A method and system for visual tracking control of a robotic arm for a moving target

Abstract

The present invention discloses a method and system for visual tracking control of a robotic arm for a moving target. The method comprises: estimating and predicting the unknown motion state information of the moving target using a Kalman filter method based on a camera perspective projection model; designing the prediction results based on a visual prediction tracking control method using a Kalman filter to obtain desired joint angular velocity information; and compensating for the uncertain dynamics and external disturbances of the robotic system using a robot dynamics control method based on an adaptive neural network, thereby achieving dynamic tracking of the desired joint angular velocity information. The system comprises: an estimation module, a kinematic control module, and a dynamic control module. By using the present invention, the problems of unknown target motion, uncertain dynamics, and external disturbances can be simultaneously addressed. As a method and system for visual tracking control of a robotic arm for a moving target, the present invention can be widely applied to the field of robotic arm visual tracking technology.

Description

Technical Field

[0001] The present invention relates to the technical field of robot arm visual tracking, and in particular to a robot arm visual tracking control method and system for a moving target. Background Art

[0002] Visual sensors are one of the important external sensors of robots, with the characteristics of non-contact measurement, low cost, high reliability, and rich information. Robotic arm visual servo is an important method that uses visual feedback information to control robot motion. Visual servo control technology can make robots more flexible and intelligent. Existing technical methods include a four-degree-of-freedom robotic arm visual servo control method to achieve autonomous positioning of the robotic arm on a cooperative target. For example, a visual servo method based on SVM and proportional control uses an SVM-based training model to obtain the Jacobian matrix, and uses a proportional control method to drive the robot to the target position. For example, a robotic arm target tracking method based on visual servo uses an α-β filtering method to estimate the image Jacobian matrix, and uses each element of the matrix as the state of the control system to design a controller to achieve effective tracking of the target. Most previous related visual servo control research works focused on stationary targets, and the above methods did not simultaneously consider multiple uncertainty problems such as unknown target motion, uncertain dynamics, and external disturbances. Summary of the Invention

[0003] In order to solve the above technical problems, the purpose of the present invention is to provide a robotic arm visual tracking control method and system for a moving target, which can simultaneously handle the problems of unknown target motion, uncertain dynamics and external disturbances.

[0004] The first technical solution adopted by the present invention is: a method for visual tracking control of a robotic arm for a moving target, comprising the following steps:

[0005] Based on the camera perspective projection model, the unknown motion state information of the moving target is estimated and predicted by the Kalman filter method to obtain the prediction result;

[0006] Based on the visual prediction tracking control method of Kalman filter, the prediction results are designed to obtain the expected joint angular velocity information. Taking into account the nonlinear dynamics of the robot, the expected joint angular velocity information is input into the robot dynamic control system;

[0007] The robot dynamics control method based on adaptive neural network compensates for the uncertain dynamics and external disturbances of the robot system and realizes real-time dynamic tracking of the desired joint angular velocity information.

[0008] Furthermore, the step of estimating and predicting the unknown motion state information of the moving target by a Kalman filter method based on the camera perspective projection model to obtain a prediction result specifically includes:

[0009] The depth information of the feature point image is expressed as a linear form of the state parameters of the moving target;

[0010] According to the linear form of the depth information of the feature point image, the Kalman filter measurement model of the moving target is obtained by transforming the camera perspective projection model;

[0011] The image information of the moving target and the kinematic information of the robot are obtained, and the unknown position and speed of the moving target are predicted and estimated based on the Kalman filter model of the moving target to obtain the prediction result.

[0012] Furthermore, the Kalman filter model of the moving target includes a state equation and a measurement equation, wherein the position and velocity of the moving target at different moments before and after are linked to obtain the state equation of the Kalman filter model of the moving target, and the feature point image position information and robot kinematic information are substituted into the transformed camera perspective projection model to obtain the measurement equation of the Kalman filter model of the moving target.

[0013] Furthermore, the visual prediction tracking control method based on Kalman filtering designs the prediction results to obtain the expected joint angular velocity information, and considers the nonlinear dynamics of the robot to input the expected joint angular velocity information into the robot dynamic control system. The step specifically includes:

[0014] Derivative the camera perspective projection model to obtain the relationship between the feature point image velocity information and the robot's joint angular velocity information;

[0015] Discretize the visual servo model of the robotic arm to obtain a discretized visual servo model;

[0016] A visual prediction tracking controller is designed based on the discretized visual servo model and the target motion state information estimated by the Kalman filter method. The image position information of the feature points and the robot kinematic information are updated to estimate the position and velocity of the moving target at the next moment and obtain the corresponding estimated values.

