Six-axis industrial robot positioning error compensation method

Abstract

The application discloses a kind of six-axis industrial robot positioning error compensation methods, comprising: robot is kinematically modeled and joint space is divided, and the first three-axis joint space and the last three-axis joint space are respectively grid discrete, grid point is as sampling point;Laser tracker is switched to robot base coordinate system;The sampling point of the first three-axis joint space is tracked measurement, utilizes special tool to simultaneously collect three point position information at the same sampling point to calculate the position and attitude error corresponding to sampling point;Laser tracker is switched to robot wrist coordinate system;The sampling point of the last three-axis joint space is tracked measurement, calculates the positioning error corresponding to sampling point;According to the pose error data of sampling point, the pose error of target point is calculated;According to the pose error obtained, positioning error compensation is realized.The application can effectively solve the dimension disaster problem existing in the joint space grid compensation of six-axis industrial robot, and improve the positioning accuracy of robot.

Description

Technical Field

[0001] This invention relates to the field of industrial robot positioning error compensation technology, and in particular to a method for compensating positioning errors of a six-axis industrial robot. Background Technology

[0002] Six-axis industrial robots generally possess high repeatability but poor absolute positioning accuracy, which severely limits their application in fields requiring high precision. Factors affecting accuracy are multifaceted and can be categorized into two types based on their source of error: firstly, geometric parameter errors, referring to errors caused by discrepancies between the actual and theoretical values ​​of robot geometric parameters such as link length, offset, and torsion angle; and secondly, non-geometric parameter errors, referring to errors caused by factors other than geometric parameter errors, such as flexibility deformation errors caused by load and self-weight, errors caused by temperature and humidity changes, and errors caused by friction. Performance measurements of six-axis industrial robots are generally conducted using the national standard GB / T12642-2013, "Performance Specifications and Test Methods for Industrial Robots." This performance specification defines 14 parameters for determining the performance level of industrial robots and their corresponding test methods, such as pose accuracy and repeatability, distance accuracy and repeatability, and trajectory accuracy and repeatability.

[0003] For geometric parameter errors, kinematic parameter identification methods can be used to obtain the true geometric parameters and reduce their impact. However, non-geometric parameter errors are difficult to model, and compensation methods are generally used. This involves predicting the error of any pose based on the errors of certain pre-sampled poses without considering the source of the error, and then completing the compensation. Based on this, the spatial mesh method was developed. For example, Chinese patent document CN114161425A discloses an error compensation method for industrial robots, which combines the axis measurement method and the spatial mesh method to perform error compensation for industrial robots.

[0004] Chinese patent document CN113910239A discloses an absolute positioning error compensation device for an industrial robot. It proposes a two-step spatial grid partitioning strategy, establishes and updates a spatial error library in real time, calculates the absolute positioning error values ​​of key nodes on the robot's motion trajectory using an interpolation algorithm, corrects the coordinate values ​​in the control program, and completes online compensation of the robot's absolute positioning error without affecting the running cycle.

[0005] Based on the workspace type, spatial meshes can be divided into two types: Cartesian spatial meshes and joint spatial meshes. Since the inverse kinematics of industrial robots is not unique, Cartesian spatial mesh compensation can lead to poor accuracy in applications with significant changes in end-effector posture. Theoretically, joint spatial mesh compensation is more reasonable; however, for six-axis industrial robots, the end-effector pose is a function of the six joint angles, and directly creating a spatial mesh can lead to the curse of dimensionality, which limits the development and application of this compensation method. Summary of the Invention

[0006] This invention provides a positioning error compensation method for a six-axis industrial robot, which simultaneously compensates for geometric and non-geometric errors of the six-axis industrial robot to achieve higher compensation accuracy, and solves the problem of dimensional curse caused by full-joint spatial mesh compensation.

[0007] A method for compensating positioning errors in a six-axis industrial robot includes the following steps:

[0008] (1) Perform kinematic modeling and joint space segmentation on the six-axis industrial robot, and discretize the joint spaces of the first three axes and the last three axes into grids, with grid points as sampling points;

[0009] Among them, the joint space of the first three axes is from the robot base to the center of the robot wrist, and the joint space of the last three axes is from the center of the robot wrist to the robot end effector.

