A combined robot contact calibration method and system for solving global optimality
Through the combined iterative calculation of the L-M-D algorithm, the problem of global optimal solution in robot contact calibration is solved, the calibration accuracy and calculation efficiency are improved, and the high accuracy and stability of the processing system are ensured.
Patent Information
- Application Number
- CN202410873945.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-02
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2044-07-02
AI Technical Summary
The existing robot contact calibration technology is difficult to achieve global optimal solution, resulting in low calibration accuracy and low computational efficiency. Especially in the optimization of nonlinear least squares method, it is easy to fall into local optimal situations, affecting processing quality.
The combined iteration algorithm of the L-M method and the DIRECT method is adopted to establish the objective function and perform loop calculations to obtain the coordinate system origin position parameters of the ruby probe to ensure the acquisition of the global optimal solution, combining the local search efficiency of the L-M method and the DIRECT method Global convergence, the L-M-D algorithm is designed to improve calibration accuracy and computational efficiency.
It realizes high-precision calibration of the coordinate system of the robot processing system, avoids the trap of local optimal solutions, improves calibration accuracy and calculation efficiency, can quickly converge from any initial position, and has good robustness and global optimization capabilities.
Smart Images

Figure CN118404589B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of robot contact calibration, and in particular to a combined robot contact calibration method and system for solving global optimality. Background Art
[0002] Panel parts are widely used in aerospace structures and are usually used in various load-bearing applications. These parts are usually processed using CNC machine tools. In order to improve the surface quality and extend the fatigue life, it is necessary to eliminate the tool marks on the surface of the parts. In addition, a certain degree of residual burrs and bosses are often left on the surface of the parts, and the manual grinding required to remove such burrs is a time-consuming process.
[0003] The robot can clamp a variety of equipment (measuring tools, processing tools), and can also communicate with external calibration and measurement equipment. The automated matching processing of structural parts is completed through the integrated operation of calibration, measurement, and processing. In order to efficiently and high-quality complete the removal of boss residues and processing marks of wall panel parts, the robot work unit can be used to complete the above functions because it has the composite processing capabilities of milling and grinding. Due to the variety of structural parts, the cost of designing special clamping tooling is very high, and the fixture needs to be replaced frequently, resulting in low efficiency and low automation level of the solution using special tooling. Therefore, a robot grinding and polishing flexible tooling is designed. The tooling adopts a combination of vacuum adsorption and magnetic fixation, which can realize the arbitrary placement of workpieces on the workbench.
[0004] Since the workpiece can be placed at any position on the workbench and the position of the workpiece is uncertain, the workpiece needs to be calibrated and positioned before processing to determine the workpiece coordinate system before the offline program can be used for processing. To improve the robot processing accuracy, the focus is on improving the calibration accuracy of the robot processing system coordinate system. The robot processing system coordinate system calibration is mainly the calibration of the tool coordinate system and the workpiece coordinate system, and its calibration accuracy is of great significance to subsequent processing operations.
[0005] At present, in addition to visual measurement, the robot machining system coordinate system can also be calibrated through contact ruby probe calibration. The flexible fixed probe holder is calibrated with the "ball-to-ball" method instead of the traditional point-to-point method; then, the calibrated probe is used to accurately find the tool origin and accurately calibrate the tool.
[0006] The patent with publication number CN109848989A discloses a method for automatic calibration and detection of the robot execution end based on a ruby probe. The fixed ruby probe is used to touch and calibrate the surface contour of the workpiece, calculate the surface contour of the workpiece before and after processing, and transmit the calculation results to the robot controller to realize the calculation and display of the surface contour of the workpiece at the robot end. It can realize automatic calibration of the robot execution end and automatically complete the surface contour detection of the workpiece before and after processing.
[0007] The patent with publication number CN111941425A discloses a method for rapid positioning of workpieces in a robot milling system based on a laser tracker and a binocular camera. A method for rapid positioning of workpieces in a robot milling system based on a laser tracker and a binocular camera. This method can quickly convert the point cloud scanned by the binocular camera into the workpiece coordinate system, extract the feature points in the workpiece coordinate system through point cloud processing, and then guide the robot to move to the feature point position based on these feature points. This method is not only efficient, but also avoids the error of manual teaching, and improves the accuracy of workpiece positioning in the robot milling system.
