A robot calibration method based on position and distance constraints

Abstract

The present invention discloses a robot calibration method based on position and distance constraints, including establishing a robot kinematic model and a kinematic error model based on position and distance constraints; installing an end calibration device to the end of the robot, and installing a geometric constraint device in the robot's workspace; dragging the robot so that the calibration ball of the end calibration device is constrained to each ball seat on the geometric constraint device, and each ball seat touches several times in different configurations, reading and recording the joint angle data measured each time; identifying the corresponding robot kinematic model parameters; and compensating the kinematic model parameter errors obtained by identification to the robot controller. The present invention has the advantages of low cost, good portability, simple operation, and not easy to damage.

Description

Technical Field

[0001] The invention belongs to the technical field of robot calibration, and in particular relates to a robot calibration method based on position and distance constraints. Background Art

[0002] Collaborative robots are a new type of industrial robot that can work closely with people. Compared with traditional industrial robots, collaborative robots have the advantages of light weight, good flexibility, high safety, draggable teaching, easy deployment and implementation, and support for human-machine collaboration. They can not only meet the growing small-batch and multi-variety production needs of the manufacturing industry, but also be applied to the field of social services to achieve safe and friendly human-machine interaction, and have extremely broad development prospects.

[0003] However, due to certain errors in the processing and assembly of parts by collaborative robots, their absolute positioning accuracy is poor. Therefore, in order to improve the absolute positioning accuracy of collaborative robots, it is often necessary to calibrate the robots.

[0004] Domestic and foreign scholars have conducted fruitful research on robot calibration and established a robot kinematic calibration method consisting of four main steps: error modeling, posture measurement, parameter identification, and error compensation, which effectively improved the absolute positioning accuracy of industrial robots. However, most of the existing robot kinematic calibration methods rely on external precision measuring equipment such as laser trackers, arm-type three-coordinate measuring instruments, and wire-drawing measurement systems to measure the robot position or posture. These large-scale precision measuring equipment are expensive, have high use and maintenance costs, poor convenience, and are difficult to deploy and implement on site. They are difficult to meet the regular on-site calibration needs of collaborative robots.

[0005] In response to the above problems, many researchers have begun to explore low-cost, portable self-calibration devices in recent years. The industrial robot calibration device disclosed in CN107042528A contacts the target sphere fixed on the table with three detection rods fixed at the end of the robot, reads the readings of three displacement sensors, touches the same or different spheres twice, and calibrates the robot using the deviation between the nominal distance and the actual value. However, when the displacement sensor of the above device contacts the sphere, the force direction does not coincide with the measurement direction of the displacement sensor, which easily causes the force deformation of the detection rod, thereby affecting the subsequent calibration accuracy of the device.

[0006] How to provide a collaborative robot calibration solution that is low-cost, portable, and not prone to damage is an issue that needs to be addressed urgently. Summary of the invention

[0007] The main purpose of the present invention is to provide a robot calibration method based on position and distance constraints which is low-cost, portable and not prone to damage, thereby overcoming the deficiencies of the prior art.

[0008] To achieve the above-mentioned invention object, the technical solution adopted by the present invention includes: a robot calibration method based on position and distance constraints, the method is implemented based on a robot calibration device, the robot calibration device includes an end calibration device and a geometric constraint device, the end calibration device includes a connecting seat and a calibration ball installed at one end of the connecting seat, the geometric constraint device includes a constraint support seat and a plurality of ball seats arranged on the constraint support seat, the method includes:

[0009] S1, establishing a model, wherein the model includes a kinematic model of the robot, a first kinematic error model based on position constraints, and a second kinematic error model based on distance constraints;

[0010] S2, measurement and installation, including measuring a first relative posture between the calibration ball on the terminal calibration device and the connecting seat, and a second relative posture between the ball seats on the geometric constraint device, and after the measurement, installing the terminal calibration device to the terminal of the robot, and installing the geometric constraint device into the working space of the robot;