[0010] (2) Transfer the laser tracker to the robot's base coordinate system;

[0011] (3) The sampling points of the three-axis joint space are tracked and measured by a laser tracker. The three points are collected at the same sampling point using a special tooling to calculate the position and attitude error corresponding to the sampling point.

[0012] (4) Transfer the laser tracker to the robot's wrist coordinate system;

[0013] (5) Track and measure the sampling points in the rear three-axis joint space using a laser tracker, and calculate the positioning error corresponding to the sampling points;

[0014] (6) Calculate the pose error of the target point based on the pose error of the sampling point obtained in step (3) and the positioning error of the sampling point obtained in step (5);

[0015] (7) Compensate for the positioning error of the target point based on the pose error obtained in step (6).

[0016] This invention calculates and compensates for the pose error of the target point based on the pose data of two subspaces. First, the position and error of the grid points in the first three-axis joint space are sampled, and the positioning error of the grid points in the second three-axis joint space is sampled. Then, the inverse distance weighting method is used to calculate and compensate for the error of the target point in the two subspaces respectively.

[0017] This invention effectively solves the problem of the curse of dimensionality in joint space mesh compensation of six-axis industrial robots, making the compensation method practically feasible, and the positioning accuracy of the robot can be significantly improved after compensation.

[0018] In step (1), the specific method for kinematic modeling and joint space segmentation of the six-axis industrial robot is as follows:

[0019] (1-1) The DH method is used to model the kinematics of the robot. The transformation relationship between adjacent links is as follows:

[0020]

[0021] Where T represents the transformation relationship between the link coordinate systems, x and z represent the x-axis and z-axis of the link coordinate system, i represents the link number, Rot represents rotation about the corresponding axis, Trans represents translation along the corresponding axis, θ is the joint angle, d is the link offset, a is the link length, and α is the link torsion angle. Expanded into the general form:

[0022]

[0023] Where, cθ i =cosθ i sθ i =sinθ i ca i =cosa i ,sa i =sina i The transformation matrix of the robot's end effector relative to the base is obtained by multiplying the transformation matrices of adjacent joints in sequence:

[0024]

[0025] (1-2) With the wrist joint as the center, the robot joint space is divided into two subspaces: the front 3-axis joint space and the rear 3-axis joint space. The front 3-axis joint space is from the robot base to the wrist, and its kinematic chain is as follows:

[0026]

[0027] The posterior 3-axis joint space extends from the robot wrist to the end effector, and its kinetic chain is as follows:

[0028]

[0029] In this context, subscripts B, W, and F represent the robot base, wrist, and end effector, respectively; the overall kinematic chain from the robot base to the end effector is represented as:

[0030]

[0031] Considering the pose error and expressing it in differential form, ignoring higher-order derivatives, the equivalent form on the right side of the equation is:

[0032]

[0033] The physical meaning of this equation is that the differential transformation or superposition of differential errors of the two subspaces is equivalent to the joint space before partitioning, and the joint space partitioning is completed.

[0034] In step (1), the step size and volume of the discretization space are determined according to the working range and accuracy requirements of the task; the mesh obtained by the first three-axis joint space discretization is similar to a fan-shaped body in the Cartesian coordinate system, and the mesh points obtained by the second three-axis joint space discretization are distributed on the sphere in the Cartesian coordinate system because the axes of the three joints intersect at the center of the wrist.

[0035] The specific process of step (2) is as follows:

[0036] (2-1) Move each joint of the robot to the set angle. In this state, the end of the robot is not in the Z-axis direction of the base coordinate system. Keep the other joints still and control joint 1 to rotate. Measure the coordinate value of the end of the robot at regular intervals. Perform circle fitting on the measured set of spatial points to obtain the center O1 and its normal. O1 is on the Z-axis of the base coordinate system and its normal is parallel to the Z-axis of the base coordinate system.

[0037] (2-2) Rotate robot joint 1 to the zero position and move other joints to the set angle. In this state, the robot end is not in the Y-axis direction of the base coordinate system. Keep the other joints stationary and control joint 2 to perform rotational motion. Measure the coordinate value of the robot end at regular intervals. Perform circle fitting on the measured set of spatial points to obtain the circle center O2 and its normal. The Z-axis coordinate value of O2 in the base coordinate system is 0, and its normal is parallel to the Y-axis of the base coordinate system.

