CN112720480B - Robot track correction method and system based on grading errors - Google Patents
Robot track correction method and system based on grading errors Download PDFInfo
- Publication number
- CN112720480B CN112720480B CN202011532552.9A CN202011532552A CN112720480B CN 112720480 B CN112720480 B CN 112720480B CN 202011532552 A CN202011532552 A CN 202011532552A CN 112720480 B CN112720480 B CN 112720480B
- Authority
- CN
- China
- Prior art keywords
- robot
- error
- positioning
- parameter
- errors
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1656—Programme controls characterised by programming, planning systems for manipulators
- B25J9/1664—Programme controls characterised by programming, planning systems for manipulators characterised by motion, path, trajectory planning
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1602—Programme controls characterised by the control system, structure, architecture
- B25J9/1607—Calculation of inertia, jacobian matrixes and inverses
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1602—Programme controls characterised by the control system, structure, architecture
- B25J9/161—Hardware, e.g. neural networks, fuzzy logic, interfaces, processor
Landscapes
- Engineering & Computer Science (AREA)
- Robotics (AREA)
- Mechanical Engineering (AREA)
- Automation & Control Theory (AREA)
- Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Software Systems (AREA)
- Numerical Control (AREA)
- Manipulator (AREA)
Abstract
The invention discloses a robot track correction method and system based on grading errors, belonging to the field of robot control and comprising the following steps: determining an optimal test point by taking the minimum condition number as a target, and measuring a first positioning error of the robot under the optimal test point; correcting a DH parameter in the robot controller according to the DH parameter error in the first positioning error; dividing joint space of front three axes of the robot into a plurality of grids, and measuring second positioning errors of the robot under each grid; identifying joint angle compensation values corresponding to the front three axes from the second positioning error, and training a preset neural network by using the joint angle compensation values and joint angles corresponding to each grid; and correcting the programming track by using the trained neural network until the actual track obtained by correction is consistent with the target track. Geometric errors and non-geometric errors of the robot are compensated and corrected in a grading mode, the track accuracy of the robot is improved, and the robot has the advantages of being wide in applicability, good in stability, high in accuracy and the like.
Description
Technical Field
The invention belongs to the field of robot control, and particularly relates to a robot track correction method and system based on grading errors.
Background
Industrial robots are important components of the robot industry and play a vital role in the fields of electrical and electronics, medical aerospace, home appliances, automobiles, and the like. As industrial robots enter the field of precision equipment such as medical aerospace, electrical electronics and the like, the positioning precision and the trajectory precision of the robots are more and more important. The positioning accuracy is the basis of the track accuracy, and the positioning errors of the robot are mainly divided into geometric parameter errors and non-geometric parameter errors according to sources. The geometric parameter error is an error caused by the mismatching of the robot kinematic model parameters and the parameter values of the robot real model. The non-geometric parameter errors are errors caused by factors except kinematic geometric parameter errors, and errors which cannot be directly used for modeling analysis by kinematic geometric parameters mainly comprise errors caused by load, dead weight, temperature and humidity, transmission errors, mechanical wear errors and the like.
The accurate kinematics model is the precondition of high-precision positioning control of the robot, and the method for compensating the parameter error of the kinematics model of the robot through kinematics calibration is a common method for improving the accuracy of the kinematics model. However, the ordinary calibration can only compensate the geometric parameter error, and cannot compensate the non-geometric parameter error. To compensate for non-geometric parameter errors, a parametric-free kinematic calibration may be employed. The parameter-free kinematic calibration is to establish a relation between the pose of the robot and an error by collecting a large amount of data, and estimate and compensate the error by combining a mathematical method. The accuracy and generalization capability of the model established by the method are very dependent on data, and the acquisition of a large amount of data is very time-consuming.
Disclosure of Invention
Aiming at the defects and improvement requirements of the prior art, the invention provides a robot track correction method and system based on grading errors, and aims to compensate geometric parameter errors and non-geometric parameter errors of a robot, improve the positioning accuracy and track accuracy of the robot and avoid long time consumption caused by acquiring a large amount of data.