[0011] S3, data acquisition, including changing the position of the geometric constraint device in the workspace of the robot for multiple times, dragging the robot at each position, constraining the calibration ball of the end calibration device to each ball seat on the geometric constraint device, and each ball seat touches several times in different configurations, reading and recording the joint angle data after each measurement operation is stabilized;

[0012] S4, parameter identification, including dividing the joint angle data into several groups according to different geometric constraint device positions, and further dividing the joint angle data into several small groups according to different ball seat positions, substituting the data of the same group into the first kinematic error model in pairs, substituting the data of the same group but different groups into the second kinematic error model in pairs, and then mixing the two error model data to identify the kinematic model parameters of the corresponding robot;

[0013] S5, error compensation, including compensating the identified kinematic model parameter errors into the robot controller.

[0014] In a preferred embodiment, in S1, the correspondence between the robot end posture and the joint angle, joint rotation and initial posture is established through the global exponential product formula to construct the kinematic model; the correspondence between the position constraint error and the joint rotation error and the initial posture error is established through the accompanying transformation matrix to construct the first kinematic error model; the correspondence between the distance constraint error and the joint rotation error and the initial posture error is established through the accompanying transformation matrix to construct the second kinematic error model.

[0015] In a preferred embodiment, the global exponential product formula is expressed as follows:

[0016]

[0017] Among them, T 0，n+1 Represents the coordinates of the robot's end position in the base coordinate system, s i (i=1, 2, ..., n) represents the coordinates of the robot's joint rotation in the base coordinate system, q i (i=1, 2, ..., n) represents the rotation angle of each joint of the robot, that is, the joint angle, T 0，n+1 (0) represents the initial position of the robot end relative to the base coordinate system.

[0018] In a preferred embodiment, the first kinematic error model is expressed as follows:

[0019] Y PC =A PC X;

[0020] in, represents the end position error under two different configurations, X represents the model parameter error to be identified, and A PC Represents the position constraint relationship matrix between the terminal position error and the model parameter error, the A PC It is expressed as:

[0021]

[0022] Where p is the nominal end position coordinate, I 3 represents the 3×3 identity matrix, Ad(·) represents the adjoint transformation matrix corresponding to a homogeneous transformation matrix, A j (j=1, 2) is expressed as:

[0023]

[0024] In a preferred embodiment, the second kinematic error model is expressed as follows:

[0025] Y DC =A DC X,;

[0026] in, represents the deviation between the nominal end distance and the actual value under two different configurations, X represents the model parameter error to be identified, and A DC Represents the distance constraint relationship matrix between the terminal distance error and the model parameter error, the A DC It is expressed as:

[0027]

[0028] In a preferred embodiment, the first kinematic error model and the second kinematic error model are mixed to form a mixed error model, and the mixed error model is expressed as:

[0029] Y = AX;

[0030]

[0031]

[0032] Where Y is m 1 The end position error and m 2 The vector composed of the terminal distance errors, A represents the vector composed of m 1 position constraint relationship matrix and m 2 The combined relationship matrix composed of distance constraint relationship matrices, m 1 and m 2 is an integer greater than or equal to 1.

[0033] In a preferred embodiment, in S4, the kinematic model parameter error is identified by iterative least square method, and the least square method formula is expressed as:

[0034] X=(A T A) -1 A T Y.

[0035] In a preferred embodiment, in S5, the identified kinematic model parameter errors are compensated to the robot controller by direct compensation or indirect compensation.

[0036] In a preferred embodiment, the direct compensation method includes directly modifying the kinematic model parameters in the controller.

[0037] In a preferred embodiment, the indirect compensation method includes correcting the target posture by identifying the kinematic model parameters, inputting the corrected target posture into the original controller, and performing error compensation on the original kinematic model.

