A robot image visual servo robust control method based on adaptive update law

Abstract

This invention discloses a robust visual servo control method for robots based on an adaptive update law, comprising: associating the pixel coordinates of feature points with their 3D coordinates in the camera coordinate system according to a pinhole camera projection model, and establishing a global linear regression model containing unknown position parameter vectors; designing an adaptive update law for the estimated values ​​of the unknown position parameter vectors based on the global linear regression model, used to estimate the unknown depth information parameters of feature points in real time, and reconstructing the visual Jacobian matrix online based on the estimated depth information parameters; constructing a visual state observer to obtain smoothed image feature estimates; defining a joint space-assisted tracking error based on the smoothed image feature estimates based on the reconstructed visual Jacobian matrix; and designing an integral sliding mode robust feedback term based on the integral sliding mode in the robot joint driving torque expression based on the lumped disturbance in robot dynamics, combined with feedforward compensation, to generate the final joint driving torque to drive the robot to perform servo tracking tasks. This invention can achieve highly robust and high-precision asymptotic tracking of target feature points under complex working conditions.

Description

Technical Field

[0001] This invention relates to a robust control method for robot image vision servoing based on an adaptive update law, belonging to the fields of industrial robot control, machine vision and intelligent automation technology. Background Technology

[0002] With the popularization of automation technology, visual servoing technology has become a core technology for improving the autonomy of industrial robots in complex and unstructured environments. Among them, image-based visual servoing (IBVS) constructs control laws directly in the feature space of a two-dimensional image, and compared with position-based visual servoing methods, it exhibits stronger robustness to camera calibration errors and robot kinematic uncertainties.

[0003] In existing technologies, image-based visual servo controllers typically employ linear error feedback strategies such as proportional-derivative (PD) to adjust image feature errors. These methods are simple in structure, easy to implement, and can achieve a certain degree of image tracking control in ideal environments. However, PD control is essentially a linear feedback adjustment method, and its control performance is highly dependent on the accuracy of the system model and the stability of the external environment. When the robot is in a complex working scenario, relying solely on linear feedback of image errors is insufficient to fully compensate for the nonlinear effects caused by the robot itself and the external environment.

[0004] In reality, robots are inherently highly nonlinear, strongly coupled, and complex dynamic systems. Lumped disturbances such as joint static friction, unmodeled dynamic characteristics, and external time-varying loads are prevalent and difficult to model accurately. The image-based visual servoing methods described above, based on simple linear error feedback strategies, lack effective suppression of unstructured dynamic disturbances. When the system is affected by external disturbances or model uncertainties, it often struggles to eliminate steady-state errors, which can even lead to closed-loop performance degradation and poor dynamic robustness. Furthermore, existing visual servoing control frameworks struggle to effectively integrate visual parameter estimation with dynamic disturbance rejection control, failing to overcome the impacts of unknown depth parameters, visual measurement uncertainties, and unstructured physical disturbances at the robot's underlying structure, especially in the absence of prior 3D geometric information.

[0005] Therefore, in view of the shortcomings of existing technologies, there is an urgent need to design a robust robot vision servoing method that can maintain a high level of anti-interference capability under harsh conditions such as missing depth information and complex dynamics and external interference, and achieve stable and high-precision asymptotic tracking of visual targets. This has important practical application value for promoting the industrial application of robot vision technology. Summary of the Invention

[0006] To address the issues of insufficient system robustness caused by the lack of 3D depth information parameters, complex nonlinear dynamic characteristics of robots, and strong external environmental interference in robot visual servoing tasks, this invention provides a robust control method for robot image visual servoing based on an adaptive update law. It aims to achieve high robustness and high precision asymptotic tracking of target feature points under complex working conditions through the collaborative design of online depth information parameter reconstruction and nonlinear disturbance rejection (external disturbance, friction) control.