[0008] Although the laser tracker and binocular camera used in the above patents can meet the requirements of automatic measurement and calibration of workpieces, the overall equipment cost is relatively high, making it difficult to achieve low-cost widespread application. In addition, the use of nonlinear least squares method to optimize TCP calibration results is prone to fall into local optimal situations, resulting in inaccurate results of parameter solutions, affecting the final calibration quality. Summary of the invention
[0009] To this end, the present invention provides a combined robot contact calibration method and system for solving the global optimal solution, which obtains the global optimal solution through a mutual iteration process, ensures the efficiency of the algorithm, and maximizes the global optimal solution capability. The calibration accuracy of the robot machining system coordinate system is of great significance to the subsequent machining operations.
[0010] In order to solve the above technical problems, the present invention provides a contact calibration method for a combined robot for solving the global optimum, comprising:
[0011] Obtain the homogeneous transformation matrix from the coordinate system of the robot wrist end to the robot base coordinate system when the robot's ruby probe contacts the standard ball;
[0012] Establishing a calibration model based on the positions of the coordinate system origin of the ruby probe and the center of the standard ball in the robot base coordinate system;
[0013] Based on the calibration model, an objective function is established according to a nonlinear least squares optimization method;
[0014] Based on the objective function, the LM method and the DIRECT method are used to perform cyclic iterative calculations to obtain the global optimal solution of the objective function, including:
[0015] Setting initial parameters of the LM method and iteratively calculating, in response to the objective function converging to a local minimum through iterative calculation of the LM method, using a local minimum point corresponding to the local minimum as an initial parameter of the DIRECT method;
[0016] In response to the function value of the objective function iteratively calculated by the DIRECT method being less than the local minimum value, updating the initial parameters of the LM method, taking the point corresponding to the function value as the new initial parameters of the LM method, and iterating the calculation again;
[0017] Repeat the iterative calculation steps until a global optimal solution of the objective function is obtained by the LM method;
[0018] Based on the global optimal solution, the coordinate system origin position parameters of the ruby probe are obtained.
[0019] In one embodiment of the present invention, when the ruby probe of the robot contacts the standard ball, the homogeneous transformation matrix from the coordinate system of the robot wrist end to the robot base coordinate system is obtained, including:
[0020] Install the standard ball on a magnetic base and attach it to a fixed bracket within the reach of the robot through a magnet;
[0021] The robot moves toward the standard ball through different positions and postures;
[0022] The robot controller obtains the homogeneous transformation matrix from the coordinate system {E} of tool0 at the end of the robot wrist to the robot base coordinate system {B} each time the ruby probe contacts the standard ball. .
[0023] In one embodiment of the present invention, a calibration model is established based on the positions of the coordinate system origin of the ruby probe and the center of the standard ball in the robot base coordinate system, including:
[0024] Taking the standard ball as the reference target, the calibration model is established as follows:
[0025]
[0026] in, is the position of the coordinate origin TCP of the ruby probe in the robot base coordinate system {B}; is the position of the center of the standard sphere in the robot base coordinate system {B}, and its value is unknown; R and r are the radius of the standard sphere and the radius of the ruby probe respectively.
[0027] In one embodiment of the present invention, the objective function is:
[0028]
[0029] .
[0030] In one embodiment of the present invention, the coordinate system origin position parameter of the ruby probe is , there is the following relationship:
[0031] ,have to:
[0032] , which is simplified to:
[0033] ;
[0034] in, is the posture matrix of the robot wrist end, is the position matrix of the robot wrist end.
[0035] In one embodiment of the present invention, based on the objective function, the LM method and the DIRECT method are used to perform cyclic iterative calculation, further comprising:
[0036] Set the iterative convergence condition, the minimum step size for iterative termination, the maximum number of iterations, and the maximum number of cycles of the LM method;
[0037] After each iteration, determine whether the iteration step is less than the minimum step;
[0038] If it is less than the minimum step size, determine whether the convergence condition is met. If the convergence condition is met, terminate the iteration and output the global optimal solution of the objective function. If the convergence condition is not met, output the local minimum value of the objective function.
