Industrial robot calibration method and device based on dial indicator

Abstract

The invention provides an industrial robot calibration method and device based on a dial indicator, and the method comprises the steps: collecting the position data of the tail end of a mechanical arm through the dial indicator, and carrying out the preprocessing of the position data of the tail end of the mechanical arm; calculating the tail end position error of the mechanical arm based on the tail end position data of the mechanical arm, and determining an error threshold value and a training iteration round number upper limit; initializing a DH parameter of the mechanical arm, and setting an iteration round number control variable; judging whether the current DH parameter updating meets an updating condition or not; judging whether the iteration round number control variable is greater than a training iteration round number upper limit or not; if the DH parameter meets the updating condition and the iteration round number control variable does not exceed the training iteration round number upper limit, a robot calibration model composed of the robot DH parameter and the joint angle is constructed; and calibrating the DH parameters of the mechanical arm by using an LM algorithm based on the robot calibration model. According to the method, the precision and efficiency of robot calibration are remarkably improved by introducing the elastic equilibrium factor and self-adaptive parameter control.

Description

Technical Field

[0001] The invention relates to the technical field of industrial robots, and in particular to an industrial robot calibration method and device based on a micrometer. Background Art

[0002] Industrial robot technology is developing rapidly, playing a significant role in welding, handling, assembly, and testing in industrial production. It is constantly penetrating into mechanical processing, electronic appliances, food safety, and life services, improving the quality of human life.

[0003] At present, the repeatability of industrial robots is very high, but the absolute positioning accuracy of robots is low at this stage, and its error can reach 10mm or even higher. This also greatly limits the application of industrial robots in the fields of aviation, chip processing, etc. Therefore, improving the absolute positioning accuracy of industrial robots plays a vital role in the development of industrial robots. There are many factors that affect the absolute positioning accuracy of robots, which can be mainly divided into geometric factors and non-geometric factors. Geometric factors mainly refer to the errors between the nominal geometric parameters and the actual values ​​of robots caused by the processing and assembly of robots. Non-geometric factors mainly refer to the elastic deformation caused by the load and the deadweight of the connecting rod, the error of the tightness between adjacent joints, and the error caused by the external environment. Among them, the geometric error of industrial robots has the greatest impact, accounting for about 90% of all error sources.

[0004] At present, robot calibration technology is a common method to improve the absolute positioning accuracy of robots. Among them, laser tracker is a traditional measuring instrument with an accuracy of ±0.1mm, but its high price limits its wide application. Summary of the invention

[0005] The present invention aims to at least solve the technical problem of high cost in the prior art, and in particular innovatively proposes an industrial robot calibration method and device based on a micrometer.

[0006] In order to achieve the above-mentioned object of the present invention, the present invention provides an industrial robot calibration method based on a micrometer, the method comprising:

[0007] S1. Using a micrometer to collect the position data of the end of the robot arm, and preprocessing the position data of the end of the robot arm;

[0008] S2. Calculate the position error of the end of the manipulator based on the end position data of the manipulator, and determine the error threshold and the upper limit of the number of training iterations;

[0009] S3, initialize the robot arm DH parameters and set the iteration round number control variable;

[0010] S4. Determine whether the current DH parameter update meets the update condition;

[0011] S5, judging whether the iteration round control variable is greater than the upper limit of the training iteration round;

[0012] S6. If the DH parameters meet the update conditions and the iteration round control variable does not exceed the upper limit of the training iteration rounds, a robot calibration model consisting of the robot DH parameters and joint angles is constructed;

[0013] S7. Calibrate the DH parameters of the robot arm using the Levenberg-Marquardt algorithm based on the robot calibration model.

[0014] As an optional embodiment of the present invention, optionally, constructing a robot calibration model consisting of robot DH parameters and joint angles in step S6 includes:

[0015] S601, constructing a loss function based on the end position data of the robotic arm;

[0016] S602, using the least squares method and the opposite direction of the objective function gradient as the learning direction, locally updating the DH parameters to obtain preliminary DH parameter estimates;

[0017] S603, using the Levenberg-Marquardt algorithm to introduce a flexible adaptive control factor, and performing adaptive parameter control on the preliminary DH parameters to obtain updated corresponding DH parameters.

[0018] As an optional embodiment of the present invention, optionally, the expression of the loss function in step S601 is:

[0019]

[0020] Y i =f i (w)+ΔP

[0021] ΔP=(D1,D2,D3) T

[0022] Among them, argmin() means solving the minimum value of the objective function, RSE means the loss value, n means the number of measurement points, i means the i-th measurement point, |||| 2 represents the square of the two-norm, Y i represents the actual measurement position corresponding to the i-th measurement point in the robot end position data set, f i() represents the theoretical position corresponding to the i-th measurement point in the robot end position data set, w represents the DH parameter, ΔP represents the vector of the error components in the x, y, and z directions displayed by the three micrometers of the robot end in the base coordinate system, D1 represents the error value measured in the x-axis direction, D2 represents the error value measured in the y-axis direction, D3 represents the error value measured in the z-axis direction, and T represents transpose.