[0007] The technical solution of this invention is:

[0008] According to a first aspect of the present invention, a robust control method for robot image visual servoing based on an adaptive update law is provided, comprising:

[0009] Step S1: Extract image feature points of the target object. Based on the pinhole camera projection model, associate the pixel coordinates of the feature points with their 3D coordinates in the camera coordinate system, and establish a global linear regression model containing unknown position parameter vectors through the linear parameterization method.

[0010] Step S2: Based on the global linear regression model, design an adaptive update law for the unknown location parameter vector estimate, which is used to estimate the unknown depth information parameters of feature points in real time, and reconstruct the visual Jacobian matrix online based on the depth information parameter estimate.

[0011] Step S3: Construct a visual state observer, and use the reconstructed visual Jacobian matrix to estimate the image feature vector constructed from the pixel coordinates of the feature points to obtain smooth image feature estimates.

[0012] Step S4: Based on the reconstructed visual Jacobian matrix, define the normalized Jacobian transpose mapping matrix; based on the normalized Jacobian transpose mapping matrix, define the joint space-assisted tracking error based on the smoothed image feature estimation.

[0013] Step S5: For the lumped disturbance in robot dynamics, design an integral sliding mode robust feedback term based on the integral sliding mode in the expression of robot joint driving torque according to the joint space auxiliary tracking error, and combine it with feedforward compensation to generate the final joint driving torque to drive the robot to perform servo tracking task.

[0014] Further, step S1 specifically includes:

[0015] For a A robot visual servoing system with degrees of freedom defines the image feature points of the target object extracted by the camera on the robot visual servoing system. The pixel value in the horizontal direction of the pixel plane is The vertical pixel value Based on the pinhole camera projection model, feature points are... pixel coordinates Its relation to its 3D coordinates in the camera coordinate system:

[0016] ;

[0017] in, For feature points homogeneous pixel coordinates; For feature points Relative to the camera's depth information parameters, The third row represents the camera intrinsic parameter matrix. and These are the rotation matrix and translation vector from the robot's base coordinate system to the camera coordinate system, respectively. For feature points Constant coordinates in the base coordinate system; This represents the intrinsic parameter matrix of the camera; Indicates transpose;

[0018] Define auxiliary matrix The pinhole camera projection model is transformed into a linear parameterized form:

[0019] ;

[0020] in, For the output vector, This is the regression matrix;

[0021] Will Augmenting the feature points yields a global linear regression model:

[0022] ;

[0023] in, The augmented output vector, The augmented regression matrix, It is an unknown position parameter vector, that is, an augmented constant coordinate vector.

[0024] Further, step S2 specifically includes:

[0025] Based on the global linear regression model, an auxiliary matrix with a forgetting factor is introduced. and auxiliary vector The first derivative expressions for the two are as follows:

[0026] ;

[0027] in, Forgetting factor; This is the augmented output vector in the global linear regression model. This is the augmented regression matrix in the global linear regression model;

[0028] Based on the auxiliary matrix with forgetting factor and auxiliary vector Construct error variables related to estimation error :

[0029] ;

[0030] in, unknown position parameter vector Estimated value The estimation error of the unknown position parameter vector;

[0031] With error variable To drive this process, the adaptive update law for the estimated unknown position parameter vector is designed as follows:

[0032] ;

[0033] in, It is a symmetric positive definite adaptive gain matrix;

[0034] Integrating the adaptive update law allows for online updating of the estimated value of the unknown position parameter vector. And then based on Obtain estimated values ​​of depth information parameters. Then, the visual Jacobian matrix is ​​reconstructed based on the estimated values ​​of the depth information parameters. ;in, The third row represents the camera intrinsic parameter matrix. and These are the rotation matrix and translation vector from the robot's base coordinate system to the camera coordinate system, respectively. express The estimated value, For feature points Constant coordinates in the base coordinate system.