To achieve the above object, according to an aspect of the present invention, there is provided a grading error-based robot trajectory correction method, including: s1, determining an optimal test point by taking the condition number of the robot positioning error relative to the differential Jacobian matrix of the DH parameters as a target, and measuring a first positioning error of the robot under the optimal test point; s2, recognizing a DH parameter error from the first positioning error, and correcting a DH parameter in a robot controller according to the DH parameter error to compensate the geometric parameter error of the robot; s3, dividing joint space of the front three axes of the robot into a plurality of grids, and measuring second positioning errors of the robot under each grid, wherein the second positioning errors are errors caused by non-geometric factors; s4, identifying joint angle compensation values corresponding to the front three axes from second positioning errors corresponding to each grid, and training a preset neural network by using the joint angles corresponding to each grid and the joint angle compensation values; and S5, correcting the programmed trajectory in the robot controller by using the trained neural network until the actual trajectory of the robot is consistent with the target trajectory under the control of the corrected programmed trajectory.
Further, the identifying a DH parameter error from the first positioning error in S2 includes: carrying out QR decomposition on the differential Jacobian matrix of the robot positioning error relative to the DH parameters to remove redundant parameters in the DH parameter error through the R matrix after the QR decomposition, wherein the decomposition is as follows:
identifying a DH parameter error from the first positioning error by using a least square method, wherein the DH parameter error is as follows:
Δq＝(J T J) -1 J T ΔP
wherein Δ q is the DH parameter error, Δ P is the first positioning error, J is a differential Jacobian matrix of the robot positioning error with respect to the DH parameter, J is T Is a transposed matrix of J, Q is an orthogonal matrix, R is an upper triangular matrix, and O is a 0 matrix.
Further, in the step S2, a DH parameter is identified from the first positioning errorThe error includes: identifying a DH parameter error from the first positioning error by using a particle swarm optimization, wherein the fitness function of the particle swarm optimization is the positioning error of the robot under all the test points; and along with the superposition of iteration times, the inertia weight of the particle swarm optimization is linearly reduced, and the acceleration constant c is increased 1 Linear decrease, acceleration constant c 2 Increasing linearly.
Further, the inertial weight decreases linearly from 0.9 to 0.5, the acceleration constant c 1 Linearly decreasing from 2.5 to 0.5, the acceleration constant c 2 Increasing linearly from 0.5 to 2.5.
Further, the joint space of the front three axes of the robot is divided into a plurality of meshes with a rotation step of 20 ° in S3.
Further, the step of identifying the joint angle compensation values corresponding to the front three axes from the second positioning errors corresponding to each grid in S4 includes: constructing an over-determined equation set based on second positioning errors of 8 vertexes and central points of each grid to obtain joint angle compensation values corresponding to the front three axes, wherein the over-determined equation set is as follows:
wherein, Delta theta is a vector formed by the joint angle compensation values corresponding to the front three axes, J R For a differential jacobian matrix of positioning errors with respect to the anterior three-axis corresponding joint angles,
is J R Δ P is a vector formed by the second positioning errors of the 8 vertices and the center point.
Further, the condition number is:
cond(J)＝||J||·||J + ||
wherein cond (J) is the condition number, J is a differential Jacobian matrix of the robot positioning error with respect to the DH parameters, J + Is a generalized inverse matrix of J.
Further, the correcting the programmed trajectory in the robot controller by using the trained neural network in S5 includes: and correcting each interpolation point of the programming track by using the trained neural network to obtain a pseudo target position corresponding to each interpolation point.
Further, the preset neural network comprises two layers, wherein the first layer adopts a tansig hyperbolic tangent transfer function, the second layer adopts a purelin linear transfer function, the training function adopts a Levenberg _ Marquardt function, and the network learning function adopts a Learngdm function with additional momentum.