[0023] As an optional embodiment of the present invention, optionally, in step S603, the expression for introducing the elastic adaptive control factor using the Levenberg-Marquardt algorithm is:

[0024]

[0025] Among them, argmin() means solving the minimum value of the objective function, RSE means the loss value, n means the number of measurement points, i means the i-th measurement point, |||| 2 represents the square of the two-norm, Y i represents the actual measurement position corresponding to the i-th measurement point in the robot end position data set, f i () represents the theoretical position corresponding to the i-th measurement point in the robot end position data set, w represents the DH parameter, represents f i () The derivative of w, w t+1 Indicates the w parameter of the next iteration in the update calculation process, w t represents the w parameter at the current moment in the update calculation process, λ represents the regularization term control parameter, and β represents the elastic adaptive control factor.

[0026] As an optional embodiment of the present invention, optionally, the expression for local update in step S602 is:

[0027]

[0028] Among them, w t+1 Indicates the w parameter of the next iteration in the update calculation process, w t Indicates the w parameter at the current moment in the update calculation process, n indicates the number of measurement points, i indicates the i-th measurement point, represents the gradient obtained at the nominal position of the i-th measurement point, express The transpose of i represents the actual measurement position corresponding to the i-th measurement point in the robot end position data set, f i () represents the theoretical position corresponding to the i-th measurement point in the robot end position data set, and w represents the DH parameter.

[0029] As an optional embodiment of the present invention, optionally, the updated expression corresponding to the DH parameter obtained in step S603 is:

[0030]

[0031] Among them, w t+1 Indicates the w parameter of the next iteration in the update calculation process, w t Indicates the w parameter at the current moment in the update calculation process, n indicates the number of measurement points, i indicates the i-th measurement point, represents the gradient obtained at the nominal position of the i-th measurement point, Indicates ▽ w f i The transpose of , λ represents the restriction effect of the regularization term on the model, β represents the elastic adaptive control factor, I represents the unit vector, and Y i represents the actual measurement position corresponding to the i-th measurement point in the robot end position data set, f i () represents the theoretical position corresponding to the i-th measurement point in the robot end position data set, and w represents the DH parameter.

[0032] As an optional embodiment of the present invention, optionally, judging whether the current DH parameter update meets the update condition in step S4 includes:

[0033] S401. Determine whether the error between the current calibration accuracy and the previous calibration accuracy is less than the error threshold. If the error is greater than the error threshold, the DH parameter meets the update condition; if the error is less than the error threshold, the DH parameter does not meet the update condition and the training ends.

[0034] As an optional embodiment of the present invention, optionally, updating the DH parameters includes:

[0035] S4011. Each time the DH parameter is updated, a column vector is added to the DH parameter, and the error is recalculated. If the error is less than the error threshold, the DH parameter is updated.

[0036] S4012, if the error is greater than the error threshold, repeat step S4011;

[0037] S4013: If the error is always greater than the error threshold, then when the number of training iterations reaches an upper limit, the updating of the DH parameters is stopped.

[0038] On the other hand, the present invention further provides an industrial robot calibration device based on a micrometer, the device is based on an industrial robot calibration method based on a micrometer; the device also includes:

[0039] A data preprocessing module, used to collect the position data of the end of the robot arm using a micrometer, and preprocess the position data of the end of the robot arm;

[0040] A data cache module, connected to the data preprocessing module, for storing the position data of the end of the robot arm;

[0041] A parameter control module, connected to the data cache module, for determining whether the DH parameters meet the update conditions;

[0042] An adaptive parameter control module, connected to the parameter control module, for training the robot calibration model according to the end position data of the robotic arm and the DH parameters;

[0043] A calibration data generation module, connected to the parameter control module and the adaptive parameter control module, for generating final DH parameters;

[0044] The data output module is connected to the calibration data generation module and is used to output the final DH parameters.

[0045] As another optional embodiment of the present invention, optionally, the parameter control module includes:

[0046] an initialization unit, used to initialize the training parameters of the device;

[0047] The parameter control unit is used to determine whether the DH parameters meet the update conditions.

[0048] The beneficial effect of the present invention is that the present invention significantly improves the accuracy and efficiency of robot calibration by introducing elastic balance factors and adaptive parameter control. Specifically, by combining the least squares algorithm and the Levenberg-Marquardt algorithm, not only can the local optimal solution be quickly approached, but also the introduction of the elastic factor enables the model to have better generalization ability and adaptability when facing different measurement data. In addition, the addition of the regularization term further enhances the stability and robustness of the model, and effectively avoids the occurrence of overfitting. Therefore, the robot calibration method of the present invention not only reduces the calibration cost, but also shortens the calibration time, providing an effective technical means for the rapid and efficient calibration of industrial robots.