[0017] Combined with the depth-independent image Jacobian matrix, the design is based on the estimated values ​​of the current position and velocity of the moving target to obtain the expected joint angular velocity information of the robot;

[0018] Considering the nonlinear dynamics of the robot, the expected joint angular velocity information is input into the robot dynamic control system.

[0019] Furthermore, the expression of the visual prediction tracking controller of the discretized visual servoing model is specifically as follows:

[0020]

[0021] e(k+i|k)=y(k+i|k)-y d

[0022] In the above formula, Q and R represent weight matrices, N p represents the prediction time domain, N c represents the control time domain, y(k+i|k) represents the predicted output calculated by the estimated target motion state value, and y d represents the expected image position of the feature point, w(k+i|k) represents the optimal control sequence in the rolling time domain, and e T Represents the image position error of the feature point.

[0023] Furthermore, the nonlinear dynamics of the robot is specifically expressed as follows:

[0024]

[0025] In the above formula, B(q) represents the inertia matrix, represents the centripetal force and Coriolis force matrix, G(q) represents the gravity matrix, τ represents the joint torque, d represents the external disturbance, Indicates the robot's joint angular velocity information.

[0026] Furthermore, the robot dynamics control method based on the adaptive neural network compensates for the uncertain dynamics and external disturbances of the robot system and realizes the step of real-time dynamic tracking of the desired joint angular velocity information, which specifically includes:

[0027] Estimate the unknown terms of the robot's nonlinear dynamics through the RBF neural network and design the estimated weight values ​​of the neural network;

[0028] Define the error value of the robot's desired joint angular velocity information and the robot dynamics controller;

[0029] The robot dynamics controller is substituted into the nonlinear dynamics of the robot. Based on the error value of the robot's expected joint angular velocity information, the robot system is iteratively updated according to the estimated weight value to achieve real-time dynamic tracking of the expected joint angular velocity information.

[0030] Furthermore, the expression of the RBF neural network is specifically as follows:

[0031]

[0032]

[0033] In the above formula, represents the input vector of the RBF neural network, c i and δ i represents the center vector and width of the neuron, η represents the optimal weight value, represents the activation function, and ξ represents a constant.

[0034] Furthermore, the expression of the robot dynamics controller is specifically as follows:

[0035]

[0036] In the above formula, K v represents a positive symmetric matrix, ε represents a constant, B0(q) represents a rough inertia matrix, represents the rough centripetal and Coriolis force matrices, G0(q) represents the rough gravity matrix, represents the estimated value of the unknown quantity of the system, τ represents the joint torque, e v Represents the robot joint angular velocity tracking error.

[0037] The second technical solution adopted by the present invention is: a robot arm visual tracking control system for a moving target, comprising:

[0038] The estimation module estimates and predicts the unknown motion state information of the moving target through the Kalman filter method based on the camera perspective projection model to obtain the prediction result;

[0039] The kinematic control module designs the prediction results based on the visual prediction tracking control method of Kalman filter to obtain the expected joint angular velocity information. Taking the nonlinear dynamics of the robot into consideration, the expected joint angular velocity information is input into the robot dynamic control system.

[0040] The dynamics control module uses a robot dynamics control method based on an adaptive neural network to compensate for the uncertain dynamics and external interference of the robot system, and realizes real-time dynamic tracking of the desired joint angular velocity information.