According to an aspect of the present invention, there is provided a grading error-based robot trajectory correction system, including: the system comprises a first measurement module, a second measurement module and a third measurement module, wherein the first measurement module is used for determining an optimal test point by taking the minimum condition number of a differential Jacobian matrix of a DH parameter related to a robot positioning error as a target, and measuring a first positioning error of the robot under the optimal test point; the first correction module is used for identifying a DH parameter error from the first positioning error and correcting a DH parameter in a robot controller according to the DH parameter error so as to compensate the geometric parameter error of the robot; the second measurement module is used for dividing the joint space of the front three axes of the robot into a plurality of grids and measuring a second positioning error of the robot under each grid, wherein the second positioning error is an error caused by non-geometric factors; the training module is used for identifying joint angle compensation values corresponding to the front three axes from second positioning errors corresponding to the grids and training a preset neural network by using the joint angles corresponding to the grids and the joint angle compensation values; and the second correction module is used for correcting the programmed track in the robot controller by using the trained neural network until the actual track of the robot is consistent with the target track under the control of the corrected programmed track.
Generally, by the above technical solution conceived by the present invention, the following beneficial effects can be obtained: compensating and correcting geometric errors and non-geometric errors of the robot in a grading manner, and specifically, providing a joint grid-based method on the basis of a first-stage error modelThe second-stage model error model of the division and neural network can compensate and correct geometric errors and non-geometric errors of the robot, so that the track precision of the robot is improved, and the robot has the advantages of wide applicability, good stability, high precision and the like; the data required by the established second-stage error model is greatly less than that of the existing parameter-free kinematic calibration, so that the time consumption caused by acquiring a large amount of data is avoided, the angle compensation values of all joints of the front three axes can be predicted, and the positioning accuracy of the robot is improved to improve the track accuracy of the robot; the optimal test point is screened based on the condition number, so that the influence of measurement errors, noise interference and the like during testing is reduced, and the stability of parameter identification is improved; in addition, when parameter identification is carried out by a particle swarm optimization method, the inertia weight is linearly decreased from 0.9 to 0.5 along with the increase of the iteration times; acceleration constant c 1 The acceleration constant c decreases linearly from 2.5 to 0.5 as the number of iterations increases 2 As the number of iterations increases linearly from 0.5 to 2.5, trapping in local minima can be avoided and the final optimal solution is made more accurate.
Drawings
Fig. 1 is a flowchart of a robot trajectory correction method based on a grading error according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a coordinate system of a robot link according to an embodiment of the present invention;
FIG. 3 is a front view of the range of travel of an HSR-JR605 robot in an embodiment of the present invention;
FIG. 4 is a top view of the range of travel of the HSR-JR605 robot in accordance with an embodiment of the present invention;
FIG. 5 is a schematic diagram of a trajectory modification in an embodiment of the present invention;
fig. 6 is a block diagram of a system for correcting a robot trajectory based on a grading error according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
In the present application, the terms "first," "second," and the like (if any) in the description and the drawings are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order.
Fig. 1 is a flowchart of a robot trajectory correction method based on a grading error according to an embodiment of the present invention. Referring to fig. 1, a detailed description will be given of the robot trajectory correction method based on the grading error in the present embodiment with reference to fig. 2 to 5.
In this embodiment, the method for correcting the trajectory of the robot based on the hierarchical error will be described in detail by taking a Huazhong numerical control HSR-JR605 robot as an example. The link coordinate system of the robot is shown in fig. 2, and the link transformation matrix is:
wherein the content of the first and second substances,
a transformation matrix of a connecting rod coordinate system { i } relative to { i-1 }; rot (x, alpha) i-1 ) Representing the coordinate system i-1 around x i-1 Rotation alpha i-1 An angle; trans (x, a) i-1 ) Representing the coordinate system i-1 along x i-1 Axial movement a i-1 ；Rot(z,θ i ) Representing a coordinate system i-1 around z i Rotation theta i An angle; trans (z, d) i ) Representing a coordinate system i-1 along z i Axial movement d i . The number of joints of the HSR-JR605 robot is 6, the travel range is shown in figures 3 and 4, and the specific link parameters are shown in table 1.
TABLE 1
In Table 1, a i-1 Represents the length of link i-1; alpha is alpha i-1 Represents from z i-1 To z i Around x i-1 The angle of rotation; d i Represents from x i-1 To x i Along the measured distance; theta i Represents from x i-1 To x i Around z i The angle of rotation. The robot trajectory correction method based on the grading error includes operations S1-S4.