[0049] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0051] Figure 1 It is a flow chart of the industrial robot calibration method based on the micrometer in Example 1 of the present invention.

[0052] Figure 2 It is a flow chart of the calibration of the DH parameters of the robot arm in Example 1 of the present invention.

[0053] Figure 3 It is a schematic diagram of the process of calculating the error Di between the nominal position and the actual position of the end of the robot arm through the robot arm DH parameters and the joint angle in Example 1 of the present invention.

[0054] Figure 4 It is a comparison chart of the model convergence execution time before and after the application of the present invention in Example 1.

[0055] Figure 5 It is a comparison chart of RMSE in the data analysis process before and after the application of the present invention in Example 1. RMSE is a measure of prediction error. The smaller the RMSE, the higher the accuracy.

[0056] Figure 6 It is a structural schematic diagram of an industrial robot calibration device based on a micrometer according to Example 2 of the present invention. DETAILED DESCRIPTION

[0057] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0058] Example 1

[0059] like Figure 1 and 2 As shown, a calibration method for an industrial robot based on a micrometer, the method comprising:

[0060] S1. Using a micrometer to collect the position data of the end of the robot arm, and preprocessing the position data of the end of the robot arm;

[0061] It should be noted that in step S1, the end position data is obtained from the robot, and the collected data is processed into data that can be directly used and stored; in the collected data, the end position coordinates of the robot arm are marked as x, y, and z, and the joint angles of the six-axis robot are recorded as q1, q2, q3, q4, q5, and q6 respectively. The error values ​​measured in the three directions of x, y, and z on the base coordinate system displayed by the three micrometers are recorded as D1, D2, and D3, and a robot arm end position data set is established based on this, and the data set is stored in the data storage module. The Levenberg-Marquardt algorithm is used to identify the robot parameter error, and the DH parameters and parameter characteristics of the robot after calibration are obtained respectively. Among them, the DH parameter w includes the connecting rod length a i , connecting rod angle a i , connecting rod offset d i , joint angle θ i , and store it at the same time.

[0062] S2. Calculate the position error of the end of the manipulator based on the end position data of the manipulator, and determine the error threshold and the upper limit of the number of training iterations;

[0063] It should be noted that in step S2, the calculation of the error is based on the difference between the end position data of the robot arm and the theoretical model. The error threshold is set to ensure the accuracy and efficiency of the calibration process and avoid endless iterations. The upper limit of the number of training iterations is set to prevent overfitting and over-calculation and ensure that the calibration process is completed within a reasonable time. The theoretical model means that in actual applications, the robot operator sets a reasonable error threshold based on the difference between the end position data of the robot arm and the theoretical model. This threshold is usually determined based on the actual application scenario and accuracy requirements to ensure that the calibration process is both accurate and efficient. The upper limit of the number of training iterations is set to prevent overfitting and over-calculation and ensure that the calibration process is completed within a reasonable time.

[0064] S3, initialize the robot arm DH parameters and set the iteration round number control variable;

[0065] It should be noted that in step S3, the initialization process includes setting the initial DH parameters of the robot arm, which are based on the theoretical values ​​provided when the robot is manufactured. After initialization, the iteration round number control variable is set to zero to start the iteration process. The iteration round number control variable is set to 0.

[0066] S4. Determine whether the current DH parameter update meets the update condition;

[0067] It should be noted that in step S4, whether the DH parameter update meets the update condition is determined by comparing the current iteration number with the set upper limit of the training iteration number. If the current iteration number has not reached the upper limit, the parameter update continues; if it has reached the upper limit, the iteration process is stopped. The setting of the update condition is to ensure that after reaching a certain number of iterations, if the theoretical model converges, the iteration is stopped to avoid unnecessary waste of computing resources.

[0068] S5, judging whether the iteration round control variable is greater than the upper limit of the training iteration round;

[0069] It should be noted that in step S5, if the iteration control variable is greater than the upper limit of the training iteration number, it means that a sufficient number of iterations have been performed, and the iteration process should be stopped at this time. If the iteration control variable does not reach the upper limit, proceed to the next step.

[0070] S6. If the DH parameters meet the update conditions and the iteration round control variable does not exceed the upper limit of the training iteration rounds, a robot calibration model consisting of the robot DH parameters and joint angles is constructed;

[0071] It should be noted that in step S6, the robot calibration model is constructed by combining the collected robot end position data with the theoretical model, and optimizing the model parameters using the least squares algorithm and the Levenberg-Marquardt algorithm. In the optimization process, by introducing elastic factors and regularization terms, it is ensured that the model has better generalization ability and adaptability when facing different measurement data, while avoiding the occurrence of overfitting. The optimized model can more accurately reflect the actual position of the end of the robot arm, thereby improving the accuracy of calibration.