[0041] The beneficial effects of the method and system of the present invention are as follows: the present invention can handle the unknown motion problem of the target by incorporating the Kalman filter method into the visual prediction tracking control, and estimate and predict the unknown position and speed of the moving target. In order to consider the nonlinearity and uncertain dynamics of the robot, the visual prediction tracking control method based on Kalman filtering is combined with the robot dynamic control method based on adaptive neural network. When dealing with uncertain dynamics, the robot dynamic model does not need to meet the parameter linearization conditions, and can simultaneously handle the unknown motion of the target, uncertain dynamics, unmodeled dynamics and external disturbance problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a flowchart of the steps of a visual tracking control method of a robotic arm for a moving target according to the present invention;

[0043] Figure 2 This is a structural block diagram of a robotic arm visual tracking control system for a moving target according to the present invention;

[0044] Figure 3 It is a structural diagram of the visual tracking control system of the robotic arm of the present invention;

[0045] Figure 4 Schematic diagram of the structure of the RBF neural network of a specific embodiment of the present invention;

[0046] Figure 5 is a schematic diagram of a simulation of image errors of feature points on an image plane in a specific embodiment of the present invention;

[0047] Figure 6 1 is a schematic diagram of a simulation of the trajectory of the estimated position of a moving target in a specific embodiment of the present invention;

[0048] Figure 7 1 is a schematic diagram of a simulation of a trajectory of estimated velocity of a moving target in a specific embodiment of the present invention;

[0049] Figure 8 1 is a schematic diagram of a simulation of changes in weight estimation values ​​of an adaptive neural network in a specific embodiment of the present invention;

[0050] Figure 9 is a schematic diagram of a simulation of image errors of feature points on an image plane in a specific embodiment of the present invention;

[0051] Figure 10 1 is a schematic diagram of a simulation of the trajectory of the estimated position of a moving target in a specific embodiment of the present invention;

[0052] Figure 11 1 is a schematic diagram of a simulation of a trajectory of estimated velocity of a moving target in a specific embodiment of the present invention;

[0053] Figure 12 It is a schematic diagram of a simulation of the change of the weight estimation value of the adaptive neural network in a specific embodiment of the present invention. DETAILED DESCRIPTION

[0054] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. The step numbers in the following embodiments are provided for ease of description only and do not limit the order of the steps. The order of execution of the steps in the embodiments can be adaptively adjusted based on the understanding of those skilled in the art.

[0055] Reference Figure 1 and Figure 3 The present invention provides a method for visual tracking control of a robot arm for a moving target, the method comprising the following steps:

[0056] S1. Transform the camera perspective projection model to obtain the Kalman filter model of the moving target;

[0057] S2, Kalman filter predicts and estimates the unknown motion state of the target based on the image feature coordinate information and robot kinematic information;

[0058] Specifically, the image coordinates of the feature point are y=[u,v] T , under camera perspective projection, the image coordinates can be expressed as:

[0059]

[0060] In the above formula, z c Represents the depth information of the feature point, M represents the internal and external parameter matrix of the camera, T represents the robot forward kinematics matrix, and p represents the position coordinates of the feature point;

[0061] Among them, the depth information of the feature point can be expressed as:

[0062]

[0063] In the above formula, represents the i-th row of the camera internal and external matrix M;

[0064] In summary, the Kalman filter state model of the moving target is expressed as:

[0065] x k =Ax k-1 +γ k

[0066] In the above formula, represents the state vector, X, Y and Z represent the position of the moving target, and represents the speed of the moving target, γ k represents process noise, which has a zero-mean Gaussian distribution and a covariance matrix of E k ;

[0067] in,

[0068]

[0069] Substitute the depth information of the feature points into the camera perspective projection model and transform it to obtain the measurement model of the Kalman filter:

[0070] z k =Cx k +ξ k

[0071] in,

[0072]

[0073]

[0074] In the above formula, ξ k represents the measurement noise, which has a zero-mean Gaussian distribution and a covariance matrix of F k ,ρ i =(m i1 m i2 m i3 ), R i =(r 1i r 2i r 3i ) T , β=(t1 t2 t3 1) T , m ij is the (i, j)th element of the camera's internal and external parameter matrix, r ii is the (i, j)th element of the robot’s forward kinematics rotation matrix R, t i is the i-th element of the robot’s forward kinematic displacement vector t;

[0075] The recursive Kalman filter algorithm consists of prediction and estimation parts, and the prediction update is:

[0076]

[0077] P k+1,k =AP k,k A T +E k

[0078] Estimated update:

[0079]

[0080] P k+1,k+1 =P k+1,k -K k+1 CP k+1,k

[0081] The Kalman gain matrix at time k+1 is:

[0082]

[0083] The position and velocity of the moving target are estimated based on the Kalman filter.