Operation S1 is to determine an optimal test point with a target of a minimum condition number of a differential jacobian matrix of robot positioning errors with respect to DH parameters, and measure a first positioning error of the robot under the optimal test point.
The differential jacobian matrix of the robot positioning error with respect to the DH parameters is J, and its condition number cond (J) is:
cond(J)＝||J||·||J + ||
wherein, J + Is a generalized inverse matrix of J. Let the test point with the smallest condition number cond (j) be the best test point.
The test point is a group of joint angles used by the test robot, the tail end of the robot can move to a position in a Cartesian space by inputting the group of joint angles, the position error of the tail end of the robot can be measured by external measuring equipment, and the position error is the first positioning error.
And operation S2, recognizing a DH parameter error from the first positioning error, and correcting a DH parameter in the robot controller according to the DH parameter error to compensate for the geometric parameter error of the robot.
The DH parameters include [ alpha, d, theta, a ], and the DH parameter errors include [ delta alpha, delta d, delta theta, delta a ]. In an embodiment of the invention, a DH parameter error is identified from the first positioning error by using a least square method. Because the matrix J is not column full rank, redundancy analysis needs to be performed on the matrix J to remove redundant parameters, specifically, the redundant parameters are removed through QR decomposition, and the decomposed matrix J is:
wherein Q is an orthogonal matrix, R is an upper triangular matrix, and O is a 0 matrix. And (3) parameters corresponding to columns with elements of 0 on the diagonal line of the matrix R obtained through QR decomposition are redundant parameters, and are removed.
Then, based on the matrix J after QR decomposition, a DH parameter error is identified from the first positioning error by using a least square method, wherein the DH parameter error delta q is as follows:
Δq＝(J T J) -1 J T ΔP
where Δ P is the first positioning error.
In another embodiment of the present invention, a DH parameter error is identified from the first positioning error by using a least square method. Taking the HSR-JR605 robot shown in Table 1 as an example, the number of parameters to be identified is 24, so that the particle swarm search space dimension D is 24, a cluster is formed by N particles, and the ith particle is represented as a vector X of dimension D i ＝(x i1 ,x i2 ,…,x iD ) I ═ 1,2, …, N; the velocity of the ith particle is also a D-dimensional vector V i ＝(v i1 ,v i2 ,…,v iD ) I ═ 1,2, …, N; the optimum position searched by the ith particle is an individual extreme value which is marked as p i ＝(p i1 ,p i2 ,…,p iD ) I ═ 1,2, …, N; the optimal positions searched by all the particles are global extremum and are marked as g best ＝(g 1 ,g 2 ,…,g D ) (ii) a Each particle updates its velocity and position according to the following formula:
v ij (t+1)＝v ij (t)+c 1 r 1 (t)[p ij (t)-x ij (t)]+c 2 r 2 (t)[p gj (t)-x ij (t)]
x ij (t+1)＝x ij (t)+v ij (t+1)
wherein, c 1 And c 2 Is an acceleration constant, r 1 And r 2 Is [0,1 ]]A uniform random number within the range. In this embodiment, the particle swarm optimization key parameters are set as follows: adaptation toThe degree function is the positioning error of the robot under all the test points; and along with the superposition of iteration times, the inertia weight of the particle swarm optimization is linearly reduced, and the acceleration constant c is increased 1 Linear decrease, acceleration constant c 2 Increasing linearly. Further, particle population size N is, for example, 150; the maximum number of iterations is, for example, 300; the inertia weight w is linearly decreased from 0.9 to 0.5 along with the increase of the iteration times, so that the inertia weight is larger at the beginning, the global search capability is strong, and the inertia weight is smaller and the local optimization capability is strong at the end; acceleration constant c 1 The acceleration constant c decreases linearly from 2.5 to 0.5 as the number of iterations increases 2 Linearly increasing from 0.5 to 2.5 as the number of iterations increases, thus starting with an acceleration constant c 1 Larger, large influence of individual experience of the particle, and the acceleration constant c at the end of the process 2 The particle population experience is large, so that the final optimal solution is more accurate; and stopping the algorithm when the iteration times reach the maximum iteration times or the global optimal value of the iteration times of the previous and next 50 times is less than 0.01.