[0072] S7. Calibrate the DH parameters of the robot arm using the Levenberg-Marquardt algorithm based on the robot calibration model.

[0073] It should be noted that in step S7, the calibration of the DH parameters of the manipulator using the Levenberg-Marquardt algorithm is achieved through an iterative process. In each iteration, the algorithm calculates the predicted value of the end position of the manipulator based on the current DH parameters and joint angles, and compares it with the actual position data collected. By minimizing the error between the predicted position and the actual position, the algorithm gradually adjusts the DH parameters until the predetermined convergence standard is reached. During the calibration process, the algorithm considers the dynamic characteristics of the manipulator, such as joint friction, elastic deformation and other factors, to ensure the accuracy and reliability of the calibration results. The error between the end position of the manipulator after calibration and the theoretical model is calculated, and it is determined whether the predetermined accuracy requirements are met; the error between the end position of the manipulator after calibration and the theoretical model is calculated to verify whether the calibration process is successful. By comparing the difference between the end position of the manipulator after calibration and the theoretical model, it can be determined whether the calibration has reached the predetermined accuracy requirements.

[0074] like Figure 1 , 2 As shown in Figure 3, in summary, the data collected by the micrometer is preprocessed, and then the robot parameter error is identified and calibrated by the Levenberg-Marquardt algorithm. During the calibration process, the algorithm takes into account the dynamic characteristics of the robot arm, such as joint friction, elastic deformation and other factors, to ensure the accuracy and reliability of the calibration results. The error between the end position of the robot arm after calibration and the theoretical model is calculated, and it is determined whether the predetermined accuracy requirements are met; the error between the end position of the robot arm after calibration and the theoretical model is calculated to verify whether the calibration process is successful. By comparing the difference between the end position of the robot arm after calibration and the theoretical model, it can be determined whether the calibration has achieved the expected accuracy.

[0075] As an optional embodiment of the present invention, optionally, constructing a robot calibration model consisting of robot DH parameters and joint angles in step S6 includes:

[0076] S601, constructing a loss function based on the end position data of the robotic arm;

[0077] It should be noted that in step S601, the construction of the loss function is based on the difference between the robot end position data and the theoretical model. The purpose of the loss function is to quantify the error between the model prediction value and the actual observation value, and provide a basis for subsequent parameter optimization. When constructing the loss function, the sum of squares of the errors is considered to ensure that data points with larger errors have a greater impact on the overall optimization process.

[0078] S602, using the least squares method and the opposite direction of the objective function gradient as the learning direction, locally updating the DH parameters to obtain preliminary DH parameter estimates;

[0079] It should be noted that, in step S602, the least square method is a mathematical optimization technique that seeks the best function match for the data by minimizing the sum of squares of the errors. In this embodiment, the least square method is used to process the end position data of the manipulator to construct a loss function. The construction of the loss function is based on the difference between the actual measurement data and the theoretical model, and its purpose is to find a set of DH parameters that minimizes the error between the predicted value of the end position of the manipulator and the actual measurement value.

[0080] S603, using the Levenberg-Marquardt algorithm to introduce a flexible adaptive control factor, and performing adaptive parameter control on the preliminary DH parameters to obtain updated corresponding DH parameters.

[0081] It should be noted that in step S603, the Levenberg-Marquardt algorithm is an optimization algorithm that combines the advantages of the gradient descent method and the Gauss-Newton method. It is particularly suitable for solving nonlinear least squares problems, and performs well in parameter estimation and data fitting. In the present invention, the Levenberg-Marquardt algorithm is used to perform adaptive parameter control on preliminary DH parameters. By introducing an elastic adaptive control factor, the algorithm can dynamically adjust the step size of parameter update according to the performance of the current iteration. The introduction of this adaptive control factor enables the algorithm to better balance the convergence speed and stability during the iteration process, thereby improving the efficiency and accuracy of the calibration process.

[0082] As an optional embodiment of the present invention, optionally, the expression of the loss function in step S601 is:

[0083]

[0084] Y i =f i (w)+ΔP

[0085] ΔP=(D1,D2,D3) T

[0086] Among them, argmin() means solving the minimum value of the objective function;

[0087] RSE represents the loss value;

[0088] n represents the number of measurement points;

[0089] i represents the i-th measurement point;

[0090] |||| 2 represents the square of the two-norm;

[0091] Yi represents the actual measurement position corresponding to the i-th measurement point in the robot end position data set;

[0092] f i () represents the theoretical position corresponding to the i-th measurement point in the robot end position data set;

[0093] w represents DH parameter;

[0094] ΔP represents the vector of the error components in the x, y, and z directions displayed by the three micrometers at the end of the robot arm in the base coordinate system;

[0095] D1 represents the error value measured in the x-axis direction;

[0096] D2 represents the error value measured in the y-axis direction;

[0097] D3 represents the error value measured in the z-axis direction;

[0098] T stands for transpose.