[0038] (2-3) Determine the X-axis direction according to the right-hand rule, and determine the origin of the coordinate system according to O1 and O2; O1 determines the x and y values ​​of the origin, and O2 determines the z value of the origin. That is, the robot base coordinate system is obtained under the coordinate system of the tracker, and the transformation relationship is used to complete the station transfer.

[0039] In step (3), the special tooling refers to the tooling with three tracker target ball seats fixed. The tooling is installed at the end of the robot. The three target ball seats are not collinear. During measurement, the target ball is placed in three positions at the same sampling point, and the position coordinates are measured in sequence.

[0040] In step (3), information from three points is collected simultaneously at the same sampling point to calculate the position and attitude error corresponding to the sampling point. The specific process is as follows:

[0041] (3-1) At a certain sampling point, the measured location of the fixed point numbered i is P. i The theoretical location is P. i ′, i=1,2,3, calculating the pose error of the sampling point is equivalent to finding the transformation relationship between the theoretical point set and the measured point set, and the problem is transformed into a rigid point set matching problem;

[0042] (3-2) This problem is solved using the least squares method, and the objective function is:

[0043]

[0044] Where ΔR is a 3×3 rotation matrix and ΔP is a 3×1 translation vector; solving this equation yields the position and attitude errors ΔP and ΔR of the sampling point.

[0045] The specific process of step (4) is as follows:

[0046] (4-1) Fix the first three joints of the robot at a preset angle. In this state, the last three joints reach all the sampling points planned in step (1).

[0047] (4-2) Move joints 5 and 6 to the set angle. In this state, the robot end is not in the Z-axis direction of the wrist coordinate system. Keep the other joints still and control joint 4 to rotate. Measure the coordinate value of the robot end at regular intervals. Perform circle fitting on the measured set of spatial points to obtain the center O4 and its normal. O4 is on the Z-axis of the wrist coordinate system and its normal is parallel to the Z-axis of the wrist coordinate system.

[0048] (4-3) Rotate the robot's 4th joint to the zero position, keep the other joints still, control the 5th joint to rotate, measure the robot's end coordinate value at regular intervals, perform circle fitting on the measured set of spatial points, and obtain the circle center O5 and its normal. O5 is on the Y-axis of the wrist coordinate system, its Z-axis coordinate value is 0, and its normal is parallel to the Y-axis of the wrist coordinate system.

[0049] (4-4) Determine the X-axis direction according to the right-hand rule, and determine the location of the origin according to O4. That is, obtain the robot wrist coordinate system under the tracker coordinate system, and use its transformation relationship to complete the station transfer.

[0050] The specific process of step (6) is as follows:

[0051] (6-1) By sorting the Euclidean distances between the target point and the sampling point, the front three-axis joint space grid and the back three-axis joint space grid where the target point is located are determined respectively;

[0052] (6-2) Calculate the pose error of the target point using the position and pose error data of the mesh vertices, and determine the weight λ of each vertex using the inverse distance weighting method. i ,Right now:

[0053]

[0054] Where the subscripts i and j both range from 1 to 8, representing the grid vertex numbers, D(P i ) represents the target point P target The Euclidean distance to the i-th nearest grid vertex, where θ represents the joint angle of the target point. The subscripts 1, 2, and 3 represent the three joint angles. For the first three axes of the joint space, these correspond to joints 1 to 3, and for the last three axes of the joint space, they correspond to joints 4 to 6.

[0055] (6-3) Using the weights λ of each vertex i Its error ζ(P) i ) Calculate the target point error ζ(P) target ):

[0056] ζ(P target )=∑λ i ζ(P i )

[0057] (6-4) The calculated pose errors in the first three axes of joint space are denoted as μ. x1 μ y1 μ z1 ε x1 ε y1 ε z1 The first three terms represent position errors, and the last three terms represent attitude errors, all relative to the base coordinate system. Representing this in matrix form, we get:

[0058]

[0059] ΔT B It is defined in the base coordinate system, therefore:

[0060]

[0061] (6-5) The compensation matrix in the rear three-axis joint space is consistent with that in the front three-axis joint space, except that the attitude error term is 0:

[0062]

[0063] (6-6) The overall compensation transformation can be obtained through the above two equations:

[0064]

[0065] Compared with the prior art, the present invention has the following beneficial effects:

[0066] 1. This invention achieves dimensionality reduction through joint space partitioning, which reduces the workload of joint space mesh compensation methods while ensuring the positioning accuracy of the six-axis industrial robot after compensation.