[0035] Further, step S3 specifically includes:

[0036] Define the global observation error as Based on the global observation error and the reconstructed visual Jacobian matrix, the visual state observer is constructed as follows:

[0037] ;

[0038] in, express The first derivative, for The estimated value, For image feature vectors, Representing feature points pixel coordinates transpose, , Indicates the number of feature points; For the reconstructed visual Jacobian matrix; For the robot's joint angular velocity, Indicates the position of the robot's joint angles; This is the gain matrix of the positive definite observer.

[0039] Furthermore, S4 specifically includes:

[0040] Define the normalized Jacobian transpose mapping matrix based on the reconstructed visual Jacobian matrix. :

[0041] ;

[0042] in, Represents the reconstructed visual Jacobian matrix transpose, Normalization factor;

[0043] Based on the normalized Jacobian transpose mapping matrix, the joint space auxiliary tracking error is defined. :

[0044] ;

[0045] in, For the robot's joint angular velocity; To control the gain; For the tracking error in the defined image space, For smooth image feature estimates, These are the desired pixel coordinates.

[0046] Further, step S5 specifically includes:

[0047] Robot joint drive torque for:

[0048] ;

[0049] in, For gravity compensation, For Coriolis force and centrifugal force terms, Represents the matrix of Coriolis force and centrifugal force. For the robot's joint angular velocity, Indicates the position of the robot's joint angles. For integral sliding mode robust feedback; based on joint space auxiliary tracking error The integral sliding mode robust feedback term in the robot joint driving torque expression is designed as follows:

[0050] ;

[0051] in, For design gain, robust term gain , This represents the lumped disturbance derivative term, which is composed of the rate of change of external disturbance and the rate of change of frictional force when moving along the desired trajectory. Represents the infinite norm; For symbolic functions, Joint space auxiliary tracking error exist The initial value at time.

[0052] According to a second aspect of the present invention, a robust control system for robot image visual servoing based on an adaptive update law is provided, comprising modules of the robust control method for robot image visual servoing based on an adaptive update law as described in any of the preceding claims.

[0053] According to a third aspect of the present invention, a processor is provided for running a program that, when running, performs the steps of the robot image visual servo robust control method based on adaptive update law as described in any one of the preceding claims.

[0054] The beneficial effects of this invention are as follows: Compared with existing visual servo control technologies, the technical solution proposed in this invention achieves rapid online reconstruction of depth information parameters through an adaptive update law driven by position parameter estimation errors, fundamentally solving the problem of IBVS's dependence on prior depth information; combining a visual state observer and a normalized Jacobian transpose mapping, this invention completely eliminates the risk of singularity from a mathematical structure perspective while suppressing image noise; the core driving torque controller can effectively absorb robot friction and complex external disturbances, achieving high-precision asymptotic tracking with zero steady-state error while ensuring smooth control torque and no jitter, significantly improving the robustness and operational performance of the visual servo system in complex industrial environments. Attached Figure Description

[0055] Figure 1 This is a schematic diagram of an eye-on-hand robot system.

[0056] Figure 2 This is a block diagram of a robot image vision servo robust control based on an adaptive update law according to an embodiment of the present invention.

[0057] Figure 3A flowchart of a robot image vision servo robust control method based on an adaptive update law;

[0058] Figure 4 This is a schematic diagram illustrating the convergence effect of adaptive depth estimation according to the embodiment.

[0059] Figure 5 This is a schematic diagram of the state estimation error norm provided according to the embodiment;

[0060] Figure 6 This is a schematic diagram of the pixel tracking trajectory provided according to the embodiment;

[0061] The numbers in the diagram are: 1-camera, 2-robotic arm, 3-target object, 4-base. Detailed Implementation

[0062] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be arbitrarily combined with each other.

[0063] Figure 1 This is a simplified diagram of an image-based robot visual servoing system with the eye on the hand. As shown, camera 1 is fixed to the end of robotic arm 2. Based on the pixel coordinates of the target object 3 acquired by camera 1, the robot joint driving torque is designed to control the target object to reach the desired pixel coordinates within the camera's imaging plane. However, traditional image-based visual servoing control methods do not consider external disturbances and unmodeled robot dynamics (such as frictional changes), making it difficult to achieve high control accuracy under complex force control conditions. To improve the robustness of image-based visual servoing control systems, this invention proposes a novel robust control method for robot image-based visual servoing based on an adaptive update law.