And correcting the DH parameters in the robot controller by using the DH parameter errors identified by a least square method or a particle swarm method so as to compensate the geometric parameter errors of the robot.
Operation S3 is to divide the joint space of the front three axes of the robot into a plurality of grids, and measure a second positioning error of the robot under each grid, where the second positioning error is an error caused by non-geometric factors.
Because the influence of the front three axes of the industrial robot on the position of the tail end is large, in order to divide the grid conveniently and not to influence the testing accuracy, in this embodiment, the joint space of the front three axes of the robot is divided into the grid, specifically, the joint space of the front three axes of the robot is divided into a plurality of grids by a rotation step length of 20 degrees, and the positioning error data of the robot under each grid is obtained through retesting.
The joint space of the front three shafts of the robot is divided into a plurality of grids, namely the robot is controlled to enable the front three shafts (the joint 1, the joint 2 and the joint 3) to rotate, other shafts are fixed, and the space passed by the tail end of the robot is one grid. For the HSR-JR605 robot shown in table 1, the joint variable ranges of the front three axes are [ -200 °, +200 ° ], [ -180 °, 0 ° ], [ -60 °, +100 ° ], and the three axes are gridded in rotation steps of 20 °. And retesting to obtain second positioning error data corresponding to each grid.
Operation S4 is performed to identify the joint angle compensation values corresponding to the front three axes from the second positioning errors corresponding to the grids, and train the preset neural network using the joint angles and the joint angle compensation values corresponding to the grids.
The joint angle is an input value, and each grid has a corresponding joint angle. According to the embodiment of the invention, an over-determined equation set is constructed based on the second positioning errors of 8 vertexes and central points of each grid to obtain joint angle compensation values corresponding to the front three axes, and the over-determined equation set is as follows:
wherein, Δ θ is a vector formed by the joint angle compensation values corresponding to the first three axes, and Δ θ ═ Δ θ 1 ，Δθ 2 ，Δθ 3 ]；J R To be a differential jacobian matrix of the positioning error with respect to the anterior three axis corresponding joint angles,
is J R Δ P is a vector formed by the second positioning errors of the 8 vertices and the center point. Thus, the joint angle compensation value Δ θ corresponding to each grid is obtained.
And substituting the joint angle compensation value delta theta and the joint angle corresponding to each grid into a preset neural network for training. The preset neural network is a BP neural network, the BP neural network consists of two processes of information forward propagation and error backward propagation, the input is the angle of the space center of each joint, the BP neural network comprises two layers, a first Layer hidden Layer adopts a tansig hyperbolic tangent transfer function, a second Layer Output Layer adopts a purelin linear transfer function, the Output is a joint angle compensation value corresponding to each joint point, a Levenberg _ Marquardt function is adopted as a training function, and a network learning function adopts a Learndm function with additional momentum, so that the trapping of the partial minimum value in the backward propagation process can be effectively avoided, and the training time can be possibly reduced.
And operation S5, correcting the programmed trajectory in the robot controller by using the trained neural network until the actual trajectory of the robot under the control of the corrected programmed trajectory is consistent with the target trajectory.
Specifically, each interpolation point of the programmed trajectory in the robot controller is corrected by using the trained neural network, so as to obtain a pseudo target position corresponding to each interpolation point, and a pseudo target trajectory formed by each pseudo target position is shown in fig. 5. The robot controller controls the robot to move according to the pseudo target track to form an actual track of the robot, so that the actual track of the robot is consistent with the target track, and the track precision of the robot is improved.
Fig. 6 is a block diagram of a robot trajectory correction system based on a grading error according to an embodiment of the present invention. Referring to fig. 6, the grading error based robot trajectory correction system includes a first measurement module 610, a first correction module 620, a second measurement module 630, a training module 640, and a second correction module 650.
The first measurement module 610 performs, for example, operation S1 for targeting a condition number minimization of the robot positioning error with respect to a differential jacobian matrix of DH parameters, determining an optimal test point, and measuring a first positioning error of the robot under the optimal test point.