[0099] As an optional embodiment of the present invention, optionally, in step S603, the expression for introducing the elastic adaptive control factor using the Levenberg-Marquardt algorithm is:

[0100]

[0101] Among them, argmin() means solving the minimum value of the objective function;

[0102] RSE represents the loss value;

[0103] n represents the number of measurement points;

[0104] i represents the i-th measurement point;

[0105] |||| 2 represents the square of the two-norm;

[0106] Y i represents the actual measurement position corresponding to the i-th measurement point in the robot end position data set;

[0107] f i () represents the theoretical position corresponding to the i-th measurement point in the robot end position data set;

[0108] w represents DH parameter;

[0109] represents f i () The derivative of w;

[0110] w t+1 Indicates the w parameter for the next iteration in the update calculation process;

[0111] w t Indicates the w parameter at the current moment in the update calculation process;

[0112] λ represents the regularization term control parameter;

[0113] β represents the elastic adaptive control factor.

[0114] It should be noted that in order to enhance the generalization performance of the model and improve the calibration accuracy, a regularization term is usually added to the RSE and the objective function is restricted at the same time, and the elasticity factor is used to adaptively control the parameters of the objective function; in order to adaptively control the restriction effect of the regularization coefficient on the model, the calibration accuracy of the model for unknown data is maximized.

[0115] As an optional embodiment of the present invention, optionally, the expression for local update in step S602 is:

[0116]

[0117] Among them, w t+1 Indicates the w parameter for the next iteration in the update calculation process;

[0118] w t Indicates the w parameter at the current moment in the update calculation process;

[0119] n represents the number of measurement points;

[0120] i represents the i-th measurement point;

[0121] represents the gradient obtained at the nominal position of the i-th measurement point;

[0122] express The transpose of

[0123] Y i represents the actual measurement position corresponding to the i-th measurement point in the robot end position data set;

[0124] f i () represents the theoretical position corresponding to the i-th measurement point in the robot end position data set;

[0125] w represents the DH parameter.

[0126] It should be noted that, through the least squares algorithm, during a round of training, for the parameter w, the opposite direction of the gradient is used as the learning direction, and it is locally updated, so as to approach the local optimal value at the fastest speed.

[0127] As an optional embodiment of the present invention, optionally, the updated expression corresponding to the DH parameter obtained in step S603 is:

[0128]

[0129] Among them, w t+1 Indicates the w parameter for the next iteration in the update calculation process;

[0130] w t Indicates the w parameter at the current moment in the update calculation process;

[0131] n represents the number of measurement points;

[0132] i represents the i-th measurement point;

[0133] represents the gradient obtained at the nominal position of the i-th measurement point;

[0134] express The transpose of

[0135] λ represents the measure of the restrictive effect of the regularization term on the model;

[0136] β represents the elastic adaptive control factor;

[0137] I represents a unit vector;

[0138] Y i represents the actual measurement position corresponding to the i-th measurement point in the robot end position data set;

[0139] f i () represents the theoretical position corresponding to the i-th measurement point in the robot end position data set;

[0140] w represents the DH parameter.

[0141] It should be noted that, since the control parameters need to be adapted, an elasticity factor is added to this expression, and the corresponding DH parameters are solved by the Levenberg-Marquardt algorithm. This is to adaptively control the restrictive effect of the regularization coefficient on the model, so that the calibration accuracy of the model for unknown data is maximized.

[0142] As an optional embodiment of the present invention, optionally, judging whether the current DH parameter update meets the update condition in step S4 includes:

[0143] S401. Determine whether the error between the current calibration accuracy and the previous calibration accuracy is less than the error threshold. If the error is greater than the error threshold, the DH parameter meets the update condition; if the error is less than the error threshold, the DH parameter does not meet the update condition and the training ends.

[0144] As an optional embodiment of the present invention, optionally, updating the DH parameters includes:

[0145] S4011. Each time the DH parameter is updated, a column vector is added to the DH parameter, and the error is recalculated. If the error is less than the error threshold, the DH parameter is updated.

[0146] It should be noted that in step S4011, the error threshold can be a pre-set fixed value, or can be dynamically adjusted according to the accuracy requirements in the actual calibration process. For example, the error threshold can be set to 0.01 mm, which means that when the error of the calibration accuracy is less than 0.01 mm, the current DH parameters are considered to be sufficiently accurate and do not need to be further updated.

[0147] In practical applications, in order to improve the efficiency and accuracy of calibration, a multi-level iteration method can be used. Specifically, the entire calibration process can be divided into multiple stages, and different error thresholds are set for each stage. In the initial stage, the error threshold can be set relatively large to quickly narrow the error range; as the number of iterations increases, the error threshold is gradually reduced to improve the calibration accuracy.