[0038] Compared with the prior art, the beneficial effects of the present invention are at least:

[0039] For the position constraint of the robot, the present invention drags the calibration ball at the end to the same ball seat in two different robot configurations, and uses the terminal position error calculated twice to calibrate the robot; for the distance constraint, the calibration ball at the end is dragged to two different ball seats, and the deviation between the terminal distance calculated twice and the actual distance is used to calibrate the robot. Compared with traditional external calibration devices, the present invention has the characteristics of low cost and good portability; compared with most existing self-calibration devices, the present invention does not contain additional sensors, and has the advantages of simple operation and not easy to damage. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0041] Figure 1 is a schematic diagram of the structure of the device according to an embodiment of the present invention;

[0042] Figure 2 is a schematic flow chart of a method in one embodiment of the present invention;

[0043] Figure 3 It is a schematic diagram of the structure of the terminal calibration device of the present invention;

[0044] Figure 4 It is a structural schematic diagram of the geometric constraint device of the present invention;

[0045] Figure 5 It is a schematic diagram of the calibration ball being positioned on the ball seat;

[0046] Figure 6 is a schematic diagram of the calibration ball being positioned on the ball seat under different robot configurations;

[0047] Figure 7 It is a schematic diagram of the calibration ball being positioned on different ball seats.

[0048] Reference numerals:

[0049] 1. Robot, 2. Connecting seat, 3. Calibration ball, 4. Ball seat, 5. Constraint support seat. DETAILED DESCRIPTION

[0050] By attaching Figure 1The present invention will be more fully understood by reading the following detailed description. Detailed embodiments of the present invention are disclosed herein; however, it should be understood that the disclosed embodiments are only exemplary of the present invention, and the present invention can be embodied in various forms. Therefore, the specific functional details disclosed herein should not be interpreted as limiting, but only as the basis of the claims and as a representative basis for teaching those skilled in the art to adopt the present invention in different ways in virtually any appropriate detailed embodiment.

[0051] The present invention discloses a robot calibration method based on position and distance constraints, which is implemented based on a robot calibration device. In view of the problems of the existing calibration devices, such as high price, poor portability, and easy damage, the present invention utilizes the characteristics of collaborative robots that can be dragged for teaching, and provides a calibration device for collaborative robots that is low-cost, portable, and not easy to damage. For general industrial robots, if the dragging teaching function can be realized, the present invention can also be applied.

[0052] like Figure 1 As shown, the robot calibration device includes an end calibration device and a geometric constraint device, wherein the robot calibration device includes an end calibration device and a geometric constraint device. Figure 3 As shown, the end calibration device is installed at the end of the robot 1, which specifically includes a connecting seat 2 and a calibration ball 3 installed at one end of the connecting seat 2; Figure 4 As shown, the geometric constraint device is installed in the working space of the robot 1, which mainly includes a constraint support seat 5 and a plurality of ball seats 4 arranged on the constraint support seat 5. During calibration, the end calibration device is installed at the end of the robot 1, specifically, the connecting seat 2 is installed at the end of the robot 1; the geometric constraint device is installed in the working space of the robot 1. The robot 1 is dragged so that the calibration ball 3 on the end calibration device is adsorbed in the ball seat 4 on the geometric constraint device through a magnet (not shown). For position constraints, the calibration ball 3 at the end is dragged to the same ball seat 4 in two different robot configurations, and the robot 1 is calibrated using the end position error calculated twice; for distance constraints, the calibration ball 3 at the end is dragged to two different ball seats 4, and the robot 1 is calibrated using the error between the end distance calculated twice and the actual distance.

[0053] like Figure 2 As shown, the present invention discloses a robot calibration method based on position and distance constraints, which specifically includes the following steps:

[0054] S1, establishing a model, wherein the model includes a kinematic model of the robot, a first kinematic error model based on position constraints, and a second kinematic error model based on distance constraints.