[0084] S3, the visual prediction tracking control algorithm based on Kalman filtering calculates the control variable according to the estimated unknown motion state of the target and outputs the expected joint angular velocity to the robot system;

[0085] Specifically, the visual prediction tracking control algorithm based on Kalman filtering uses the estimated target motion state to calculate the control variable and outputs the desired joint angular velocity to the robot system;

[0086] The relationship between the velocity of the feature point in the image plane and the joint velocity is obtained by deriving the camera perspective projection model:

[0087]

[0088] In the above formula, q represents the joint angle of the robot, is the joint angular velocity of the robot;

[0089] Among them, the depth-independent image Jacobian matrix L is:

[0090]

[0091] In order to design a visual tracking controller using the model predictive control method, the visual servo model is discretized and the estimated target motion state is used to obtain:

[0092]

[0093] In the above formula, T e represents the sampling time, Represents the estimated feature point depth information, represents the estimated depth-independent image Jacobian matrix;

[0094] The control quantity is calculated using a visual prediction tracking controller, which can be described as:

[0095]

[0096] In the above formula, Q>0, R>0 represents the weight matrix, N p represents the prediction time domain, N c represents the control time domain, e(k+i|k)=y(k+i|k)-y d , y(k+i|k) represents the predicted output calculated by the estimated target motion state value, y d represents the desired image position of the feature point, w(k+i|k) is the optimal control sequence in the rolling time domain, and w(k)=w(k|k) is the first element of the optimal control sequence and is applied to the system;

[0097] The desired joint angular velocity is designed to be:

[0098]

[0099] Among them, the reference velocity of the feature point in the image plane is:

[0100]

[0101] In the above formula, is the expected feature point velocity in the image plane, μ is a constant;

[0102] The estimated compensated depth-independent image Jacobian matrix is:

[0103]

[0104] and yes The pseudo-rebellion, The expression is:

[0105]

[0106] S4. The dynamic controller based on adaptive neural network estimates and compensates for uncertain dynamics and external disturbances, and tracks the desired joint angular velocity.

[0107] Specifically, the robot dynamics controller based on adaptive neural networks estimates and compensates for the system's uncertain dynamics and external disturbances, and tracks the desired joint angular velocity.

[0108] The dynamic equation of the robot is:

[0109]

[0110] In the above formula, B(q) represents the inertia matrix, represents the centripetal force and Coriolis force matrices, G(q) represents the gravity matrix, τ represents the joint torque, and d represents the external disturbance;

[0111] However, it is difficult to obtain an accurate dynamic model of the robot, and external disturbances often exist in the robot system. Usually, only a rough robot dynamic model can be obtained, using B0(q), and G0(q), its expression is:

[0112]

[0113] In summary, the calculation is

[0114] in

[0115] In fact, f(·) is unknown and needs to be estimated and compensated. RBF neural network is used to estimate and compensate the unknown nonlinear term f(·) online. The structure diagram of RBF neural network is shown in the figure below. Figure 4 As shown;

[0116] The RBF neural network algorithm can be described as:

[0117]

[0118]

[0119] In the above formula, x represents the input vector of the RBF neural network, c i and δ i represents the center vector and width of the neuron, η represents the optimal weight value, represents the activation function, ξ is a very small constant, and the output of the RBF neural network can be used to approximate in;

[0120]

[0121] in represents the estimated value of the weight matrix η, ||η|| F ≤η max ,and

[0122]

[0123] The robot joint angular velocity tracking error is defined as:

[0124]

[0125] In order to deal with uncertain dynamics and external disturbances, a robot velocity tracking dynamics controller based on an adaptive neural network is designed to estimate system uncertainty and compensate for it. The input vector of the RBF adaptive neural network is defined as The controller is designed to:

[0126]

[0127] In the above formula, K v represents a positive symmetric matrix, ε represents a constant, B0(q) represents a rough inertia matrix, represents the rough centripetal and Coriolis force matrices, G0(q) represents the rough gravity matrix, Represents the estimated value of the unknown quantity of the system;

[0128] Substituting the above formula into the robot dynamics equation, we get;

[0129]

[0130] Subtract the left and right sides of the above equation get;

[0131]

[0132] in:

[0133]

[0134] The adaptive update law of the RBF neural network weights is designed as:

[0135]

[0136] Where λ>0.