[0148] In addition, in order to ensure the stability and reliability of the calibration process, some constraints can be introduced. For example, the maximum number of iterations for DH parameter updates can be set to prevent infinite loops caused by certain abnormal situations. At the same time, the update range of DH parameters can also be limited to avoid system instability caused by excessive update ranges.

[0149] S4012, if the error is greater than the error threshold, repeat step S4011;

[0150] S4013: If the error is always greater than the error threshold, then when the number of training iterations reaches an upper limit, the updating of the DH parameters is stopped.

[0151] It should be noted that in step S4013, the upper limit of the number of training iterations is set to avoid infinite loops and ensure that the calibration process is completed within a reasonable time. When the upper limit of the iteration is reached, if the error is still greater than the threshold, the entire calibration system may need to be checked and adjusted to ensure the accuracy and reliability of the calibration process.

[0152] like Figure 4 As shown, Figure 4 The following is a comparison chart of the data analysis time of the calibration model before and after applying the method of this embodiment. Figure 4 It can be concluded that after applying the calibration device and method of the present invention, the execution time of the model for the end position data of the robot arm is much shorter than that when the method of this embodiment is not applied. Figure 4 It can be seen that after using the calibration device and method of the present invention, the execution time of the model without the calibration device and method of the present invention is about 3 times that of the model using the calibration device and method of the present invention. That is, after using the calibration device and method of the present invention, the execution time of the model is improved by 3 times, which greatly improves the operation efficiency of the model.

[0153] like Figure 5 As shown, Figure 5 This is a comparison chart of RMSE in the data analysis process before and after the application of the method of this embodiment. RMSE is a measure of calibration error. The smaller the RMSE, the higher the accuracy. Figure 5 It can be concluded that the application of the method of this embodiment greatly improves the accuracy of the model in the robot arm calibration. In practical applications, it can better provide the industrial field with a safe, reliable, practically guaranteed and personalized calibration service.

[0154] Through the above calibration methods and processes, the calibration accuracy of the robot can be improved, which is of great significance to the promotion of industrialization.

[0155] Example 2

[0156] like Figure 6 As shown, an industrial robot calibration device based on a micrometer, the device is based on an industrial robot calibration method based on a micrometer; the device also includes:

[0157] A data preprocessing module, used to collect the position data of the end of the robot arm using a micrometer, and preprocess the position data of the end of the robot arm;

[0158] A data cache module, connected to the data preprocessing module, for storing the position data of the end of the robot arm;

[0159] A parameter control module, connected to the data cache module, for determining whether the DH parameters meet the update conditions;

[0160] An adaptive parameter control module, connected to the parameter control module, for training the robot calibration model according to the end position data of the robotic arm and the DH parameters;

[0161] A calibration data generation module, connected to the parameter control module and the adaptive parameter control module, for generating final DH parameters;

[0162] The data output module is connected to the calibration data generation module and is used to output the final DH parameters.

[0163] The parameter control module includes:

[0164] an initialization unit, used to initialize the training parameters of the device;

[0165] The parameter control unit is used to determine whether the DH parameters meet the update conditions.

[0166] like Figure 6 As shown above, in summary, the data preprocessing module obtains the end position data from the robot, processes the collected data into data that can be directly used and stores it in the data storage module; the data storage module is used to store the preprocessed input data, parameter feature matrix, control model parameters, etc.; the parameter control module is used to determine whether the decision parameters meet the construction and update conditions of the industrial robot calibration model based on the micrometer; the adaptive parameter control module constructs the industrial robot calibration model based on the micrometer according to the end position data of the robot arm and the model control parameters and notifies the calibration data generation module to perform data training on the corresponding model; the calibration data generation module calibrates the robot DH parameters and stores the generated parameter data in the data storage module; the data output module outputs the calibrated DH parameters and parameter characteristics.

[0167] Furthermore, the parameter control module includes an initialization unit and a parameter control unit. The initialization unit is used to initialize relevant parameters involved in the calibration process of the industrial robot based on the micrometer; the parameter control unit combines the initialization data to determine whether the decision parameters meet the construction and update conditions of the industrial robot calibration model based on the micrometer.

[0168] In the data preprocessing module, the end position coordinates of the robot are marked as x, y, and z in the collected data. The joint angles of the six-axis robot are recorded as q1, q2, q3, q4, q5, and q6 respectively. The error values ​​measured in the three directions of x, y, and z on the base coordinate system shown by the three micrometers are recorded as D1, D2, and D3. A robot end position data set is established based on this, and the data set is stored in the data storage module. The Levenberg-Marquardt algorithm is used to identify the robot parameter error, and the DH parameters and parameter characteristics of the robot after calibration are obtained respectively. Among them, the DH parameter w includes the connecting rod length a i , connecting rod angle a i , connecting rod offset d i , joint angle θ i , and store it in the data storage module at the same time.