[0067] 2. The six-axis industrial robot positioning error compensation method of the present invention can simultaneously compensate for the robot's geometric and non-geometric errors, and can achieve higher positioning accuracy after compensation. Attached Figure Description

[0068] Figure 1 This is a schematic diagram of a grid division according to the present invention;

[0069] Figure 2 This is a flowchart of a positioning error compensation method for a six-axis industrial robot according to the present invention. Detailed Implementation

[0070] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be noted that the embodiments described below are intended to facilitate the understanding of the present invention and do not constitute any limitation thereof.

[0071] like Figure 1 As shown, the meshes obtained by discretizing the front three-axis joint space and the rear three-axis joint space of a six-axis industrial robot are illustrated. The meshes obtained by discretizing the front three-axis joint space are shaped like a sector in the Cartesian coordinate system, while the meshes obtained by discretizing the rear three-axis joint space are distributed on a sphere in the Cartesian coordinate system because the axes of the three joints intersect at the center of the wrist.

[0072] like Figure 2 As shown, a positioning error compensation method for a six-axis industrial robot includes the following steps:

[0073] S101, perform kinematic modeling and joint space segmentation on the robot, and discretize the joint spaces of the front three axes and the rear three axes into meshes, using mesh points as sampling points. The method for kinematic modeling and joint space segmentation of the robot is as follows:

[0074] First, the DH method is used to model the kinematics of the robot, and the transformation relationship between adjacent links is as follows:

[0075]

[0076] Where T represents the transformation relationship between the link coordinate systems, x and z represent the x-axis and z-axis of the link coordinate system, i represents the link number, Rot represents rotation about the corresponding axis, Trans represents translation along the corresponding axis, θ is the joint angle, d is the link offset, a is the link length, and α is the link torsion angle. Expanded into the general form:

[0077]

[0078] Where, cθ i =cosθ i sθ i =sinθ i ca i =cosθ i ,sa i =sinθ i The transformation matrix of the robot's end effector relative to the base can be obtained by multiplying the transformation matrices of adjacent joints in sequence:

[0079]

[0080] Then, with the wrist joint as the center, the robot's joint space is divided into two subspaces: the front 3-axis joint space and the rear 3-axis joint space. The front 3-axis joint space runs from the robot base to the wrist, and its kinematic chain is as follows:

[0081]

[0082] The posterior 3-axis joint space extends from the robot wrist to the end effector, and its kinetic chain is as follows:

[0083]

[0084] In this context, subscripts B, W, and F represent the robot base, wrist, and end effector, respectively. The overall kinematic chain from the robot base to the end effector can be represented as:

[0085]

[0086] Considering the pose error and expressing it in differential form, ignoring higher-order derivatives, the equivalent form on the right side of the equation is:

[0087]

[0088] The physical meaning of this equation is that the differential transformation or superposition of differential errors of the two subspaces is equivalent to the joint space before partitioning, and the joint space partitioning is completed.

[0089] S102, Transfer the laser tracker to the robot's base coordinate system as follows:

[0090] First, each joint of the robot moves to a suitable angle. In this state, the robot's end effector is not in the Z-axis direction of the base coordinate system. Keeping the other joints stationary, control joint 1 to perform rotational motion. Measure the coordinate values ​​of the robot's end effector at regular intervals. Perform circle fitting on the measured set of spatial points to obtain the circle center O1 and its normal. O1 is on the Z-axis of the base coordinate system, and its normal is parallel to the Z-axis of the base coordinate system.

[0091] Then, rotate robot joint 1 to its zero position and move the other joints to appropriate angles. In this state, the robot's end effector is not in the Y-axis direction of the base coordinate system. Keeping the other joints stationary, control joint 2 to perform rotational motion, and measure the coordinate values ​​of the robot's end effector at regular intervals. Perform circle fitting on the measured set of spatial points to obtain the center O2 and its normal. The Z-axis coordinate value of O2 in the base coordinate system is 0, and its normal is parallel to the Y-axis of the base coordinate system.