[0137] The simulation experiment of the present invention is as follows:

[0138] In order to verify the effectiveness of the proposed method, a simulation is performed on the robotic arm eye-in-hand vision system. The transformation matrix of the robotic arm end effector coordinate system relative to the camera coordinate system is:

[0139]

[0140] The position of the feature point relative to the robot base coordinate system is (0.06, 0.05, 1.5) T m, in order to show that the proposed method can track targets with unknown motion, the control purpose is to lock the moving target at a specific position in the image plane. The rough length parameter of the robot arm is Rough quality parameters of the robotic arm The rough moment of inertia parameters of the robot arm are The process noise covariance and measurement noise covariance in the Kalman filter are E k =10 -7 I 6×6 , F k =10 -7 I 2×2 , the weight matrix of visual prediction tracking control is Q = 0.0001I 2×2 , R=0.0001I 6×6 , the number of neurons in the hidden layer is 5, and the center vector c of the Gaussian kernel function is i is [-2 -1 0 1 2], and the width value δ i is 3, when the target's moving speed is 5cm / s, the image error of the feature point on the image plane is as follows Figure 5 As shown, the estimated position and velocity of the moving target are Figure 6 and7 As shown, the weight estimates of the adaptive neural network are as follows Figure 8 As shown, in order to further verify the robustness and anti-interference ability of the proposed method, the amplitude is [3,3] T When the external interference N is added to the input of the robot arm, when the target is at a higher moving speed of 10 cm / s, the image error of the feature point on the image plane is as follows: Figure 9 As shown, the estimated target position and velocity are Figure 10 and 11 As shown, the weight estimates of the adaptive neural network are as follows Figure 12 As shown in Figure 3, the results show that the proposed method can handle unknown target motion, uncertain dynamics and external disturbances well.

[0141] Reference Figure 2 , a robot arm visual tracking control system for a moving target, comprising:

[0142] The estimation module estimates and predicts the unknown motion state information of the moving target through the Kalman filter method based on the camera perspective projection model to obtain the prediction result;

[0143] The kinematic control module designs the prediction results based on the visual prediction tracking control method of Kalman filter to obtain the expected joint angular velocity information. Taking the nonlinear dynamics of the robot into consideration, the expected joint angular velocity information is input into the robot dynamic control system.

[0144] The dynamics control module uses a robot dynamics control method based on an adaptive neural network to compensate for the uncertain dynamics and external interference of the robot system, and realizes real-time dynamic tracking of the desired joint angular velocity information.

[0145] The contents of the above method embodiments are all applicable to the present system embodiments. The functions specifically implemented by the present system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0146] The above is a specific description of the preferred implementation of the present invention, but the invention is not limited to the embodiments. Those skilled in the art can make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.

Claims

1. A method for visual tracking control of a robotic arm for a moving target, characterized in that: The following steps are involved: Based on the camera perspective projection model, the unknown motion state information of the moving target is estimated and predicted by the Kalman filter method to obtain the prediction result; Based on the visual prediction tracking control method of Kalman filter, the prediction results are designed to obtain the expected joint angular velocity information. Taking into account the nonlinear dynamics of the robot, the expected joint angular velocity information is input into the robot dynamic control system; The robot dynamics control method based on adaptive neural network compensates for the uncertain dynamics and external disturbances of the robot system and realizes real-time dynamic tracking of the desired joint angular velocity information. The visual prediction and tracking control method based on Kalman filtering designs the prediction results to obtain the expected joint angular velocity information, and considers the nonlinear dynamics of the robot to input the expected joint angular velocity information into the robot dynamic control system. The step specifically includes: Derivative the camera perspective projection model to obtain the relationship between the feature point image velocity information and the robot's joint angular velocity information; Discretize the visual servo model of the robotic arm to obtain a discretized visual servo model; A visual prediction tracking controller is designed based on the discretized visual servo model and the target motion state information estimated by the Kalman filter method. The image position information of the feature points and the robot kinematic information are updated to estimate the position and velocity of the moving target at the next moment and obtain the corresponding estimated values. Combined with the depth-independent image Jacobian matrix, the design is based on the estimated values ​​of the current position and velocity of the moving target to obtain the expected joint angular velocity information of the robot; Considering the nonlinear dynamics of the robot, the expected joint angular velocity information is input into the robot dynamic control system.