[0064] like Figures 2-6 As shown, according to a first aspect of the present invention, a robust control method for robot image visual servoing based on an adaptive update law is provided, the specific implementation process of which is as follows:

[0065] Step S1: Extract image feature points of the target object. Based on the pinhole camera projection model, associate the pixel coordinates of the feature points with their 3D coordinates in the camera coordinate system, and establish a global linear regression model containing unknown position parameter vectors through the linear parameterization method.

[0066] Step S2: Based on the global linear regression model, design an adaptive update law for the unknown location parameter vector estimate, which is used to estimate the unknown depth information parameters of feature points in real time, and reconstruct the visual Jacobian matrix online based on the depth information parameter estimate.

[0067] Step S3: Construct a visual state observer, and estimate the image feature vector constructed from the pixel coordinates of the feature points using the reconstructed visual Jacobian matrix to obtain smooth image feature estimates. ;

[0068] Step S4: Based on the reconstructed visual Jacobian matrix, define the normalized Jacobian transpose mapping matrix; based on the normalized Jacobian transpose mapping matrix, define the joint space auxiliary tracking error based on the smoothed image feature estimation value; the normalized Jacobian transpose mapping matrix introduced in this invention can map the tracking error in the image space to the joint space of the robotic arm by normalizing the reconstructed visual Jacobian matrix, thereby generating a joint space auxiliary tracking error signal.

[0069] Step S5: For the lumped disturbance in robot dynamics, design an integral sliding mode robust feedback term based on the joint space auxiliary tracking error in the robot joint driving torque expression, and combine it with feedforward compensation (gravity compensation term, Coriolis force and centrifugal force term) to generate the final joint driving torque to drive the robot to perform servo tracking task.

[0070] It should be noted that before designing the joint drive torque, the intrinsic and extrinsic parameters of the visual servo system can be calibrated based on the robot's forward kinematics and hand-eye calibration methods to obtain the camera's intrinsic parameter matrix. extrinsic rotation matrix and extrinsic translation vector .

[0071] For example, consider as follows Figure 1 The image-based robot vision servoing system shown depicts an eye on a hand, where the robot has 7 degrees of freedom. The robot includes a base 4 and a robotic arm 2 mounted on the base. A camera 1 is mounted at the end of the robotic arm 2. Four red dots are shown on the target object 3 as target feature points. (Combined with...) Figure 2 , Figure 3 The robust control method for robot image visual servoing based on adaptive update law provided by the present invention is further described below:

[0072] I. Constructing a global linear regression model

[0073] Extracting target objects using a camera The pixel coordinates of image feature points. Among these feature points... The pixel coordinates are Its homogeneous pixel coordinate form Based on the pinhole camera projection model, feature points Pixel coordinate values ​​satisfy:

[0074] ;

[0075] in, For feature points homogeneous pixel coordinates, express Transpose of; For feature points Relative to the camera's depth information parameters, , and These are the rotation matrix and translation vector from the robot's base coordinate system to the camera coordinate system, respectively. For feature points Constant three-dimensional coordinates in the base coordinate system; The intrinsic parameter matrix of the camera ( (Represents the third row of the intrinsic parameter matrix); The dimension is The same applies to others; Indicates transpose;

[0076] Define auxiliary matrix The pinhole camera projection model is transformed into a linear parameterized form:

[0077] ;

[0078] in, For the output vector, This is the regression matrix;

[0079] Will Augmenting the feature points yields a global linear regression model:

[0080] ;

[0081] in, For the augmented output vector ( express (The transpose of the others, and so on). The augmented regression matrix, For unknown position parameter vectors ( express (The transpose of the vector, and so on, is the augmented constant coordinate vector.)