The first modification module 620 performs, for example, operation S2, to identify a DH parameter error from the first positioning error, and modify a DH parameter in the robot controller according to the DH parameter error to compensate for the geometric parameter error of the robot.
The second measurement module 630 performs, for example, operation S3, to divide the joint space of the front three axes of the robot into a plurality of grids, and measure a second positioning error of the robot under each grid, where the second positioning error is an error caused by a non-geometric factor.
The training module 640 performs operation S4, for example, to identify the joint angle compensation values corresponding to the front three axes from the second positioning errors corresponding to each grid, and train the preset neural network using the joint angles and the joint angle compensation values corresponding to each grid.
The second modification module 650 performs, for example, operation S5, to modify the programmed trajectory in the robot controller by using the trained neural network until the actual trajectory of the robot under the control of the modified programmed trajectory is consistent with the target trajectory.
The grading error based robot trajectory modification system 600 is used to perform the grading error based robot trajectory modification method in the embodiments shown in fig. 1-5. For details that are not described in the present embodiment, please refer to the method for correcting a robot track based on a grading error in the embodiments shown in fig. 1 to fig. 5, which is not described herein again.
It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (9)
1. A robot track correction method based on grading errors is characterized by comprising the following steps:
s1, determining an optimal test point by taking the condition number of the robot positioning error relative to the differential Jacobian matrix of the DH parameters as a target, and measuring a first positioning error of the robot under the optimal test point;
s2, recognizing a DH parameter error from the first positioning error, and correcting a DH parameter in a robot controller according to the DH parameter error to compensate the geometric parameter error of the robot;
s3, dividing joint space of the front three axes of the robot into a plurality of grids, and measuring second positioning errors of the robot under each grid, wherein the second positioning errors are errors caused by non-geometric factors;
s4, identifying joint angle compensation values corresponding to the front three axes from second positioning errors corresponding to the grids, and training a preset neural network by using the joint angles corresponding to the grids and the joint angle compensation values to obtain a second-stage error model based on joint grid division and the neural network so as to predict all joint angle compensation values of the front three axes;
s5, correcting the programming track in the robot controller by using the trained neural network until the actual track of the robot is consistent with the target track under the control of the corrected programming track;
the correcting the programmed trajectory in the robot controller by using the trained neural network in S5 includes: and correcting each interpolation point of the programming track by using the trained neural network to obtain a pseudo target position corresponding to each interpolation point.
2. The grading error-based robot trajectory modification method of claim 1, wherein the identifying a DH parameter error from the first positioning error in S2 comprises:
carrying out QR decomposition on the differential Jacobian matrix of the robot positioning error relative to the DH parameters to remove redundant parameters in the DH parameter error through the R matrix after the QR decomposition, wherein the decomposition is as follows:
identifying a DH parameter error from the first positioning error by using a least square method, wherein the DH parameter error is as follows:
Δq＝(J T J) -1 J T ΔP
wherein Δ q is the DH parameter error, Δ P is the first positioning error, J is a differential Jacobian matrix of the robot positioning error with respect to the DH parameter, J is T Is a transposed matrix of J, Q is an orthogonal matrix, R is an upper triangular matrix, and 0 is a 0 matrix.
3. The grading error-based robot trajectory modification method of claim 1, wherein the identifying a DH parameter error from the first positioning error in S2 comprises:
identifying a DH parameter error from the first positioning error by using a particle swarm optimization, wherein the fitness function of the particle swarm optimization is the positioning error of the robot under all the test points; and along with the superposition of iteration times, the inertia weight of the particle swarm optimization is linearly reduced, and the acceleration constant c is increased 1 Linear decrease, acceleration constant c 2 Increasing linearly.
4. The grading error based robot trajectory correction method of claim 3, wherein the inertial weight is linearly reduced from 0.9 to 0.5, and the acceleration constant c is 1 Linearly decreasing from 2.5 to 0.5, the acceleration constant c 2 Increasing linearly from 0.5 to 2.5.
5. The grading error-based robot trajectory modification method of claim 1, wherein the joint space of the front three axes of the robot is divided into a plurality of meshes with a rotation step of 20 ° in S3.