[0055] Specifically, the present invention calibrates the robot based on position and distance constraints respectively. In this embodiment, a global exponential product formula is used to establish a kinematic model of the robot, specifically, by establishing a corresponding relationship between the robot's end position and the joint angle, joint rotation and initial position, to construct the kinematic model. The global exponential product formula is expressed as follows:

[0056]

[0057] Among them, T 0，n+1 Represents the coordinates of the robot's end position in the base coordinate system, s i (i=1, 2, ..., n) represents the coordinates of the robot's joint rotation in the base coordinate system, q i (i=1, 2, ..., n) represents the rotation angle of each joint of the robot, that is, the joint angle, T 0，n+1 (0) represents the initial position of the robot end relative to the base coordinate system. The base coordinate determination process here mainly includes: the installation position of the robot relative to the position of the geometric constraint device can be predicted (such as can be determined by measuring methods such as three-dimensional scanners, laser trackers, and three-coordinate measuring instruments). The actual base coordinate position can be calibrated by the calibration device of the present invention and then calibrated by the calibration method of the present invention, and then the base coordinate system of the robot can be calibrated again.

[0058] In the above robot kinematic model, q i (i=1,2,...,n) can be directly read by the robot encoder, and s i (i=1, 2, ..., n) and T 0，n+1 Parameter identification is required. i (i=1, 2, ..., n) and T 0，n+1 (0) can be expressed as 6 parameter variables to be identified, so there are 6(n+1) parameters to be identified. The errors t corresponding to these parameters are 1 , t 2 , ..., t n , t 0 Represented as a vector to be identified By deducing certain formulas, the relationship between the end position error and the kinematic model parameter error under two different configurations can be established, that is, the first kinematic error model is expressed as follows:

[0059] Y PC =A PC X,;

[0060] in, represents the end position error under two different configurations, X represents the model parameter error to be identified, and A PCRepresents the position constraint relationship matrix between the terminal position error and the model parameter error, the A PC It is expressed as:

[0061]

[0062] Where p is the nominal end position coordinate, I 3 represents the 3×3 identity matrix, Ad(·) represents the adjoint transformation matrix corresponding to a homogeneous transformation matrix, A j (j=1, 2) is expressed as:

[0063]

[0064] In this embodiment, similarly, a distance-based robot kinematic error model can also be established, that is, the second kinematic error model is expressed as follows:

[0065] Y DC =A DC X,;

[0066] in, represents the deviation between the nominal end distance and the actual value under two different configurations, X also represents the model parameter error to be identified, A DC Represents the distance constraint relationship matrix between the terminal distance error and the model parameter error, the A DC It is expressed as:

[0067]

[0068] Furthermore, the first kinematic error model and the second kinematic error model are mixed to form a mixed error model. The mixed error model specifically integrates m 1 The samples with position constraints are m 2 The mixed error model is specifically expressed as:

[0069] Y = AX;

[0070]

[0071]

[0072] Where Y is m 1 The end position error and m 2 The vector composed of the terminal distance errors, A represents the vector composed of m 1 position constraint relationship matrix and m 2 The combined relationship matrix composed of distance constraint relationship matrices, m 1 and m 2 is an integer greater than or equal to 1.

[0073] S2, measurement and installation, including measuring the first relative posture between the calibration ball on the end calibration device and the connecting seat, and the second relative posture between the ball seats on the geometric constraint device. After measurement, the end calibration device is installed to the end of the robot, and the geometric constraint device is installed in the working space of the robot.

[0074] During implementation, a three-dimensional scanner or other equipment may be used to measure the first relative posture and the second relative posture, where the first relative posture and the second relative posture are used as reference data for subsequent parameter identification.

[0075] S3, data acquisition, includes changing the position of the geometric constraint device in the workspace of the robot multiple times, dragging the robot at each position, so that the calibration ball of the end calibration device is constrained to each ball seat on the geometric constraint device, and each ball seat is touched several times in different configurations, and reading and recording the joint angle data after each measurement operation is stabilized.

[0076] Specifically, in combination Figure 5 and Figure 6 As shown in the figure, the process of position constraint data collection is as follows: a calibration ball of the end calibration device is moved to any ball seat on the geometric constraint device, and then moved to the same ball seat with different robot configurations, and the joint angle readings after each operation is stabilized are read and recorded, so that a set of position constraint calibration data is collected. In order to reduce the influence of measurement error and non-geometric error, the position and posture of the geometric constraint device in the robot workspace can be changed multiple times to collect multiple sets of data.