[0082] Second, design an adaptive update law for the estimated value of the unknown position parameter vector, specifically as follows:

[0083] Based on the global linear regression model, an auxiliary matrix with a forgetting factor is introduced. and auxiliary vector The first derivative expressions for both are as follows (for simplicity, the parameter related to time has been simplified, for example, ...). , , , Abbreviated as , , , ):

[0084] ;

[0085] in, This is a forgetting factor used to prevent integral saturation (taken as 0.1 in this embodiment of the invention);

[0086] Based on the auxiliary matrix with forgetting factor and auxiliary vector Construct error variables related to estimation error :

[0087] ;

[0088] in, unknown position parameter vector Estimated value The estimation error of the unknown position parameter vector;

[0089] With error variable To drive this process, the adaptive update law for the estimated unknown position parameter vector is designed as follows:

[0090] ;

[0091] in, It is a symmetric positive definite adaptive gain matrix;

[0092] Integrating the adaptive update law allows for online updating of the estimated value of the unknown position parameter vector. And then based on Obtain estimated values ​​of depth information parameters. Then, the visual Jacobian matrix is ​​reconstructed based on the estimated values ​​of the depth information parameters. ;in, The third row represents the camera intrinsic parameter matrix. and These are the rotation matrix and translation vector from the robot's base coordinate system to the camera coordinate system, respectively. express The estimated value, For feature points Constant coordinates in the base coordinate system.

[0093] The reconstructed visual Jacobian matrix expression above is: ;in, It is the Jacobian matrix of the image that maps translational motion. It is the Jacobian matrix of the image that maps rotational motion. It is the Jacobian matrix that maps the robot's joint space to its workspace.

[0094] As can be seen from the above technical solution, the method of the present invention differs from the traditional gradient descent method, as it is directly driven by parameter estimation error to obtain the position estimate. Then, estimates of the depth information parameters are obtained accordingly. Based on this, the visual Jacobian matrix is ​​reconstructed. It has faster convergence characteristics.

[0095] III. Construct a visual state observer to obtain smooth image feature estimates. Specifically:

[0096] Define the global observation error as (For the sake of simplicity, the following...) Abbreviated as Based on the global observation error and the reconstructed visual Jacobian matrix, the visual state observer is constructed as follows:

[0097] ;

[0098] in, express The first derivative, for The estimated value (i.e., the smoothed image feature estimate), This is the image feature vector (i.e., the augmented image feature vector based on the pixel coordinates of feature points). Representing feature points pixel coordinates transpose, , This refers to global observation errors; For the reconstructed visual Jacobian matrix; The angular velocity of the robot joints (i.e. (first derivative), Indicates the position of the robot's joint angles; The gain matrix of the positive definite observer (taken in the embodiment of the invention) ), Represents the identity matrix.

[0099] IV. Define the joint space auxiliary tracking error as follows:

[0100] Based on the reconstructed visual Jacobian matrix, the normalized Jacobian transpose mapping matrix is ​​defined as follows:

[0101] ;

[0102] in, Represents the reconstructed visual Jacobian matrix transpose, Normalization factor; express Summing the elements on the main diagonal.

[0103] Based on the normalized Jacobian transpose mapping matrix, the joint space auxiliary tracking error is defined. :

[0104] ;

[0105] in, The angular velocity of the robot joints (i.e. (first derivative), Indicates the position of the robot's joints; The gain is a normal number (taken as 2 in this embodiment of the invention); For the tracking error in the defined image space, for The estimated value, For image feature vectors, Representing feature points pixel coordinates transpose, , These are the desired pixel coordinates.

[0106] As can be seen from the above. Used to Tracking error in 3D image space is transformed into The auxiliary tracking error of the dimension is reduced, and due to its normalization property, it can essentially avoid the singularity problem that may occur in the parameter convergence process of traditional pseudo-inverse methods.