[0092] Finally, the X-axis direction is determined using the right-hand rule, and the origin of the coordinate system is determined using O1 and O2. O1 determines the x and y values ​​of the origin, and O2 determines the z value of the origin. This allows us to obtain the robot's base coordinate system in the tracker's coordinate system, and the transformation relationship can be used to complete the station transfer.

[0093] S103, the fixture with three tracker target ball holders is installed at the robot's end effector. During measurement, the target ball is placed in three positions at the same sampling point. The method for simultaneously obtaining the position and attitude errors from the measurement results of the three positions is as follows:

[0094] First, at a certain sampling point, the measured position of the fixed point numbered i (i = 1, 2, 3) is P. i The theoretical location is P. i Calculating the pose error of the sampling point is equivalent to finding the transformation relationship between the theoretical point set and the measured point set, and the problem is transformed into a rigid point set matching problem.

[0095] Then, the least squares method is used to solve this problem, and the objective function is:

[0096]

[0097] Where ΔR is a 3×3 rotation matrix and ΔP is a 3×1 translation vector. Solving this equation yields the position and attitude errors ΔP and ΔR of the sampling point.

[0098] S104, Transfer the laser tracker to the robot wrist coordinate system as follows:

[0099] First, fix the robot’s first three joints at a suitable angle. In this state, the last three joints can reach all the sampling points planned in step (1).

[0100] Then, move joints 5 and 6 to a suitable angle. In this state, the robot's end effector is not in the Z-axis direction of the wrist coordinate system. Keep the other joints stationary and control joint 4 to perform rotational motion. Measure the coordinate values ​​of the robot's end effector at regular intervals. Perform circle fitting on the measured set of spatial points to obtain the circle center O4 and its normal. O4 is on the Z-axis of the wrist coordinate system, and its normal is parallel to the Z-axis of the wrist coordinate system.

[0101] Next, rotate the robot's 4th joint to the zero position, keep the other joints still, control the 5th joint to rotate, measure the robot's end coordinate value at regular intervals, perform circle fitting on the measured set of spatial points to obtain the circle center O5 and its normal. O5 is on the Y-axis of the wrist coordinate system, its Z-axis coordinate value is 0, and its normal is parallel to the Y-axis of the wrist coordinate system.

[0102] Finally, the X-axis direction is determined according to the right-hand rule, and the location of the origin is determined according to O4. The robot wrist coordinate system can then be obtained in the tracker coordinate system, and the transformation relationship can be used to complete the station change.

[0103] S105 uses a laser tracker to track and measure the sampling points in the rear three-axis joint space, and calculates the positioning error corresponding to the sampling points.

[0104] S106, Calculate the pose error of the target point based on the pose error of the sampling points and the positioning error, using the following method:

[0105] First, by sorting the Euclidean distances between the target point and the sampling point, the front three-axis joint space grid and the back three-axis joint space grid where the target point is located are determined respectively.

[0106] The pose error of the target point is calculated using the pose error data of the mesh vertices, and the weight λ of each vertex is determined using the inverse distance weighting method. i ,Right now:

[0107]

[0108] Where the subscripts i and j both range from 1 to 8, representing the grid vertex numbers, D(P i ) represents the target point P target The Euclidean distance to the i-th nearest grid vertex, where θ represents the joint angle of the target point. The subscripts 1, 2, and 3 represent the three joint angles. For the first three axes of the joint space, these correspond to joints 1 to 3, and for the last three axes of the joint space, they correspond to joints 4 to 6.

[0109] Using the weights λ of each vertex i Its error ζ(P) i ) Calculate the target point error ζ(P) target ):

[0110] ζ(P target )=∑λ i ζ(P i )

[0111] Then, the calculated pose errors in the first three axes of joint space are denoted as μ. x1 μ y1 μ z1 ε x1 ε y1 ε z1 The first three terms represent position errors, and the last three terms represent attitude errors (all relative to the base coordinate system). Representing this in matrix form, we get:

[0112]

[0113] ΔT B It is defined in the base coordinate system, therefore:

[0114]

[0115] The compensation matrix in the latter three-axis joint space is the same as that in the former three-axis joint space, except that the attitude error term is 0:

[0116]

[0117] Finally, the overall compensation transformation can be obtained through the above two equations:

[0118]

[0119] S107, compensate for the positioning error of the target point based on the pose error of the target point.