[0039] If it is not less than the minimum step size, determine whether the maximum number of iterations has been reached;
[0040] If the maximum number of iterations has not been reached, the iterative calculation continues;
[0041] If the maximum number of iterations is reached, the iteration is stopped and the result is output.
[0042] In one embodiment of the present invention, when the local minimum value of the target function is output, it is determined whether the maximum number of loops is reached. If reached, the loop is terminated. If not reached, the local minimum value is used as the initial parameter of the DIRECT method.
[0043] In one embodiment of the present invention, the iteration convergence condition is that the iteration result converges to within 0.1 mm.
[0044] In one embodiment of the present invention, based on the objective function, the LM method and the DIRECT method are used to perform cyclic iterative calculation, further comprising:
[0045] Set the iteration termination condition and search range control parameters of the DIRECT method, and use the local minimum value obtained by the LM method as the initial parameters of the DIRECT method;
[0046] Determine whether the function value obtained by iterative calculation is less than the local minimum value. If it is less than the local minimum value, update the initial parameters of the LM method. If it is not less than the local minimum value, determine whether the iteration termination condition is met.
[0047] If the iteration termination condition is not met, the iterative calculation continues. If the iteration termination condition is met, it is determined whether the maximum search range has been reached. If it has been reached, the optimal solution has not been found and the result is output. If it has not been reached, the search range is expanded by changing the value of the search range control parameter and the iterative calculation is repeated.
[0048] The present invention also provides a combined robot contact calibration system for solving global optimality, comprising:
[0049] A homogeneous transformation matrix acquisition module is used to obtain a homogeneous transformation matrix from the coordinate system of the robot wrist end to the robot base coordinate system when the ruby probe of the robot contacts the standard ball;
[0050] A calibration model building module, used to build a calibration model based on the positions of the coordinate system origin of the ruby probe and the center of the standard ball in the robot base coordinate system;
[0051] An objective function establishment module, used to establish an objective function based on the calibration model and according to a nonlinear least squares optimization method;
[0052] A loop iterative calculation module, used to perform loop iterative calculation based on the objective function using the LM method and the DIRECT method to obtain a global optimal solution of the objective function;
[0053] The position parameter acquisition module is used to obtain the coordinate system origin position parameters of the ruby probe based on the global optimal solution.
[0054] The above technical solution of the present invention has the following advantages compared with the prior art:
[0055] The present invention discloses a combined robot contact calibration method and system for solving the global optimal problem. Taking into account the actual characteristics of the robot probe TCP calibration, the LM algorithm and the DIRECT algorithm are combined to obtain an optimization algorithm, namely the LMD algorithm. The algorithm has good global convergence and computational efficiency. Compared with other algorithms, the LMD algorithm combined optimization method does not depend on the initial value, has global optimization capability, and has a fast iteration speed. It can effectively solve nonlinear equation system problems such as robot calibration.
[0056] This method does not require a relatively accurate initial solution, and can start iterative calculations from any position. Even when the initial solution is not sufficiently close to the problem solution, it can converge quickly. The algorithm has good robustness and can handle ill-conditioned iterative matrix problems.
[0057] The present invention combines the advantages of two algorithms and solves the problem through a combined algorithm. The designed algorithm uses the LM method and the DIRECT algorithm to perform mutual cyclic iteration to obtain a combined algorithm for global optimization. This method takes into account the local search efficiency of LM and the global convergence of DIRECT. The entire combined algorithm consists of two stages: the local minimization stage and the local minimum jump stage. The LM method is used for local optimization, and the DIRECT algorithm is used to jump out of the local minimum. BRIEF DESCRIPTION OF THE DRAWINGS
[0058] In order to make the contents of the present invention more clearly understood, the present invention is further described in detail below based on specific embodiments of the present invention in conjunction with the accompanying drawings.
[0059] Figure 1 It is a schematic diagram of calibration between the robot contact probe and the standard ball of the present invention.
[0060] Figure 2 It is a schematic diagram of 2D segmentation of the DIRECT method of the present invention.
[0061] Figure 3 It is a flow chart of the combined calibration algorithm of the present invention. DETAILED DESCRIPTION
[0062] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it, but the embodiments are not intended to limit the present invention.