[0107] V. Design a robot joint drive torque controller, specifically:

[0108] Considering the unmodeled dynamic disturbances such as external time-varying load disturbances and frictional forces experienced by the robot's robotic arm, the robot joint drive torque controller is designed as follows:

[0109] ;

[0110] in, For gravity compensation, For Coriolis force and centrifugal force terms, Represents the matrix of Coriolis force and centrifugal force. For integral sliding mode robust feedback; based on joint space auxiliary tracking error The integral sliding mode robust feedback term in the robot joint driving torque expression is designed as follows:

[0111] ;

[0112] in, , For design gain (in embodiments of the invention) Take 1, Take 10), For robustness term gain; To assist in tracking errors in joint space exist The initial value at time t, and other values ​​are similar; It is a symbolic function.

[0113] Stability analysis based on Lyapunov theory proves that the controller can ensure tracking error in the image space. Global observation error and position parameter vector estimation error All converge asymptotically to zero.

[0114] The controller considers A robot dynamics model with degrees of freedom can be described in the following form:

[0115] ;

[0116] in, , and These are the robot's joint angular positions, velocities, and accelerations, respectively. The inertia matrix; The matrix represents the Coriolis force and the centrifugal force. It is the gravity vector; It is the driving torque of the robot joint, i.e., the input torque; Friction; This is an external disturbance.

[0117] Furthermore, the stability of the controller is analyzed using Lyapunov theory, with the following specific steps:

[0118] Define auxiliary variables and They are respectively:

[0119] ;

[0120] ;

[0121] in, Representing vectors The j-th element in The lumped disturbance derivative term, which is composed of the rate of change of external disturbance and the rate of change of frictional force when moving along the desired trajectory, is defined as follows:

[0122] ;

[0123] in, It is the desired joint angular velocity. This represents the rate of change of external disturbances. This represents the rate of change of frictional force when moving along the desired trajectory. Assume... and its first derivative Bounded, through design, to make the gain This can guarantee the auxiliary variables .

[0124] The Lyapunov function is defined as follows:

[0125] ;

[0126] right The derivative ultimately yields the following form:

[0127] ;

[0128] in, , It is a constant greater than 0. Therefore, the tracking error in the image space... Joint space auxiliary tracking error Position parameter vector estimation error and global observation error All of them will converge to 0.

[0129] According to a second aspect of the present invention, a robust control system for robot image visual servoing based on an adaptive update law is provided, comprising modules of the robust control method for robot image visual servoing based on an adaptive update law described in any of the preceding embodiments. Specifically, it comprises: a first module for performing step S1: extracting image feature points of the target object, associating the pixel coordinates of the feature points with their 3D coordinates in the camera coordinate system based on a pinhole camera projection model, and establishing a global linear regression model containing unknown position parameter vectors through a linear parameterization method; a second module for performing step S2: designing an adaptive update law for the estimated value of the unknown position parameter vector based on the global linear regression model, used to estimate the unknown depth information parameters of the feature points in real time, and reconstructing the visual Jacobian matrix online based on the estimated depth information parameters; and a third module for performing step S3: constructing a visual state observer, utilizing the reconstructed visual Jacobian... The first module estimates the image feature vector constructed from the pixel coordinates of feature points using a matrix, obtaining smooth image feature estimates. The second module executes step S4: based on the reconstructed visual Jacobian matrix, it defines a normalized Jacobian transpose mapping matrix; based on the normalized Jacobian transpose mapping matrix, it defines a joint space-aided tracking error based on the smooth image feature estimates. The third module executes step S5: for lumped disturbances in robot dynamics, it designs an integral sliding mode robust feedback term based on the joint space-aided tracking error in the robot joint driving torque expression, and combines it with feedforward compensation to generate the final joint driving torque, driving the robot to perform servo tracking tasks. All modules in the above-mentioned robot image visual servo robust control system based on the adaptive update law can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0130] According to a third aspect of the present invention, a processor is provided, the processor being configured to run a program, wherein the program, when running, executes the steps of the robot image visual servo robust control method based on adaptive update law as described in any one of the preceding embodiments.