[0120] The embodiments described above provide a detailed explanation of the technical solutions and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, additions, and equivalent substitutions made within the scope of the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A six-axis industrial robot positioning error compensation method, characterized by, The method comprises the following steps: (1) kinematic modeling and joint space partitioning are performed on a six-axis industrial robot, and the front three-axis joint space and the rear three-axis joint space are discretized by meshing respectively, and the mesh points are used as sampling points; wherein the front three-axis joint space is from the robot base to the center of the robot wrist, and the rear three-axis joint space is from the center of the robot wrist to the end of the robot; (2) the laser tracker is converted to the robot base coordinate system; the specific process is as follows: (2-1) the joints of the robot are moved to the set angles, and in this state, the end of the robot is not in the Z-axis direction of the base coordinate system; the other joints are kept unchanged, the 1 joint is controlled to rotate, the coordinate values of the end of the robot are measured every certain angle, the measured space point set is circularly fitted to obtain the center O1 and the normal thereof, O1 is on the Z-axis of the base coordinate system, and the normal thereof is parallel to the Z-axis of the base coordinate system; (2-2) the 1 joint of the robot is rotated to zero position, and the other joints are moved to the set angles, and in this state, the end of the robot is not in the Y-axis direction of the base coordinate system; the other joints are kept unchanged, the 2 joint is controlled to rotate, the coordinate values of the end of the robot are measured every certain angle, the measured space point set is circularly fitted to obtain the center O2 and the normal thereof, O2 is in the Z direction of the base coordinate system, and the normal thereof is parallel to the Y-axis of the base coordinate system; (2-3) the X-axis direction is determined according to the right-hand rule, and the origin of the coordinate system is determined according to O1 and O2; O1 determines the x and y values of the origin, and O2 determines the z value of the origin, that is, the robot base coordinate system is obtained in the tracker coordinate system, and the conversion relationship is used to complete the conversion; (3) the sampling points of the front three-axis joint space are tracked and measured by the laser tracker, three point position information is collected at the same sampling point by using a special tool, and the position and attitude errors corresponding to the sampling points are calculated; the special tool refers to a tool fixed with three tracker target ball seats, the tool is installed at the end of the robot, the three target ball seats are not collinear, the target balls are placed at the three positions at the same sampling point during measurement, and the position coordinates are measured in turn; (4) the laser tracker is converted to the robot wrist coordinate system; the specific process is as follows: (4-1) the first three joints of the robot are fixed at the preset angles, and in this state, the last three joints reach all the sampling points planned in step (1); (4-2) the 5 and 6 joints are moved to the set angles, and in this state, the end of the robot is not in the Z-axis direction of the wrist coordinate system; the other joints are kept unchanged, the 4 joint is controlled to rotate, the coordinate values of the end of the robot are measured every certain angle, the measured space point set is circularly fitted to obtain the center O4 and the normal thereof, O4 is on the Z-axis of the wrist coordinate system, and the normal thereof is parallel to the Z-axis of the wrist coordinate system; (4-3) the 4 joint of the robot is rotated to zero position, the other joints are kept unchanged, the 5 joint is controlled to rotate, the coordinate values of the end of the robot are measured every certain angle, the measured space point set is circularly fitted to obtain the center O5 and the normal thereof, O5 is on the Y-axis of the wrist coordinate system, the Z-axis coordinate value thereof is 0, and the normal thereof is parallel to the Y-axis of the wrist coordinate system; (4-4) According to the right-hand rule, the X-axis direction is determined, and the position of the origin is determined according to O4, that is, the robot wrist coordinate system is obtained under the tracking instrument coordinate system, and the conversion relationship is used to complete the station conversion; (5) The sampling points in the rear three-axis joint space are tracked and measured by the laser tracker, and the positioning errors corresponding to the sampling points are calculated; (6) The pose error of the target point is calculated according to the sampling point pose error obtained in step (3) and the sampling point positioning error obtained in step (5); the specific process is: (6-1) The size of the Euclidean distance between the target point and the sampling point is sorted to determine the front three-axis joint space grid and the rear three-axis joint space grid where the target point is located; (6-2) Calculate the pose error of the target point using the position and pose error data of the grid vertices, and determine the weight λ of each vertex using the inverse distance weighting method i That is: wherein, the subscript i, j are both in the range of 1-8, representing the grid vertex number, D(P i ) represents the Euclidean distance from the target point P target to the nearest i-th grid vertex, θ represents the joint angle of the target point, represents the joint angle of the grid vertex, the subscripts 1, 2, 3 represent three joint angles, for the first three-axis joint space, corresponding to joints 1-3, for the last three-axis joint space, corresponding to joints 4-6; (6-3) Using each vertex weight λ i The error ζ(P i ) of the target point is calculated. target ) ζ(P target ) = ∑λ i ζ(P i ) (6-4) The calculated pose errors in the first three axis joint spaces are denoted as μ x1 , μ y1 , μ z1 , ε x1 , ε y1 , and ε z1 , respectively, where the first three are position errors and the last three are attitude errors, both relative to the base coordinate system; and written in matrix form as follows: ΔT B is defined in the base coordinate system, so that: (6-5) The compensation matrix in the rear three-axis joint space is consistent with that in the front three-axis joint space, except that the attitude error term is 0: (6-6) Through the above two formulas, the overall compensation transformation is obtained: Wherein, the subscripts B, W and F respectively represent the robot base, wrist and end; (7) The positioning error of the target point is compensated according to the pose error obtained in step (6).