[0063] Example 1
[0064] Reference Figure 3 As shown, this embodiment provides a contact calibration method for a combined robot for solving a global optimum, comprising:
[0065] Step S1, obtaining a homogeneous transformation matrix from the coordinate system of the robot wrist end to the robot base coordinate system when the ruby probe of the robot contacts the standard ball.
[0066] First, build the calibration environment. Figure 1 As shown, the standard ball is mounted on a magnetic base and adsorbed on a fixed bracket within the reach of the robot through a magnet; the robot model used is IRB ABB 6660-1.95, the standard ball used is 38.05mm in diameter, and the ruby probe uses a Renishaw standard precision probe, model RENISHAWOMP40-2, with a nominal repeatability of 1 μm.
[0067] Run the robot probe calibration program, install the probe on the corresponding tool holder, install the spindle at the end of the robot, automatically grab the tool holder in the tool library, start the robot TCP calibration program, automatically turn on the probe, and the robot moves to the standard ball through different positions and postures through the calibration program;
[0068] The ruby probe triggers a signal after contacting the standard ball, so that the robot controller can immediately obtain the homogeneous transformation matrix from the coordinate system {E} of tool0 at the end of the robot wrist to the robot base coordinate system {B} every time the ruby probe contacts the standard ball. , thereby accurately realizing the "ball-to-ball" process, the robot returns to the Home point and starts measuring the next point until the measurement of all preset points is completed.
[0069] In this embodiment, a total of 10 sets of data are collected through the TCP calibration experiment, including the tool0 position coordinates and posture parameters of each measurement point.
[0070] Step S2: establishing a calibration model based on the positions of the coordinate system origin of the ruby probe and the center of the standard sphere in the robot base coordinate system.
[0071] Different from the commonly used robot calibration using sharp points, the present invention uses a standard sphere for calibration, so the point-to-point calibration equation is , which cannot be solved using traditional solution methods.
[0072] Taking the standard ball as the reference target, the calibration model is established as follows:
[0073] (1)
[0074] in, is the position of the coordinate origin TCP of the ruby probe in the robot base coordinate system {B}; is the position of the center of the standard sphere in the robot base coordinate system {B}, and its value is unknown; R and r are the radius of the standard sphere and the radius of the ruby probe respectively.
[0075] After the calibration process is completed, the TCP parameters of the ruby probe's coordinate system origin position It can be further expanded into:
[0076] (2)
[0077] (3)
[0078] After expanding formula (3), it can be simplified to:
[0079] (4)
[0080] in, is the posture matrix of the robot wrist end, is the position matrix of the robot wrist end.
[0081] Step S3: Based on the calibration model, an objective function is established according to a nonlinear least squares optimization method.
[0082] Theoretically, the more calibrated position points are, the higher the accuracy of the TCP points obtained after solving using the least squares method.
[0083] After transforming formula (1), the objective function is established:
[0084] (5)
[0085] Find the norm of the objective function to make it optimal. The problem of solving this function is transformed into a nonlinear least squares optimization problem:
[0086] (6)
[0087] It should be noted that in the actual calibration and calibration data summary process, the following points need to be noted:
[0088] The calibration result obtained must be globally optimal. If the algorithm converges to a local optimum or fails to converge, it may have serious consequences for the subsequent measurement process.
[0089] During the calibration process, the functions involved may be multi-dimensional non-convex continuous functions.
[0090] Controllable noise is mixed into the calibration data. When the noise level is low, the high-dimensional space containing the global optimal solution and other local optimal solutions becomes relatively concentrated.
[0091] Based on the amount of calibration data collected, the residual norm of the least squares method is used to establish the termination criterion for evaluating the convergence of the results. A larger norm indicates a deviation from the global optimal solution or a problem with the calibration data, resulting in the rejection of the final iterative result. On the contrary, a small norm indicates successful convergence of the results.
[0092] Step S4: Based on the objective function, the LM method (Levenberg–Marquard) and the DIRECT method are used to perform cyclic iterative calculations to obtain a global optimal solution of the objective function.
[0093] The overall process is to find a local solution through the LM method, and in order to leave this local solution, use some iterations of the DIRECT algorithm to find the next best point, and then use this point as the initial point to start a new search using the LM method. This process continues until convergence, and finally finds the global optimal solution.