[0131] Furthermore, a 7-DOF collaborative robot is used as a platform for simulation verification. This application builds a 7-DOF robot hand-eye system model in MATLAB / Simulink, and designs a controller to ensure that the target feature point pixels reach the desired pixel coordinate values. To highlight the superiority of the proposed method, friction is introduced into the robot dynamics model. and external disturbances . Figure 4The diagram illustrates the convergence effect of adaptive depth estimation for four feature points. The results show that even when the initial depth information parameters are completely unknown and there is a preset bias, the adaptive update law designed in this invention can still guide the estimation of the depth information parameters. Perfectly track real depth information parameters in a short time .

[0132] Figure 5 and Figure 6 The convergence curve of the global observation error and the tracking trajectory of the expected pixel coordinates of the feature points are shown respectively. Simulation results demonstrate that this invention effectively smooths image noise through a visual state observer, enabling the global observation error to converge rapidly to zero. Even under extreme conditions of simultaneous joint static friction and external time-varying disturbances, the proposed driving torque controller still ensures that the trajectory of the end-effector's pixel coordinates converges smoothly and without jitter to the desired position. Compared to traditional kinematic PD control architectures, this invention completely eliminates steady-state errors caused by lumped disturbances through integral sliding mode robust feedback terms. Experimental results fully demonstrate that this invention, through integrated "observation-estimation-control" collaborative design, thoroughly solves the dual challenges of kinematic singularities and dynamic uncertainties in visual servoing systems without requiring prior geometric models.

[0133] In summary, this invention addresses the problem of traditional image-based visual servoing methods struggling to converge accurately in complex, unstructured disturbance scenarios. It proposes a novel technical solution that integrates adaptive depth estimation and integral sliding mode robust control. This solution achieves asymptotically stable control under conditions of strong uncertainty in the system dynamics model, exhibiting significant advantages such as high accuracy, strong robustness, and smooth control commands. It holds significant technical application value for improving the performance of industrial robots in complex tasks such as precision assembly and dynamic grasping.

[0134] The specific embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. A robot image visual servo robust control method based on adaptive update law, characterized in that, include: Step S1: Extract image feature points of the target object. Based on the pinhole camera projection model, associate the pixel coordinates of the feature points with their 3D coordinates in the camera coordinate system, and establish a global linear regression model containing unknown position parameter vectors through the linear parameterization method. Step S2: Based on the global linear regression model, design an adaptive update law for the unknown location parameter vector estimate, which is used to estimate the unknown depth information parameters of feature points in real time, and reconstruct the visual Jacobian matrix online based on the depth information parameter estimate. Step S3: Construct a visual state observer, and use the reconstructed visual Jacobian matrix to estimate the image feature vector constructed from the pixel coordinates of the feature points to obtain smooth image feature estimates. Step S4: Based on the reconstructed visual Jacobian matrix, define the normalized Jacobian transpose mapping matrix; based on the normalized Jacobian transpose mapping matrix, define the joint space-assisted tracking error based on the smoothed image feature estimation. Step S5: For the lumped disturbance in robot dynamics, design an integral sliding mode robust feedback term based on the integral sliding mode in the expression of robot joint driving torque according to the joint space auxiliary tracking error, and combine it with feedforward compensation to generate the final joint driving torque to drive the robot to perform servo tracking task.

2. The robot image visual servo robust control method based on adaptive update law according to claim 1, characterized in that, Step S1 specifically involves: For one A robot vision servo system with one degree of freedom, defining the image feature points of a target object extracted by a camera on the robot vision servo system The pixel value in the horizontal direction of the pixel plane is , and the pixel value in the vertical direction is ; according to the pinhole camera projection model, the pixel coordinates of the feature point are associated with its 3D coordinates in the camera coordinate system: ； in, For feature points homogeneous pixel coordinates; For feature points Relative to the camera's depth information parameters, The third row represents the camera intrinsic parameter matrix. and These are the rotation matrix and translation vector from the robot's base coordinate system to the camera coordinate system, respectively. For feature points Constant coordinates in the base coordinate system; This represents the intrinsic parameter matrix of the camera; Indicates transpose; Define auxiliary matrix The pinhole camera projection model is transformed into a linear parameterized form: ； in, For the output vector, This is the regression matrix; Will Augmenting the feature points yields a global linear regression model: ； in, The augmented output vector, The augmented regression matrix, It is an unknown position parameter vector, that is, an augmented constant coordinate vector.