6. The grading error-based robot trajectory modification method of claim 1, wherein the identifying joint angle compensation values corresponding to the front three axes from the second positioning errors corresponding to each grid in S4 comprises:
constructing an over-determined equation set based on second positioning errors of 8 vertexes and central points of each grid to obtain joint angle compensation values corresponding to the front three axes, wherein the over-determined equation set is as follows:
wherein, Delta theta is a vector formed by the joint angle compensation values corresponding to the front three axes, J R For a differential jacobian matrix of positioning errors with respect to the anterior three-axis corresponding joint angles,
is J R Δ P is a vector formed by the second positioning errors of the 8 vertices and the center point.
7. The grading error based robot trajectory modification method of claim 1, wherein the condition number is:
cond(J)＝||J||·||J + ||
wherein cond (J) is the condition number, J is a differential Jacobian matrix of the robot positioning error with respect to DH parameters, J + Is a generalized inverse matrix of J.
8. The grading error-based robot trajectory modification method of any one of claims 1-7, wherein the pre-set neural network comprises two layers, a first layer using a tansig hyperbolic tangent transfer function, a second layer using a purelin linear transfer function, a training function using a Levenberg Marquardt function, and a network learning function using a Learngdm function with additional momentum.
9. A grading error based robot trajectory modification system, comprising:
the first measurement module is used for determining an optimal test point by taking the minimum condition number of the robot positioning error relative to a differential Jacobian matrix of a DH parameter as a target, and measuring a first positioning error of the robot under the optimal test point;
the first correction module is used for identifying a DH parameter error from the first positioning error and correcting a DH parameter in a robot controller according to the DH parameter error so as to compensate the geometric parameter error of the robot;
the second measurement module is used for dividing the joint space of the front three axes of the robot into a plurality of grids and measuring a second positioning error of the robot under each grid, wherein the second positioning error is an error caused by non-geometric factors;
the training module is used for identifying joint angle compensation values corresponding to the front three axes from second positioning errors corresponding to the grids, training a preset neural network by using the joint angles corresponding to the grids and the joint angle compensation values to obtain a second-stage error model based on joint grid division and the neural network so as to predict all joint angle compensation values of the front three axes;
the second correction module is used for correcting the programming track in the robot controller by using the trained neural network until the actual track of the robot is consistent with the target track under the control of the corrected programming track;
the correcting the programming trajectory in the robot controller by using the trained neural network comprises: and correcting each interpolation point of the programming track by using the trained neural network to obtain a pseudo target position corresponding to each interpolation point.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011532552.9A CN112720480B (en) | 2020-12-22 | 2020-12-22 | Robot track correction method and system based on grading errors |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011532552.9A CN112720480B (en) | 2020-12-22 | 2020-12-22 | Robot track correction method and system based on grading errors |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112720480A CN112720480A (en) | 2021-04-30 |
| CN112720480B true CN112720480B (en) | 2022-08-16 |
Family
ID=75604169
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011532552.9A Active CN112720480B (en) | 2020-12-22 | 2020-12-22 | Robot track correction method and system based on grading errors |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112720480B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114012733B (en) * | 2021-11-29 | 2023-05-23 | 江苏科技大学 | Mechanical arm control method for scribing of PC component die |
| CN114454179B (en) * | 2022-03-18 | 2022-10-14 | 深圳威洛博机器人有限公司 | Automatic correction system and method for joints of intelligent robot hand |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102230783A (en) * | 2011-05-04 | 2011-11-02 | 南京航空航天大学 | Three-dimensional grid precision compensation method for industrial robot |