[0077] Combination Figure 5 to Figure 7 As shown in the figure, the process of distance constraint data collection is as follows: a calibration ball of the end calibration device is moved to a ball seat on the geometric constraint device, and then moved to different ball seats with different robot configurations, and the joint angle readings after each operation is stabilized are read and recorded, so that a set of distance constraint calibration data is collected. Similarly, in order to reduce the influence of measurement errors and non-geometric errors, the position of the geometric constraint device in the robot workspace can be changed multiple times to collect multiple sets of data.

[0078] In the actual data collection process, the position constraint and distance constraint calibration data can be measured at the same time, and the data processing is slightly different in the subsequent identification calculation process. For example, by repeatedly changing the position of the geometric constraint device in the robot's workspace, the robot is dragged at each position so that the calibration ball of the end calibration device is constrained to each ball seat on the geometric constraint device, and each ball seat touches several times in different configurations, and the joint angle data after each measurement operation is stabilized is read and recorded. After that, all joint angle data are divided into several groups according to different geometric constraint device postures, and then divided into several small groups according to different ball seat positions, wherein the data of the same group paired in pairs are multiple groups of position constraint calibration data, and the data of the same group but different groups paired in pairs are multiple groups of distance constraint calibration data.

[0079] S4, parameter identification, includes dividing the joint angle data into several groups according to different geometric constraint device postures, and then dividing them into several small groups according to different ball seat positions, pairing the data of the same group into the first kinematic error model, pairing the data of the same group but different groups into the second kinematic error model, and then mixing the two error model data to identify the corresponding robot's kinematic model parameters.

[0080] Specifically, in this embodiment, the least square method is used to iteratively identify the kinematic model parameter errors. For the least square method, its formula can be expressed as:

[0081] X=(A T A) -1 A T Y.

[0082] The kinematic model parameters are corrected using the model parameter errors calculated at each step, and then the corrected parameters are substituted into the first kinematic error model and the second kinematic error model for calculation again, and this process is repeated until the kinematic model parameters no longer change and the iteration is stopped. The kinematic model parameters at this point can be regarded as the actual model parameters.

[0083] S5, error compensation, including compensating the identified kinematic model parameter errors into the robot controller.

[0084] Specifically, the kinematic model parameter errors obtained by identification are compensated to the controller. Generally, there are two compensation methods: direct compensation and indirect compensation. Direct compensation is to directly modify the kinematic model parameters in the controller, while indirect compensation is to correct the target posture through the calibrated kinematic model parameters, and input the corrected target posture into the original controller to achieve error compensation for the original kinematic model.

[0085] Compared with the traditional external robot calibration scheme, the present invention has the characteristics of low cost and good portability; and compared with most existing self-calibration devices, the present invention does not contain additional sensors and has the advantages of simple operation and not easy to damage.

[0086] The various aspects, embodiments, features and examples of the present invention should be considered as illustrative in all aspects and are not intended to limit the present invention, the scope of the present invention is defined only by the claims. Other embodiments, modifications and uses will be apparent to those skilled in the art without departing from the spirit and scope of the claimed invention.

[0087] The use of headings and sections in this disclosure is not meant to limit the disclosure; each section may apply to any aspect, embodiment, or feature of the disclosure.