3. The robust control method for robot image vision servoing based on adaptive update law according to claim 1, characterized in that, Step S2 specifically involves: Based on the global linear regression model, an auxiliary matrix with a forgetting factor is introduced. and auxiliary vector The first derivative expressions for the two are as follows: ； in, Forgetting factor; This is the augmented output vector in the global linear regression model. This is the augmented regression matrix in the global linear regression model; Based on the auxiliary matrix with forgetting factor and auxiliary vector Construct error variables related to estimation error : ； in, unknown position parameter vector Estimated value The estimation error of the unknown position parameter vector; With error variable To drive this process, the adaptive update law for the estimated unknown position parameter vector is designed as follows: ； in, It is a symmetric positive definite adaptive gain matrix; Integrating the adaptive update law allows for online updating of the estimated value of the unknown position parameter vector. And then based on Obtain estimated values ​​of depth information parameters. Then, the visual Jacobian matrix is ​​reconstructed based on the estimated values ​​of the depth information parameters. ;in, The third row represents the camera intrinsic parameter matrix. and These are the rotation matrix and translation vector from the robot's base coordinate system to the camera coordinate system, respectively. express The estimated value, For feature points Constant coordinates in the base coordinate system.

4. The robust control method for robot image vision servoing based on adaptive update law according to claim 1, characterized in that, Step S3 specifically involves: Define the global observation error as Based on the global observation error and the reconstructed visual Jacobian matrix, the visual state observer is constructed as follows: ； in, express The first derivative, for The estimated value, For image feature vectors, Representing feature points pixel coordinates transpose, , Indicates the number of feature points; For the reconstructed visual Jacobian matrix; For the robot's joint angular velocity, Indicates the position of the robot's joint angles; This is the gain matrix of the positive definite observer.

5. The robust control method for robot image vision servoing based on adaptive update law according to claim 1, characterized in that, Specifically, S4 is: Define the normalized Jacobian transpose mapping matrix based on the reconstructed visual Jacobian matrix. : ； in, Represents the reconstructed visual Jacobian matrix transpose, Normalization factor; Based on the normalized Jacobian transpose mapping matrix, the joint space auxiliary tracking error is defined. : ； in, For the robot's joint angular velocity; To control the gain; For the tracking error in the defined image space, For smooth image feature estimates, These are the desired pixel coordinates.

6. The robust control method for robot image vision servoing based on adaptive update law according to claim 1, characterized in that, Step S5 specifically involves: Robot joint drive torque for: ； in, For gravity compensation, For Coriolis force and centrifugal force terms, Represents the matrix of Coriolis force and centrifugal force. For the robot's joint angular velocity, Indicates the position of the robot's joint angles. For integral sliding mode robust feedback; based on joint space auxiliary tracking error The integral sliding mode robust feedback term in the robot joint driving torque expression is designed as follows: ； in, For design gain, robust term gain , This represents the lumped disturbance derivative term, which is composed of the rate of change of external disturbance and the rate of change of frictional force when moving along the desired trajectory. Represents the infinite norm; For symbolic functions, Joint space auxiliary tracking error exist The initial value at time.

7. A robust control system for robot image vision servoing based on an adaptive update law, characterized in that, The module includes the robot image vision servo robust control method based on adaptive update law as described in any one of claims 1-6.

8. A processor, characterized in that, The processor is used to run a program, characterized in that, when the program runs, it executes the steps of the robot image visual servo robust control method based on adaptive update law as described in any one of claims 1-6.

Images