[0169] The initialization unit is configured to initialize the robot arm's thousandths and DH parameter w; initialize the calibration model adaptive control factor β; initialize the regularization factor λ; initialize the maximum number of iterations T; initialize the iteration control variable a during training; initialize the convergence termination threshold s; the maximum training iteration T is a variable that controls the upper limit of the number of training times and is initialized to a large positive integer; the iteration control variable a is initialized to 0; the convergence termination threshold s is a parameter for determining whether the training is finished and is initialized with a very small positive number. The regularization factor λ is a measure of the restrictive effect of the regularization term on the model and is initialized to a small positive number. The calibration model adaptive control factor β is initialized to a small positive number in order to adaptively control the restrictive effect of the regularization coefficient on the model so that the calibration accuracy of the model for unknown data is maximized.

[0170] The parameter control unit is configured to obtain control parameters of the industrial robot calibration model based on the dial gauge from the storage module, so as to determine whether the parameters meet the construction and update conditions of the industrial robot calibration model based on the dial gauge.

[0171] According to a preferred solution of the industrial robot calibration based on a dial gauge of the present invention, a calibration model of the industrial robot of a dial gauge is constructed in combination with initialization related parameters, specifically:

[0172] According to the robot end position data set, the objective function RSE is constructed, and the formula is as follows:

[0173]

[0174] Y i =f i (w)+ΔP

[0175] ΔP=(D1,D2,D3) T

[0176] In order to enhance the generalization performance of the model and improve the calibration accuracy, a regularization term is usually added to the RSE and the objective function is restricted. The elasticity factor is used to adaptively control the parameters of the objective function to obtain:

[0177]

[0178] In order to adaptively control the restrictive effect of the regularization coefficient on the model, the calibration accuracy of the model for unknown data can be maximized.

[0179] Combined with the calibration model constructed by the adaptive parameter control module, the calibration data is generated as follows:

[0180] After calling the adaptive elastic factor parameter control module, the calibration data of the robot arm is generated by the industrial robot calibration prediction model based on the micrometer, and the DH parameter w is trained on the cumulative error RSE using the least squares method. That is, for each known actual end position data Y of the robot arm i In the first step, the least squares algorithm is used to make the parameter w in a round of training process, using the opposite direction of the RSE gradient as the learning direction, and locally updating it, so as to approach the local optimal value at the fastest speed. The update formula is:

[0181]

[0182] In the second step, since the control parameters need to be adaptive, an elastic factor is added in the second step, and the corresponding DH parameters are solved by the Levenberg-Marquardt algorithm. The update formula is:

[0183]

[0184] where w t Represents the current w parameter value, which is solved by the Levenberg-Marquardt algorithm. In order to adaptively control the restrictive effect of the regularization coefficient on the model, the model achieves the highest calibration accuracy for unknown data.

[0185] The output module includes an output parameter characteristic unit and a calibrated DH parameter unit.

[0186] Specifically, according to the end position data and joint angle values ​​of the robot arm, the kinematic model of the robot arm is called to calculate the nominal end position coordinates of the robot arm, and the errors between the theoretical position and the actual measured position are calculated respectively. The calibrated DH parameter data are obtained by weighted accumulation according to the elastic balance factor.

[0187] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

Claims

1. A calibration method for an industrial robot based on a micrometer, characterized in that: The method comprises: S1. Using a micrometer to collect the position data of the end of the robot arm, and preprocessing the position data of the end of the robot arm; S2. Calculate the position error of the end of the manipulator based on the end position data of the manipulator, and determine the error threshold and the upper limit of the number of training iterations; S3, initialize the robot arm DH parameters and set the iteration round number control variable; S4. Determine whether the current DH parameter update meets the update condition; S5, judging whether the iteration round control variable is greater than the upper limit of the training iteration round; S6. If the DH parameters meet the update conditions and the iteration round control variable does not exceed the upper limit of the training iteration rounds, a robot calibration model consisting of the robot DH parameters and joint angles is constructed; S7. Calibrate the DH parameters of the robot arm using the Levenberg-Marquardt algorithm based on the robot calibration model.

2. The industrial robot calibration method based on a micrometer as claimed in claim 1, characterized in that: Constructing a robot calibration model consisting of robot DH parameters and joint angles in step S6 includes: S601, constructing a loss function based on the position data of the end of the robotic arm; S602, using the least squares method and the opposite direction of the objective function gradient as the learning direction, locally updating the DH parameters to obtain preliminary DH parameter estimates; S603, using the Levenberg-Marquardt algorithm to introduce a flexible adaptive control factor, and performing adaptive parameter control on the preliminary DH parameters to obtain updated corresponding DH parameters.