Claims

1. A robot calibration method based on position and distance constraints, characterized in that: The method is implemented based on a robot calibration device, the robot calibration device includes an end calibration device and a geometric constraint device, the end calibration device includes a connecting seat and a calibration ball installed at one end of the connecting seat, the geometric constraint device includes a constraint support seat and a plurality of ball seats arranged on the constraint support seat, and the method includes: S1, establishing a model, wherein the model includes a kinematic model of the robot, a first kinematic error model based on position constraints, and a second kinematic error model based on distance constraints; S2, measurement and installation, including measuring a first relative posture between the calibration ball on the terminal calibration device and the connecting seat, and a second relative posture between the ball seats on the geometric constraint device, and after the measurement, installing the terminal calibration device to the terminal of the robot, and installing the geometric constraint device into the working space of the robot; S3, data acquisition, including changing the position of the geometric constraint device in the workspace of the robot for multiple times, dragging the robot at each position, constraining the calibration ball of the end calibration device to each ball seat on the geometric constraint device, and each ball seat touches several times in different configurations, reading and recording the joint angle data after each measurement operation is stabilized; S4, parameter identification, including dividing the joint angle data into several groups according to different geometric constraint device positions, and further dividing the joint angle data into several small groups according to different ball seat positions, substituting the data of the same group into the first kinematic error model in pairs, substituting the data of the same group but different groups into the second kinematic error model in pairs, and then mixing the two error model data to identify the kinematic model parameters of the corresponding robot; S5, error compensation, including compensating the identified kinematic model parameter errors into the robot controller; The first kinematic error model is expressed as follows: AND PC =A PC X, in, represents the end position error under two different configurations, X represents the model parameter error to be identified, and A PC Represents the position constraint relationship matrix between the terminal position error and the model parameter error, the A PC It is expressed as: Where p is the nominal end position coordinate, I3 represents the 3×3 unit matrix, A j (j=1, 2) is expressed as: Where Ad(·) represents the adjoint transformation matrix corresponding to a homogeneous transformation matrix, T 0，n+1 Represents the coordinates of the robot's end position in the base coordinate system, s t (i=1, 2, ..., n) represents the coordinates of the robot's joint rotation in the base coordinate system, q i (i=1, 2, ..., n) represents the rotation angle of each joint of the robot, that is, the joint angle, T 0，n+1 (0) represents the initial position of the robot end relative to the base coordinate system; The second kinematic error model is expressed as follows: AND DC =A DC X, in, represents the deviation between the nominal end distance and the actual value under two different configurations, X represents the model parameter error to be identified, and A DC Represents the distance constraint relationship matrix between the terminal distance error and the model parameter error, the A DC It is expressed as:

2. The robot calibration method based on position and distance constraints according to claim 1, characterized in that: In S1, the correspondence between the robot's end posture and the joint angle, joint rotation and initial posture is established through the global exponential product formula to construct the kinematic model; the correspondence between the position constraint error and the joint rotation error and the initial posture error is established through the accompanying transformation matrix to construct the first kinematic error model; the correspondence between the distance constraint error and the joint rotation error and the initial posture error is established through the accompanying transformation matrix to construct the second kinematic error model.

3. The robot calibration method based on position and distance constraints according to claim 2, characterized in that: The global exponential product formula is expressed as follows:

4. The robot calibration method based on position and distance constraints according to claim 1, characterized in that: The first kinematic error model and the second kinematic error model are mixed to form a mixed error model, and the mixed error model is expressed as: Y = AX; Among them, Y is a vector composed of m1 terminal position errors and m2 terminal distance errors, A represents a combined relationship matrix composed of m1 position constraint relationship matrices and m2 distance constraint relationship matrices, and m1 and m2 are integers greater than or equal to 1.

5. The robot calibration method based on position and distance constraints according to claim 4, characterized in that: In S4, the kinematic model parameter error is identified by iterative least square method, and the least square method formula is expressed as: X=(A T TO) -1 TO T AND.

6. The robot calibration method based on position and distance constraints according to claim 1, characterized in that: In S5, the kinematic model parameter errors obtained by identification are compensated to the robot controller by direct compensation or indirect compensation.

7. The robot calibration method based on position and distance constraints according to claim 6, characterized in that: The direct compensation method includes directly modifying the kinematic model parameters in the controller.

8. The robot calibration method based on position and distance constraints according to claim 6, characterized in that: The indirect compensation method includes correcting the target posture through the kinematic model parameters obtained by identification, inputting the corrected target posture into the original controller, and performing error compensation on the original kinematic model.

Images