[0094] The LM method and the DIRECT method are explained below respectively.
[0095] The LM method (Levenberg–Marquard) is used for local optimization. The LM method is a trust region algorithm. When the convergence speed is fast, the trust threshold is increased to make the algorithm tend to the Newton method; when the convergence speed is slow, the trust threshold is reduced to make the algorithm tend to the steepest descent method. Its advantages are fast speed, only using first-order matrices, and being able to obtain the optimal solution far away from the initial value.
[0096] The LM method is an improvement on the Gauss-Newton method, with a slightly different iterative step size:
[0097] (7).
[0098] In most cases, the LM method can achieve the desired iteration effect and has a fairly fast iteration speed. However, the LM method is still a local optimization method. The algorithm may converge to the local optimal solution and cannot jump out of the local optimal solution. In particular, when the noise mixed in the data is too large or the data passes through the local minimum point of the function during the iteration process, the LM method may not converge to the global optimal solution.
[0099] Use the DIRECT method for global optimization. The DIRECT method is a deterministic global optimization algorithm that has very low requirements on the objective function. It is suitable for dealing with global optimization problems with bounded constraints. However, this global convergence may come at the expense of a large-scale and exhaustive search in the field. Another obvious disadvantage is the asymptotic invalidity of the DIRECT algorithm. The DIRECT algorithm searches for the absorption domain of the local optimal solution (including the global optimal solution) with a small amount of computational cost, but requires a lot of computational cost to be very close to or reach the optimal solution. This property is also a common property of global optimization algorithms.
[0100] In addition, the DIRECT method has a strong global search capability and can quickly find some areas containing optimal solutions in the early stages of the algorithm, so that the function value can quickly drop to a value close to the minimum function value. However, in subsequent searches, as the accuracy requirements become higher and higher, the computational cost also grows rapidly, which is the asymptotic ineffectiveness of the algorithm.
[0101] Because the above algorithms are not completely applicable to the optimization problem in the present invention due to their respective shortcomings. The goal of the present invention is to balance computational efficiency and global convergence. According to the actual characteristics of robot calibration, most calibration data can converge to the global optimum using a local optimization algorithm. However, in some cases, a function has multiple local extreme points. At this time, the local optimization algorithm cannot make the function converge to the global optimum, resulting in the obtained parameters not being consistent with the actual parameters. At this time, a global optimization algorithm is needed to make it converge to the global optimum.
[0102] In order to avoid the problems of the above two algorithms, the present invention makes corresponding improvements to the algorithms.
[0103] Step S4 specifically includes:
[0104] 1. Run the LM method part.
[0105] Set the iterative convergence condition and the minimum step size for iterative termination of the LM method , maximum number of iterations k max and the maximum number of cycles K ; Set the initial parameters of the LM method and iterate the calculation. During the initialization process, since the initial parameters are unknown, the initial parameters can be set arbitrarily, such as setting the initial value to (0, 0, 0, 0, 0, 0).
[0106] After each iteration, determine the iteration step length d x Is it less than the minimum step size? ;
[0107] If the minimum step size is satisfied , then determine whether the convergence condition is met. If the convergence condition is met, terminate the iteration and output the global optimal solution of the objective function. If the convergence condition is not met, output the local minimum of the objective function. When the local minimum of the objective function is output, determine whether the maximum number of cycles has been reached. K If it is reached, the loop ends. If it is not reached, the local minimum point corresponding to the local minimum is used as the initial parameter of the DIRECT method.
[0108] If it is not less than the minimum step size, determine whether the maximum number of iterations has been reached;
[0109] If the maximum number of iterations is not reached k max , then continue iterative calculation;
[0110] If the maximum number of iterations is reached k max , then the iteration stops and the result is output. At this time, the algorithm iteration has not converged, and it is necessary to collect data again or adjust the maximum number of iterations. k max Or minimum step size Parameters and then solve. 2. Determine whether to jump to the algorithm part.
[0111] If the minimum step size criterion is met, but the result does not meet the global optimal solution threshold, it means that LM converges to a local optimal solution, and the algorithm jumps to the DIRECT method, and the local minimum point obtained by iteratively calculating the objective function through the LM method is calculated. , local minimum As initial parameters, they are transmitted to the DIRECT algorithm through the corresponding interactive interface, ready for the next step of calculation.