3. The industrial robot calibration method based on a micrometer as claimed in claim 2, characterized in that: In step S601, the loss function is expressed as: Y i =f i (w)+ΔP ΔP=(D1,D2,D3) T Among them, argmin() represents the minimum value of the objective function, RSE represents the loss value, n represents the number of measurement points, i represents the i-th measurement point, || || 2 represents the square of the two-norm, Y i represents the actual measurement position corresponding to the i-th measurement point in the robot end position data set, f i ( ) represents the theoretical position corresponding to the i-th measurement point in the robot end position data set, w represents the DH parameter, ΔP represents the vector of the error components in the x, y, and z directions displayed by the three micrometers of the robot end in the base coordinate system, D1 represents the error value measured in the x-axis direction, D2 represents the error value measured in the y-axis direction, D3 represents the error value measured in the z-axis direction, and T represents transposition.

4. The industrial robot calibration method based on a micrometer as claimed in claim 2, characterized in that: In step S603, the expression of the elastic adaptive control factor introduced by using the Levenberg-Marquardt algorithm is: Among them, argmin() represents the minimum value of the objective function, RSE represents the loss value, n represents the number of measurement points, i represents the i-th measurement point, || || 2 represents the square of the two-norm, Y i represents the actual measurement position corresponding to the i-th measurement point in the robot end position data set, f i ( ) represents the theoretical position corresponding to the i-th measurement point in the robot end position data set, w represents the DH parameter, represents f i ( ) The derivative of w, w t+1 Indicates the w parameter of the next iteration in the update calculation process, w t represents the w parameter at the current moment in the update calculation process, λ represents the regularization term control parameter, and β represents the elastic adaptive control factor.

5. The industrial robot calibration method based on a micrometer as claimed in claim 2, characterized in that: The expression for local update in step S602 is: Among them, w t+1 Indicates the w parameter of the next iteration in the update calculation process, w t Indicates the w parameter at the current moment in the update calculation process, n indicates the number of measurement points, i indicates the i-th measurement point, represents the gradient obtained at the nominal position of the i-th measurement point, express The transpose of i represents the actual measurement position corresponding to the i-th measurement point in the robot end position data set, f i ( ) represents the theoretical position corresponding to the i-th measurement point in the robot end position data set, and w represents the DH parameter.

6. The industrial robot calibration method based on a micrometer as claimed in claim 2, characterized in that: The updated expression corresponding to the DH parameter obtained in step S603 is: Among them, w t+1 Indicates the w parameter of the next iteration in the update calculation process, w t Indicates the w parameter at the current moment in the update calculation process, n indicates the number of measurement points, i indicates the i-th measurement point, represents the gradient obtained at the nominal position of the i-th measurement point, (▽ w f i ) T Indicates ▽ w f i The transpose of , λ represents the restriction effect of the regularization term on the model, β represents the elastic adaptive control factor, I represents the unit vector, and Y i represents the actual measurement position corresponding to the i-th measurement point in the robot end position data set, f i ( ) represents the theoretical position corresponding to the i-th measurement point in the robot end position data set, and w represents the DH parameter.

7. The industrial robot calibration method based on a micrometer as claimed in claim 1, characterized in that: In step S4, judging whether the current DH parameter update meets the update condition includes: S401. Determine whether the error between the current calibration accuracy and the previous calibration accuracy is less than the error threshold. If the error is greater than the error threshold, the DH parameter meets the update condition; if the error is less than the error threshold, the DH parameter does not meet the update condition and the training ends.

8. The industrial robot calibration method based on a micrometer as claimed in claim 7, characterized in that: Updating the DH parameters includes: S4011. Each time the DH parameter is updated, a column vector is added to the DH parameter, and the error is recalculated. If the error is less than the error threshold, the DH parameter is updated. S4012, if the error is greater than the error threshold, repeat step S4011; S4013: If the error is always greater than the error threshold, then when the number of training iterations reaches an upper limit, the updating of the DH parameters is stopped.

9. An industrial robot calibration device based on a micrometer, characterized in that: The device comprises the industrial robot calibration method based on a micrometer as claimed in any one of claims 1 to 8; the device also comprises: A data preprocessing module, used to collect the position data of the end of the robot arm using a micrometer, and preprocess the position data of the end of the robot arm; A data cache module, connected to the data preprocessing module, for storing the position data of the end of the robot arm; A parameter control module, connected to the data cache module, for determining whether the DH parameters meet the update conditions; An adaptive parameter control module, connected to the parameter control module, for training the robot calibration model according to the end position data of the robotic arm and the DH parameters; A calibration data generation module, connected to the parameter control module and the adaptive parameter control module, for generating final DH parameters; The data output module is connected to the calibration data generation module and is used to output the final DH parameters.

10. The industrial robot calibration device based on a micrometer as claimed in claim 9, characterized in that: The parameter control module comprises: an initialization unit, used to initialize the training parameters of the device; The parameter control unit is used to determine whether the DH parameters meet the update conditions.

Images