2. The six-axis industrial robot positioning error compensation method of claim 1, wherein, In step (1), the method for kinematic modeling and joint space segmentation of the six-axis industrial robot is as follows: (1-1) The D-H method is used for kinematic modeling of the robot, and the transformation relationship between adjacent links is: Wherein, T represents the transformation relationship between the link coordinate systems, x and z respectively represent the x coordinate axis and z coordinate axis of the link coordinate system, i represents the link number, Rot represents rotation around the corresponding axis, Trans represents translation along the corresponding axis, θ is the joint angle, d is the link offset, a is the link length, and α is the link twist angle. Expanding into the general form: wherein cθ i = cos θ i , sθ i = sin θ i , ca i = cos a i , sa i = sin a i The transformation matrix of the robot end effector relative to the base is obtained by sequentially multiplying the adjacent joint transformation matrices: (1-2) Taking the wrist joint as the center, the robot joint space is divided into two subspaces, the front 3 joint space and the rear 3 joint space; the front 3-axis joint space is from the robot base to the wrist, and its motion chain is: The rear 3-axis joint space is from the robot wrist to the end, and its motion chain is: Wherein, the subscripts B, W and F respectively represent the robot base, wrist and end; the overall motion chain of the robot base to the end is represented as: Considering the pose error and expressing it in differential form, ignoring the high-order derivative, the equivalent form on the right side of the equation is: The physical meaning of this equation is that the differential transformation or differential error of the two subspaces is equivalent to the joint space before the joint space is divided, and the joint space segmentation is completed.

3. The six-axis industrial robot positioning error compensation method of claim 1, wherein, In step (1), the step size and volume of the discretized space are determined according to the task working range and accuracy requirements; the grid obtained by discretizing the front three-axis joint space is similar to a fan-shaped body in the Cartesian coordinate system, and the grid points obtained by discretizing the rear three-axis joint space are distributed on the spherical surface in the Cartesian coordinate system, because the axes of the three joints intersect at the wrist center.

4. The six-axis industrial robot positioning error compensation method of claim 1, wherein, In step (3), three point position information is collected at the same sampling point to calculate the position and attitude error of the sampling point, and the specific process is: (3-1) At a certain sampling point, the measured position of the fixed point numbered i is P i , the theoretical position is P i ', i = 1, 2, 3, and the pose error of the sampling point is equivalent to finding the conversion relationship between the theoretical point set and the measured point set, which is converted into a rigid point set matching problem; (3-2) The least squares method is used to solve this problem, and the objective function is: Wherein, ΔR is a 3x3 rotation matrix, and ΔP is a 3x1 translation vector; solving this equation gives the position and attitude error ΔP, ΔR of the sampling point.

Images