[0112] 3. Run the DIRECT method part.
[0113] If LM only finds the local minimum of the function , then the DIRECT method is likely to find a solution that is better than the LM method. Smaller function value. To achieve this goal, the DIRECT method is introduced and modified accordingly.
[0114] In order to combine the algorithms together, it is necessary to make appropriate modifications to the combined algorithm, including data interaction with the DIRECT method. Before the LM method transmits the data of the local optimal solution to the DIRECT algorithm, it is necessary to confirm whether the combined algorithm has reached the maximum number of loop iterations. If so, it stops iterating and outputs the result. This means that no optimal solution that meets the function iteration threshold has been found. The DIRECT algorithm also needs to determine whether it needs to transmit the initial value (i.e., the local minimum point) to the LM method. ), if the DIRECT algorithm is satisfied If the parameters of the LM method are changed, the new parameters are transmitted to the LM method through the interaction part.
[0115] This improved DIRECT method does not need to search for the global optimal solution, but only needs to iteratively find New parameters x k , that is, the iteration can be completed and the loop can be jumped out. According to the characteristics of the DIRECT algorithm, the algorithm can quickly find these areas containing the optimal solution within the constraint range in the early stage. Therefore, it can effectively avoid the asymptotic invalidity of the global optimization algorithm and quickly find a solution that is better than the local optimal solution obtained by the LM method. Then, the LM method is used to quickly iterate and obtain the optimal solution.
[0116] The core of the DIRECT algorithm is to repeat the operation of "selecting a hyperrectangle-splitting a hyperrectangle". The hyperrectangle selected for the next step of segmentation is called the potential optimal hyperrectangle. Figure 2 As shown. DIRECT selects the function value of the midpoint of the sampling interval to ensure that all subsequent samples are at the midpoint of the interval. The longest side of the region is δ, and I is the set of dimensions with the longest side. With the center point c as the center, sample in each dimension of I , is a vector whose i-th element is 1 and the other elements are 0. Then each dimension of I divides the area into three equal parts. At this time, c becomes the center of the small area in the middle, and the center points of the two small areas on both sides are When there are multiple dimensions, the dimension where the sampling point with the smallest function value is located is preferentially selected for segmentation. In the undivided dimension, this operation is repeated until all dimensions are divided into three equal parts.
[0117] Initialize DIRECT related parameters to the local minimum point As the initial point, bring in the DIRECT method, set the iteration termination conditions (including whether the optimal solution threshold is reached, the maximum number of iterations, the maximum number of superstar segmentations, etc.), set the iteration termination conditions of the DIRECT method and the search range control parameter μ, and use the local minimum value obtained by the LM method as the initial parameter of the DIRECT method.
[0118] Determine whether the function value obtained by iterative calculation is less than the local minimum value. If it is less than the local minimum value, update the initial parameters of the LM method, use the point corresponding to the function value as the new initial parameters of the LM method, iterate again, and if it is not less than the local minimum value, determine whether the iteration termination condition is met;
[0119] If the iteration termination condition is not met, the iterative calculation continues. If the iteration termination condition is met, it is determined whether the maximum search range has been reached. If it has been reached, the optimal solution has not been found and the result is output. If it has not been reached, the search range is expanded by changing the value of the search range control parameter and the iterative calculation is repeated.
[0120] In the actual calibration experiment, the gap between the local optimal solution and the global optimal solution is usually within 200mm, which can be used as the initial constraint range for iteration. Setting the search range control parameter μ can control the search range and avoid the search range being too large, which may cause more invalid calculations and waste computing resources. Starting with the initial parameter μ, first search in a small range. If a smaller local minimum value cannot be found, try changing the value of the control parameter μ. The search range can be doubled and gradually expanded.
[0121] In addition, the present embodiment sets the parameter ε of the global optimal solution threshold, and the parameter ε can ensure the global optimal result. Generally, the robot body accuracy calibrated with absolute accuracy is about 0.2mm. Based on this benchmark and combined with the previous calibration experimental data, the parameter ε of the global optimal solution threshold given at this time can be set to 0.1mm. When the iteration result converges to within 0.1mm, it is considered that the iteration converges to the global optimum and the current iteration result is recognized. The iteration convergence condition is that the iteration result converges to within 0.1mm.
[0122] Step S5: Based on the global optimal solution, obtain the coordinate system origin position parameters of the ruby probe.
[0123] After the above algorithm converges, the optimization algorithm can directly output the iterative solution result matrix, which is a 6×1 column vector, in which the first three rows are the coordinate system origin position parameters of the ruby probe to be determined, that is, The last three lines are the positions of the center of the standard ball in the robot base coordinate system {B} .
[0124] Table 1 shows the solutions of different calibration algorithms for a set of parameters measured in the experiment.
[0125] Table 1 Comparison of solution results of different algorithms
[0126]
[0127] It can be seen that the LM method has the highest computational efficiency, but global convergence cannot be guaranteed, especially when there is a lack of initial values, it is easy to fall into the local optimum. Using the DIRECT method, it can be seen that the computational efficiency is very slow. Using the DIRECT method alone requires setting the search boundary. Since the initial value of the calibration TCP is unknown, the search boundary must be set in a large space to ensure that the global optimal solution is within the search range. This further leads to high computational costs. It can be seen that the LMD algorithm of the present invention has the advantages of good global convergence and faster computational efficiency compared to other methods.
[0128] Example 2
[0129] Based on the same inventive concept, this embodiment provides a combined robot contact calibration system for solving the global optimum. The principle of solving the problem is similar to the combined robot contact calibration method for solving the global optimum, and the repeated parts will not be repeated.
[0130] This embodiment provides a combined robot contact calibration system for solving a global optimum, including:
[0131] A homogeneous transformation matrix acquisition module is used to obtain a homogeneous transformation matrix from the coordinate system of the robot wrist end to the robot base coordinate system when the ruby probe of the robot contacts the standard ball;
[0132] A calibration model building module, used to build a calibration model based on the positions of the coordinate system origin of the ruby probe and the center of the standard ball in the robot base coordinate system;
[0133] An objective function establishment module, used to establish an objective function based on the calibration model and according to a nonlinear least squares optimization method;
[0134] A loop iterative calculation module, used to perform loop iterative calculation based on the objective function using the LM method and the DIRECT method to obtain a global optimal solution of the objective function;
[0135] The position parameter acquisition module is used to obtain the coordinate system origin position parameters of the ruby probe based on the global optimal solution.
[0136] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.
[0137] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0138] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0139] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0140] Finally, it should be noted that the above specific implementation methods are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to examples, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.
Claims
1. A contact calibration method for a modular robot for solving global optimality, characterized in that: include: Obtain the homogeneous transformation matrix from the coordinate system of the robot wrist end to the robot base coordinate system when the robot's ruby probe contacts the standard ball; Establishing a calibration model based on the positions of the coordinate system origin of the ruby probe and the center of the standard ball in the robot base coordinate system; Based on the calibration model, an objective function is established according to a nonlinear least squares optimization method; Based on the objective function, the LM method and the DIRECT method are used to perform cyclic iterative calculations to obtain the global optimal solution of the objective function, including: Setting initial parameters of the LM method and iteratively calculating, in response to the objective function converging to a local minimum through iterative calculation of the LM method, using a local minimum point corresponding to the local minimum as an initial parameter of the DIRECT method; In response to the function value of the objective function iteratively calculated by the DIRECT method being less than the local minimum value, updating the initial parameters of the LM method, taking the point corresponding to the function value as the new initial parameters of the LM method, and iterating the calculation again; Repeat the iterative calculation steps until a global optimal solution of the objective function is obtained by the LM method; Based on the global optimal solution, the coordinate system origin position parameters of the ruby probe are obtained; Based on the positions of the coordinate system origin of the ruby probe and the center of the standard ball in the robot base coordinate system, a calibration model is established, including: Taking the standard ball as the reference target, the calibration model is established as follows: , in, is the position of the coordinate origin TCP of the ruby probe in the robot base coordinate system {B}; is the position of the center of the standard ball in the robot base coordinate system {B}, and its value is unknown; R and r are the radius of the standard ball and the radius of the ruby probe respectively; The objective function is: , 。 2. A contact calibration method for a combined robot for solving a global optimum according to claim 1, characterized in that: Get the homogeneous transformation matrix from the coordinate system of the robot wrist end to the robot base coordinate system when the robot's ruby probe contacts the standard ball, including: Install the standard ball on a magnetic base and attach it to a fixed bracket within the reach of the robot through a magnet; The robot moves toward the standard ball through different positions and postures; The robot controller obtains the homogeneous transformation matrix from the coordinate system {E} of tool0 at the end of the robot wrist to the robot base coordinate system {B} each time the ruby probe contacts the standard ball. .
3. A contact calibration method for a combined robot for solving a global optimum according to claim 2, characterized in that: The coordinate system origin position parameter of the ruby probe is , there is the following relationship: have to: , which is simplified to: ; in, is the posture matrix of the robot wrist end, is the position matrix of the robot wrist end.
4. A contact calibration method for a combined robot for solving a global optimum according to claim 1, characterized in that: Based on the objective function, the LM method and the DIRECT method are used to perform cyclic iterative calculations, which also includes: Set the iterative convergence condition, the minimum step size for iterative termination, the maximum number of iterations, and the maximum number of cycles of the LM method; After each iteration, determine whether the iteration step is less than the minimum step; If it is less than the minimum step size, determine whether the convergence condition is met. If the convergence condition is met, terminate the iteration and output the global optimal solution of the objective function. If the convergence condition is not met, output the local minimum value of the objective function. If it is not less than the minimum step size, determine whether the maximum number of iterations has been reached; If the maximum number of iterations has not been reached, the iterative calculation continues; If the maximum number of iterations is reached, the iteration is stopped and the result is output.
5. A contact calibration method for a combined robot for solving a global optimum according to claim 4, characterized in that: When the local minimum of the target function is output, it is determined whether the maximum number of loops has been reached. If so, the loop is terminated. If not, the local minimum is used as the initial parameter of the DIRECT method.
6. A contact calibration method for a combined robot for solving a global optimum according to claim 4, characterized in that: The iteration convergence condition is that the iteration result converges to within 0.1 mm.
7. A contact calibration method for a combined robot for solving a global optimum according to claim 4, characterized in that: Based on the objective function, the LM method and the DIRECT method are used to perform cyclic iterative calculations, which also includes: Set the iteration termination condition and search range control parameters of the DIRECT method, and use the local minimum value obtained by the LM method as the initial parameters of the DIRECT method; Determine whether the function value obtained by iterative calculation is less than the local minimum value. If it is less than the local minimum value, update the initial parameters of the LM method. If it is not less than the local minimum value, determine whether the iteration termination condition is met. If the iteration termination condition is not met, the iterative calculation continues. If the iteration termination condition is met, it is determined whether the maximum search range has been reached. If it has been reached, the optimal solution has not been found and the result is output. If it has not been reached, the search range is expanded by changing the value of the search range control parameter and the iterative calculation is repeated.
8. A combined robot contact calibration system for solving global optimality, characterized in that: A contact calibration method for a combined robot for solving a global optimum based on claim 1 comprises: A homogeneous transformation matrix acquisition module is used to obtain a homogeneous transformation matrix from the coordinate system of the robot wrist end to the robot base coordinate system when the ruby probe of the robot contacts the standard ball; A calibration model building module, used to build a calibration model based on the positions of the coordinate system origin of the ruby probe and the center of the standard ball in the robot base coordinate system; An objective function establishment module, used to establish an objective function based on the calibration model and according to a nonlinear least squares optimization method; A loop iterative calculation module, used to perform loop iterative calculation based on the objective function using the LM method and the DIRECT method to obtain a global optimal solution of the objective function; The position parameter acquisition module is used to obtain the coordinate system origin position parameters of the ruby probe based on the global optimal solution.
Citation Information
Patent Citations
Automatic calibrating and detecting method for executing end of robot based on ruby probe
CN109848989A
Robot milling system workpiece rapid positioning method based on laser tracker and binocular camera
CN111941425A
Robot hand-eye calibration method based on convex relaxation global optimization algorithm
CN110288660A
Instant mapping and positioning method, device and system and storage medium
CN111337947A
Camera attitude calibration method based on single vanishing point
CN115294210A