2. A method for visual tracking control of a moving target by a robotic arm according to claim 1, characterized in that: The step of estimating and predicting the unknown motion state information of the moving target by the Kalman filter method based on the camera perspective projection model to obtain the prediction result specifically includes: The depth information of the feature point image is expressed as a linear form of the state parameters of the moving target; According to the linear form of the depth information of the feature point image, the Kalman filter measurement model of the moving target is obtained by transforming the camera perspective projection model; The image information of the moving target and the kinematic information of the robot are obtained, and the unknown position and speed of the moving target are predicted and estimated based on the Kalman filter model of the moving target to obtain the prediction result.

3. A method for visual tracking control of a moving target by a robotic arm according to claim 2, characterized in that: The Kalman filter model of the moving target includes a state equation and a measurement equation, wherein the state equation of the Kalman filter model of the moving target is obtained by connecting the position and velocity of the moving target at different moments before and after, and the measurement equation of the Kalman filter model of the moving target is obtained by substituting the feature point image position information and the robot kinematic information into the transformed camera perspective projection model.

4. A method for visual tracking control of a moving target by a robotic arm according to claim 3, characterized in that: The expression of the visual prediction tracking controller of the discretized visual servo model is specifically as follows: e(k+i|k)=y(k+i|k)-y d In the above formula, Q and R represent weight matrices, N p represents the prediction time domain, N c represents the control time domain, y(k+i|k) represents the predicted output calculated by the estimated target motion state value, and y d represents the expected image position of the feature point, w(k+i|k) represents the optimal control sequence in the rolling time domain, and e T Represents the image position error of the feature point.

5. A method for visual tracking control of a moving target by a robotic arm according to claim 4, characterized in that: The nonlinear dynamics of the robot is expressed as follows: In the above formula, B(q) represents the inertia matrix, represents the centripetal force and Coriolis force matrix, G(q) represents the gravity matrix, τ represents the joint torque, d represents the external disturbance, Indicates the robot's joint angular velocity information.

6. A method for visual tracking control of a moving target by a robotic arm according to claim 5, characterized in that: The robot dynamics control method based on the adaptive neural network compensates for the uncertain dynamics and external disturbances of the robot system and realizes the step of real-time dynamic tracking of the desired joint angular velocity information, which specifically includes: Estimate the unknown terms of the robot's nonlinear dynamics through the RBF neural network and design the estimated weight values ​​of the neural network; Define the error value of the robot's desired joint angular velocity information and the robot dynamics controller; The robot dynamics controller is substituted into the nonlinear dynamics of the robot. Based on the error value of the robot's expected joint angular velocity information, the robot system is iteratively updated according to the estimated weight value to achieve real-time dynamic tracking of the expected joint angular velocity information.

7. A method for visual tracking control of a moving target by a robotic arm according to claim 6, characterized in that: The expression of the RBF neural network is specifically as follows: In the above formula, represents the input vector of the RBF neural network, c i and δ i represents the center vector and width of the neuron, η represents the optimal weight value, represents the activation function, and ξ represents a constant.

8. A method for visual tracking control of a moving target by a robotic arm according to claim 7, characterized in that: The expression of the robot dynamics controller is as follows: In the above formula, K v represents a positive symmetric matrix, ε represents a constant, B0(q) represents a rough inertia matrix, represents the rough centripetal and Coriolis force matrices, G0(q) represents the rough gravity matrix, represents the estimated value of the unknown quantity of the system, τ represents the joint torque, e v Represents the robot joint angular velocity tracking error.

9. A robotic arm visual tracking control system for a moving target, characterized in that: The method for executing a visual tracking control method of a robotic arm for a moving target according to claim 1 comprises the following modules: The estimation module estimates and predicts the unknown motion state information of the moving target through the Kalman filter method based on the camera perspective projection model to obtain the prediction result; The kinematic control module designs the prediction results based on the visual prediction tracking control method of Kalman filter to obtain the expected joint angular velocity information. Taking the nonlinear dynamics of the robot into consideration, the expected joint angular velocity information is input into the robot dynamic control system. The dynamics control module uses a robot dynamics control method based on an adaptive neural network to compensate for the uncertain dynamics and external interference of the robot system, and realizes real-time dynamic tracking of the desired joint angular velocity information.

Images