the above two equations, the expression of the end of the actuator in the world coordinate system can be obtained as: ； ； Among them, represents the Jacobian matrix of the end of the actuator relative to the base; and represent the identity matrix and the zero matrix of represents the angular velocity of the joint angle of the robotic arm, represents the linear velocity and angular velocity of the center of mass of the UAV in the world coordinate system; define the conversion matrix between the angular velocity of the end of the actuator and the differential of the Euler angles as and the conversion matrix between the angular velocity of the center of mass of the UAV and the differential of the Euler angles as , the expression of the pose of the end of the flight robotic arm actuator in the world coordinate system can be obtained as: ； Among them, is the position of the end of the flight robotic arm actuator in the world coordinate system of the differential, is the position of the center of mass of the UAV in the world coordinate system of the differential, according to the above formula, it can be obtained that is the controllable part of the UAV, is the controllable part of the robotic arm, rewrite the above formula into the following form: ； Among them, represents the base floating interference caused by the movement of the UAV base. When the flying robotic arm is performing tasks, the UAV base is usually in a quasi-static motion, and its state changes are relatively small. Therefore , , and are all bounded values. Therefore, the base floating interference should be a bounded interference. Let represent the upper bound of the base floating interference, that is , represents the 2-norm of the vector; Step 2: Design an angle planner for the manipulator joints at the kinematic level according to the established kinematic equation to suppress the base floating interference; Step 3: Establish a multi - source interference model for the end - pose control system of the flight manipulator actuator, and design a neural network with the manipulator state as the input to estimate the overall dynamic model of the manipulator; Step 4: Design a nonlinear disturbance observer to compensate for the estimation residual of the neural network; Step 5: At the dynamic level, by reasonably selecting the parameters of the neural network and the nonlinear disturbance observer, effectively ensure the high - precision performance of the end of the flight manipulator actuator and achieve the precise grasping task.

2. The compound anti-interference control method for the end of a flying robotic arm against multi-source interference according to claim 1, characterized in that: The specific content of the second step includes: First, define the desired position of the end of the flying robotic arm actuator in the world coordinate system , then the tracking error of the end is . Design the potential energy function related to the tracking error of the actuator end as follows: ； Among them, , and are parameters to be designed; represents the selection of and the larger value. To suppress the kinematic interference caused by the floating of the drone base, the reference input signal of the robotic arm joint angle is designed as: ； Among them, is the potential energy function is the partial derivative of the error with respect to, that is , and respectively represent the pseudo-inverse and transpose of the matrix , and are the parameters to be designed.

3. The compound anti-interference control method for the end of a flying robotic arm against multi-source interference according to claim 2, characterized in that: The specific content of the third step includes: Aiming at the model uncertainty interference and unknown external disturbances of the flight manipulator system, establish the overall dynamic model of the manipulator system, and design a neural network with the manipulator state as the input to estimate the overall dynamic model of the manipulator. The dynamic model is expressed as follows: ； Among them, the state vector represents the rotation angle, rotation angular velocity and rotation angular acceleration of each joint of the robotic arm respectively; represents the unknown external disturbance suffered by the robotic arm system; is the nominal part of the positive definite inertia matrix of the robotic arm system, contains the nominal part of the Coriolis force and centripetal force matrix, represents the nominal part of the gravity matrix; is the model uncertainty disturbance part caused by the movement of the robotic arm; represents the control input torque of the system; Use a neural network to estimate the overall dynamic model of the robotic arm: When there are enough neurons, the neural network can estimate any non-linear continuous function, and the expression is as follows: ； wherein, is the unknown ideal weight matrix of neurons in the neural network, represents the neural network activation function with the state vector as the input, represents the estimation of the overall dynamic model of the robotic arm by the neural network, which includes model uncertainties and unknown external disturbances; use to represent the neural network estimation residual; according to the above formula, the input torque of the robotic arm is designed as: ； Among them, , and are control parameter matrices to be designed; is the actual estimated output of the neural network; is the actual output of the nonlinear disturbance observer; Combining the above two equations, we get: ； Among them, represents the estimation error between the ideal weight matrix and the actual weight matrix of the neuron; the tracking error of the joint angular velocity of the robotic arm; the tracking error of the joint angular acceleration of the robotic arm; represents the estimation error of the nonlinear disturbance observer; the adaptive update rate of the designed neuron weight matrix is: ； Among them, is the adaptive update rate parameter to be designed.

4. The composite anti-interference control method for the end of a flying robotic arm against multi-source interference according to claim 3, wherein: The specific content of the fourth step includes: Aiming at the characteristics of the estimation residual, establish a nonlinear disturbance observer as follows: ； Among them, is the auxiliary variable of the non-linear jammer; is the observation gain of the non-linear disturbance observer; the estimation error of the non-linear disturbance observer Differentiating it gives: ； Then, by selecting control parameters , , , , and , and parameter matrices , and to satisfy the following inequalities to ensure that the errors , , and are ultimately uniformly bounded, enabling the end of the flying manipulator actuator to accurately track the trajectory under multi-source disturbances including base floating disturbances, model uncertainties, and unknown external disturbances: ； Among them, represents the identity matrix.

Images