| CN102607552A (en) * | 2012-01-11 | 2012-07-25 | 南京航空航天大学 | Industrial robot space grid precision compensation method based on neural network |
| CN105095962A (en) * | 2015-07-27 | 2015-11-25 | 中国汽车工程研究院股份有限公司 | Method for predicting dynamic mechanical property of material based on BP artificial neural network |
| CN110385720A (en) * | 2019-07-26 | 2019-10-29 | 南京航空航天大学 | A kind of robot localization error compensating method based on deep neural network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102013101095B4 (en) * | 2013-02-04 | 2016-12-15 | Johannes Gottlieb | Method and arrangement for correcting poses errors of a kinematics and a corresponding computer program and a corresponding computer-readable storage medium |
-
2020
- 2020-12-22 CN CN202011532552.9A patent/CN112720480B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102230783A (en) * | 2011-05-04 | 2011-11-02 | 南京航空航天大学 | Three-dimensional grid precision compensation method for industrial robot |
| CN102607552A (en) * | 2012-01-11 | 2012-07-25 | 南京航空航天大学 | Industrial robot space grid precision compensation method based on neural network |
| CN105095962A (en) * | 2015-07-27 | 2015-11-25 | 中国汽车工程研究院股份有限公司 | Method for predicting dynamic mechanical property of material based on BP artificial neural network |
| CN110385720A (en) * | 2019-07-26 | 2019-10-29 | 南京航空航天大学 | A kind of robot localization error compensating method based on deep neural network |
Non-Patent Citations (3)
| Title |
|---|
| 基于ELM的工业机器人精度补偿方法研究;冯禹铭;《中国优秀博硕士学位论文全文数据库(硕士) 信息科技辑》;20200215(第02期);第5-6、36-43页 * |
| 基于嵌入式系统的装配机械手精确定位控制算法研究;赵忠伟;《中国优秀博硕士学位论文全文数据库(硕士) 信息科技辑》;20100815(第08期);第35-41页 * |
| 工业机器人运动学参数标定及误差补偿研究;王坤;《中国优秀博硕士学位论文全文数据库（硕士） 信息科技辑》;20190615(第06期);第4-60页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112720480A (en) | 2021-04-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110815206B (en) | Kinematics calibration method for Stewart parallel robot | |
| CN112720480B (en) | Robot track correction method and system based on grading errors | |
| CN110193829A (en) | A kind of robot precision's control method of coupled motions and stiffness parameters identification | |
| CN108908335B (en) | Robot calibration method based on improved differential evolution algorithm | |
| CN110900610B (en) | Industrial robot calibration method based on LM algorithm and particle filter algorithm optimization | |
| CN112558547A (en) | Quick optimization method for geometric error compensation data of translational shaft of five-axis numerical control machine tool | |
| CN105773622A (en) | Industrial robot absolute accuracy calibrating method based on IEKF | |
| CN112549019B (en) | Industrial robot track accuracy analysis method based on continuous dynamic time warping | |
| CN111216138A (en) | Robot calibration method, robot calibration system and readable storage medium | |
| CN114147726B (en) | Robot calibration method combining geometric error with non-geometric error | |
| CN107102617A (en) | A kind of high-precision spatial elliptic curve Real-time Interpolation | |
| CN114714348B (en) | Industrial robot absolute positioning precision improvement method | |
| CN111195915B (en) | Mechanical arm absolute positioning precision calibration method and system based on genetic algorithm | |
| CN109202539B (en) | Online detection method for composite material weak-rigidity special-shaped structure | |
| CN113211436B (en) | Six-degree-of-freedom series robot error calibration method based on genetic algorithm | |
| Luo et al. | An improved minimal error model for the robotic kinematic calibration based on the POE formula | |
| CN115237056B (en) | Multi-tool rapid deviation rectifying method for industrial robot | |
| CN112643658A (en) | Calibration method for adaptive error modeling of series robot based on SIR dimension reduction DH model | |
| CN113878586B (en) | Robot kinematics calibration device, method and system | |
| Chen et al. | Configuration optimization for manipulator kinematic calibration based on comprehensive quality index | |
| CN113094642B (en) | Newly-added sampling method for estimating positioning error of mechanical arm in region | |
| CN114186477A (en) | Elman neural network-based orbit prediction algorithm | |
| CN109631813B (en) | Calibration method of large-size articulated arm type coordinate measuring machine | |
| Liang et al. | Calibration of UR5 manipulator based on kinematic models | |
| CN117901122B (en) | Robot positioning method and system based on Logistic-Tent chaotic mapping Levenberg Marquardt |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |