A robot magnetorheological polishing pose closed-loop control method based on machine vision

Abstract

The application discloses a robot magnetorheological polishing pose closed-loop control method based on machine vision and relates to the technical field of optical element ultra-precision machining, and comprises the following steps: performing uniform polishing with different immersion depths on a plane mirror respectively, and shooting magnetorheological ribbon widths under different immersion depths in the polishing process; obtaining reference ribbon widths at different immersion depths, establishing a regular function; generating a numerical control program of robot polishing reference pose and speed; executing the reference numerical control program, calculating the deviation of the immersion depth at the moment from the given immersion depth according to the regular function, and converting the deviation into a pose compensation amount; and performing pose compensation through a single-chip microcomputer to control a stepping motor, so that the immersion depth in the whole polishing process is consistent with the set value, and the pose error of the robot is eliminated. According to the scheme, the normal motion error of the robot can be obtained according to the ribbon profile image, error closed-loop compensation is performed, the pose accurate control of the robot magnetorheological polishing is realized, and the stability of the removal efficiency in the polishing process is provided.

Description

Technical Field

[0001] This invention relates to the field of ultra-precision machining technology for optical components, specifically to a robot magnetorheological polishing pose closed-loop control method based on machine vision. Background Technology

[0002] Among numerous ultra-precision machining technologies, magnetorheological polishing technology, based on the rheological properties of magnetorheological polishing fluid, forms a "flexible polishing head" on the workpiece surface. By changing the magnitude of the magnetic field strength, the polishing force is controlled, thereby polishing the workpiece surface. This ensures low surface roughness, minimal subsurface damage, and high material removal efficiency. The "flexible polishing head" can achieve good conformity with curved surfaces, enabling high-precision manufacturing of various aspherical components.

[0003] Compared to traditional CNC machine tools, robots offer advantages such as lower cost and a wider machining range, making them suitable for machining complex curved surfaces. Robotic magnetorheological polishing technology is an advanced ultra-precision machining method that combines magnetorheological polishing with industrial robots. Its core is a robotic arm equipped with a magnetorheological flexible polishing tool, which, combined with the robot's multi-degree-of-freedom motion, enables the machining of complex curved surfaces. Industrial robots possess six degrees of freedom of motion, allowing them to cover complex curved surfaces (such as aspherical and freeform surfaces) and large-diameter workpieces (such as a 2.5-meter reflector), breaking through the machining range limitations of traditional CNC machine tools.

[0004] However, industrial robots suffer from rigidity issues, and the complex and heavy magnetorheological polishing tools they carry result in sub-millimeter positioning errors, exceeding the allowable range for magnetorheological polishing positioning errors. This makes stable control of removal efficiency impossible, necessitating pose compensation technology to meet the micron-level precision requirements of magnetorheological polishing. Some pose compensation techniques have complex closed-loop systems, high costs, and poor application adaptability, failing to meet high-precision positioning requirements. Therefore, it is necessary to research a high-precision closed-loop control method for robot pose error, which can simply and effectively eliminate robot positioning errors and improve the surface convergence capability of robot magnetorheological polishing. Summary of the Invention

[0005] This invention addresses the problems in existing robotic magnetorheological polishing normal-assisted positioning devices, where the robot's motion accuracy and rigidity are poor, making it difficult to meet the positioning accuracy requirements of the magnetorheological polishing process, hindering stable control of polishing removal efficiency, and affecting the processing accuracy of optical components. The aim is to provide a machine vision-based closed-loop control method for robot magnetorheological polishing posture. This method can obtain the robot's normal motion error from the ribbon contour image and perform error closed-loop compensation, achieving precise posture control for robot magnetorheological polishing and ensuring stability of removal efficiency during the polishing process.

[0006] This invention is achieved through the following technical solution: A machine vision-based closed-loop control method for robot magnetorheological polishing pose includes the following steps: S1: Based on the given process parameters, perform uniform polishing at different depths on a plane mirror, and take images of the magnetorheological ribbon shape contour at different immersion depths during the polishing process. S2: Based on the edge detection algorithm, extract the ribbon edge contour in the image, calculate the ribbon width, obtain the reference ribbon width under different immersion depths, and establish the regular function of the robot magnetorheological polishing immersion depth and ribbon width under the process parameters. S3: Establish an aspherical model of the workpiece, calculate the workpiece surface pose at a given resolution; use the image deconvolution algorithm, and according to the initial surface shape of the workpiece and the removal function at a given immersion depth, iteratively calculate the dwell time corresponding to the workpiece surface pose and convert it into the robot feed speed, and generate the CNC program for the reference pose and speed of the robot magnetorheological polishing machine. S4: Execute the CNC program and acquire the shape and contour image of the polishing wheel ribbon in real time, extract the ribbon edge contour, calculate the real-time ribbon width, and calculate the deviation between the immersion depth and the given immersion depth at this moment according to the law function, and convert it into the pose compensation amount of the robot magnetorheological polishing machine. S5: Based on the pose compensation amount, control the robotic arm to perform pose compensation, change the contact depth between the magnetorheological polishing ribbon and the polishing workpiece, so that the immersion depth throughout the polishing process is the same as the set value, and eliminate the robot's normal pose error.

[0007] Furthermore, step S2 also includes the following specific steps: S21: Convert the original magnetorheological ribbon shape outline image into a grayscale portrait. The image was smoothed using a Gaussian filter. S22: The Sobel operator for edge detection in the image is used to calculate the image gradient, and the image gradient magnitude is classified by an adaptive thresholding method to extract the true edge information; S23: Binarize the current pixel value, identify the edge pixel position when the ribbon cuts into the workpiece using the binarized image, and calculate the immersion depth. The width of the ribbon below ; S24: Finally, a third-order polynomial fitting method is used to fit the relationship between immersion depth and ribbon width under this process parameter. .

[0008] Further optimization is achieved by using the Gaussian filter to smooth the image in step S21, as follows: ; in, , =1.5.

[0009] Furthermore, in step S22, the expression for calculating the image gradient using the edge detection Sobel operator is as follows: ; ; ; The expression for classifying image gradient magnitudes using an adaptive thresholding method and extracting true edge information is as follows: ; in, Based on pixels The average grayscale value of a local area centered at a window size of 13*13. This is an empirically adjusted parameter, set to 10.

[0010] Further optimization is achieved by using the binarization formula in step S23 as follows: .

[0011] Further optimization yields the following formula for the regularity function: .

[0012] Furthermore, step S3 also includes the following specific steps: S31: Establish the aspherical model of the workpiece, and calculate the pose of the workpiece's dwell point based on the workpiece's dimensions and aspherical parameters. ; S32: At a given immersion depth Below, based on the initial shape of the aspherical surface Remove function distribution The dwell time of the dwell point is solved iteratively by using the image deconvolution algorithm; S33: Finally, convert the obtained dwell time into feed rate. Generate the reference pose of the robot magnetorheological polishing machine With feed rate The CNC program.

[0013] Furthermore, in step S32, the formula for iteratively solving the dwell time of the dwell point using the image deconvolution algorithm is as follows: ; in, It is a smoothing factor, ranging from 0.01 to 1.

[0014] To further optimize this process, in step S33, before converting the obtained dwell time into feed rate, a bilateral constraint on the speed needs to be applied. The constraint conditions are as follows: ; in, , , , These are machine tool acceleration, minimum speed, maximum speed, and speed node spacing, respectively.

[0015] Further optimization, in step S4, the immersion depth The expression is: ;in, This is the real-time ribbon width; The formula for calculating the deviation between the actual immersion depth and the given immersion depth is: .

[0016] Compared with the prior art, the present invention has the following advantages and beneficial effects: 1. The present invention provides a robot magnetorheological polishing pose closed-loop control method based on machine vision, which can obtain the robot normal motion error based on the ribbon contour image acquired by the industrial camera and perform error closed-loop compensation, thereby realizing precise pose control of the robot magnetorheological polishing and providing stability of removal efficiency during the polishing process. 2. This invention provides a machine vision-based closed-loop control method for robot magnetorheological polishing pose. The closed-loop control device, independently built within the robot polishing machine, is not limited by the robot's system or performance, exhibiting good versatility. The closed-loop control system achieves high positioning accuracy through robot localization and real-time pose compensation based on machine vision, meeting the control requirements for stable removal efficiency in magnetorheological polishing. Optimizing the industrial camera resolution increases the pixel count of the magnetorheological ribbon image, thereby improving the measurement accuracy of the ribbon width. Optimizing the edge detection algorithm reduces computation time, ensuring rapid system response. 3. This invention provides a machine vision-based closed-loop control method for robot magnetorheological polishing pose. Compared to force feedback control methods, it is unaffected by changes in the shape and curvature of aspherical components on the detection results. The system has a simple structure, fewer model parameters, high computational efficiency, and strong applicability. It overcomes the limitations of applying magnetorheological polishing to low-precision robots, reduces the accuracy requirements for equipment positioning in magnetorheological polishing technology, and promotes the development of magnetorheological polishing technology. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings: Figure 1 This is a schematic diagram of the posture control device of the present invention; Figure 2 This is a schematic diagram of the control framework of the present invention; Figure 3 This is a schematic diagram of the closed-loop control based on machine vision according to the present invention; Figure 4 This is a schematic diagram of the workflow of the robot magnetorheological polishing pose control method based on machine vision according to the present invention. Figure 5 This is the original grayscale image captured by the industrial camera in Example 2 of the present invention; Figure 6 The width identification result of the ribbon with a immersion depth of 0.4 mm in Example 2 of this invention; Figure 7 This is a vertical projection distribution diagram of the ribbon image in Example 2 of the present invention; Figure 8 This is the fitting curve of the ribbon width at different immersion depths in Example 2 of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of this invention are only for explaining this invention and are not intended to limit this invention.

[0019] Example 1: This example 1 provides a robot magnetorheological polishing pose closed-loop control method based on machine vision. The system structure and working principle are as follows: Figures 1 to 4 As shown, it includes the following steps: S1. Under given process parameters, perform uniform polishing at different immersion depths on a plane mirror, and use an industrial camera to capture images of the magnetorheological ribbon shape contour at different immersion depths during the polishing process. S2. The edge detection algorithm in image processing is used to extract the ribbon edge contour, calculate the ribbon width, obtain the reference ribbon width under different immersion depths, and use the third-order polynomial fitting method to establish the law function between the immersion depth and the ribbon width of the robot magnetorheological polishing under the process parameters. S3. Establish an aspherical model of the workpiece, and calculate the workpiece surface pose at a given resolution; using the image deconvolution algorithm, based on the initial surface shape of the workpiece and the removal function at a given immersion depth, iteratively calculate the dwell time corresponding to the workpiece surface pose and convert it into the robot feed speed, generating the CNC program for the reference pose and speed of the robot magnetorheological polishing machine. S4. Execute the reference CNC program. The industrial camera acquires the shape and contour image of the polishing wheel ribbon in real time. The edge detection algorithm in image processing is used to extract the ribbon edge contour. The real-time ribbon width is calculated. According to the regular function, the deviation between the immersion depth at this moment and the given immersion depth is calculated and converted into the pose compensation amount of the robot magnetorheological polishing machine. S5. By controlling the robotic arm with a microcontroller to perform pose compensation, the immersion depth throughout the polishing process is the same as the set value, eliminating the robot's pose error and improving the processing accuracy of the robot's magnetorheological polishing.

[0020] In this embodiment, step S2 includes: The acquired magnetorheological fluid ribbon images include the maximum width of the ribbon in contact with the polishing wheel and the width of the ribbon in contact with the workpiece. The original images are first converted to grayscale. Gaussian filtering is used to smooth the image and reduce measured noise. Its expression is: (1.1) in , =1.5.

[0021] The Sobel operator for edge detection is used to calculate the image gradient, and its expression is: (1.2) (1.3) (1.4) Image gradient magnitudes are classified using an adaptive thresholding method to extract true edge information and further suppress noise. The expression is as follows: (1.5) in, Based on pixels The average grayscale value of a local area centered at a window size of 13*13. This is an empirically adjusted parameter, set to 10.

[0022] The formula for binarizing the current pixel value is as follows: (1.6) The edge pixel position when the ribbon cuts into the workpiece is identified by binarizing the image, and the immersion depth is calculated. The width of the ribbon below .

[0023] A third-order polynomial was used to fit the relationship between immersion depth and ribbon width under this process parameter, and its expression is as follows: (1.7) In this embodiment, step S3 includes: Calculate the pose of the workpiece's dwell point based on the workpiece's dimensions and aspherical parameters. .

[0024] At a given immersion depth Below, based on the initial shape of the aspherical surface Remove function distribution The dwell time of the dwell points is solved iteratively using the image deconvolution algorithm, and its expression is: (1.8) in It is a smoothing factor, ranging from 0.01 to 1.

[0025] To meet the dynamic performance constraints of the robot, a two-sided velocity constraint is applied, with the following conditions: (1.9) in, , , , These are machine tool acceleration, minimum speed, maximum speed, and speed node spacing, respectively.

[0026] Convert the obtained residence time into feed rate. Generate the reference pose of the robot magnetorheological polishing machine With feed rate The CNC program.

[0027] In this embodiment, step S4 includes: The industrial camera executes a reference CNC program to acquire real-time images of the shape and contour of the polishing wheel ribbon. An edge detection algorithm in image processing is used to extract the ribbon edge contour, and the real-time ribbon width is calculated. ; Calculate the immersion depth at that moment based on the regularity function. Its expression is: (1.10) Obtain the deviation between the actual immersion depth and the given immersion depth. ; The deviation between the actual immersion depth and the given immersion depth is used as the real-time coordinate pose compensation amount.

[0028] In this embodiment, step S5 includes: The microcontroller controls the stepper motor speed and working time interval, which in turn drives the polishing component to move up and down via the ball screw slide. This changes the contact depth between the magnetorheological polishing ribbon and the polishing component, until the width of the polishing ribbon is controlled within a set range. This indirectly enables position monitoring and compensation during the polishing process, ensuring the accuracy of the polishing posture.

[0029] Example 2: Based on Example 1, Example 2 provides a specific implementation case, such as... Figures 5-8 As shown.

[0030] This embodiment provides a robot magnetorheological polishing pose closed-loop control method based on machine vision, the specific steps of which are as follows: Step 1: Under given process parameters, perform uniform polishing at different immersion depths on a plane mirror, and use an industrial camera to capture images of the magnetorheological ribbon shape contour at different immersion depths during the polishing process. Specifically, uniform polishing is performed on the plane mirror with an immersion depth range of 0 to 0.5 mm, and the immersion depths for each uniform polishing step are... The experimental specimens and uniform polishing parameters are shown in Table 1: Table 1: Experimental Parts and Uniform Polishing Parameter Information

[0031] The uniform polishing trajectory was achieved using grating lines with a trajectory spacing and pitch of 1 mm. Three ribbon contour images were acquired for each experiment using an industrial camera. Figure 5 Image of the ribbon profile at a immersion depth of 0.4 mm.

[0032] Step 2: Convert the image to grayscale, filter out measurement noise using Gaussian filtering, and calculate the image gradient using the Sobel operator for edge detection. Figure 6 To determine the vertical projection distribution of the ribbon image, an adaptive thresholding method was used to classify the image gradient magnitude, extract the ribbon edge contour, count the edge pixels, and calculate the baseline ribbon width at different immersion depths. The results are shown in Table 2: Table 2: Ribbon width at various immersion depths during uniform polishing

[0033] Figure 7 The information on the ribbon contour and width extracted by edge detection is provided.

[0034] A third-order polynomial fitting method was used to establish the regular function of the robot magnetorheological polishing immersion depth and ribbon width under this process parameter; (1.11) in, =-0.02347, =-0.07348, =0.05741, =-0.00497.

[0035] Figure 8 The fitting curves of immersion depth and ribbon width are given.

[0036] Step 3: Establish an aspherical model of the workpiece, given a resolution of 0.5mm, and calculate the pose of the workpiece surface. The immersion depth is set to 0.4mm. An image deconvolution algorithm is used to determine the initial surface shape of the workpiece. and removal function Iteratively calculate the dwell time corresponding to the workpiece surface pose. And convert it into robot feed speed A numerical control program for generating the reference pose and velocity of a robot magnetorheological polishing machine; Step 4: Execute the reference CNC program. The industrial camera acquires real-time images of the shape and contour of the polishing wheel ribbon. The edge detection algorithm in image processing is used to extract the ribbon edge contour, and the real-time ribbon width is calculated. Calculate the immersion depth at that moment based on the regularity function. With a given immersion depth deviation This is converted into the pose compensation amount of the robotic magnetorheological polishing machine.

[0037] Step 5: When When the depth deviation exceeds ±0.005mm, the closed-loop control system is activated. The microcontroller sends a pulse signal to control the servo motor to drive the polishing wheel base to move up and down for posture compensation, so that the immersion depth deviation throughout the polishing process does not exceed ±0.005mm. This achieves posture monitoring and closed-loop control throughout the polishing process, eliminates robot posture errors, and ensures the processing accuracy of robot magnetorheological polishing.

[0038] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A closed-loop control method for robot magnetorheological polishing pose based on machine vision, characterized in that, Includes the following steps: S1: Based on the given process parameters, perform uniform polishing at different depths on a plane mirror, and take images of the magnetorheological ribbon shape contour at different immersion depths during the polishing process. S2: Based on the edge detection algorithm, extract the ribbon edge contour in the image, calculate the ribbon width, obtain the reference ribbon width under different immersion depths, and establish the regular function of the robot magnetorheological polishing immersion depth and ribbon width under the process parameters. S3: Establish an aspherical model of the workpiece, calculate the workpiece surface pose at a given resolution; use the image deconvolution algorithm, and according to the initial surface shape of the workpiece and the removal function at a given immersion depth, iteratively calculate the dwell time corresponding to the workpiece surface pose and convert it into the robot feed speed, and generate the CNC program for the reference pose and speed of the robot magnetorheological polishing machine. S4: Execute the CNC program and acquire the shape and contour image of the polishing wheel ribbon in real time, extract the ribbon edge contour, calculate the real-time ribbon width, and calculate the deviation between the immersion depth and the given immersion depth at this moment according to the law function, and convert it into the pose compensation amount of the robot magnetorheological polishing machine. S5: Based on the pose compensation amount, control the robotic arm to perform pose compensation, change the contact depth between the magnetorheological polishing ribbon and the polishing workpiece, so that the immersion depth throughout the polishing process is the same as the set value, and eliminate the robot's normal pose error.

2. The robot magnetorheological polishing pose closed-loop control method based on machine vision according to claim 1, characterized in that, Step S2 further includes the following specific steps: S21: Convert the original magnetorheological ribbon shape outline image into a grayscale portrait. The image was smoothed using a Gaussian filter. S22: The Sobel operator for edge detection in the image is used to calculate the image gradient, and the image gradient magnitude is classified by an adaptive thresholding method to extract the true edge information; S23: Binarize the current pixel value, identify the edge pixel position when the ribbon cuts into the workpiece using the binarized image, and calculate the immersion depth. The width of the ribbon below ; S24: Finally, a third-order polynomial fitting method is used to fit the relationship between immersion depth and ribbon width under this process parameter. .

3. The robot magnetorheological polishing pose closed-loop control method based on machine vision according to claim 2, characterized in that, In step S21, the formula for smoothing the image using Gaussian filtering is: ； in, , =1.

5.

4. The robot magnetorheological polishing pose closed-loop control method based on machine vision according to claim 3, characterized in that, In step S22, the expression for calculating the image gradient using the edge detection Sobel operator is as follows: ； ； ； The expression for classifying image gradient magnitudes using an adaptive thresholding method and extracting true edge information is as follows: ； in, Based on pixels The average grayscale value of a local area centered at a window size of 13*13. This is an empirically adjusted parameter, set to 10.

5. The robot magnetorheological polishing pose closed-loop control method based on machine vision according to claim 4, characterized in that, The binarization formula in step S23 is: 。 6. The robot magnetorheological polishing pose closed-loop control method based on machine vision according to claim 5, characterized in that, The formula for the regularity function is: 。 7. The robot magnetorheological polishing pose closed-loop control method based on machine vision according to claim 1, characterized in that, Step S3 further includes the following specific steps: S31: Establish the aspherical model of the workpiece, and calculate the pose of the workpiece's dwell point based on the workpiece's dimensions and aspherical parameters. ; S32: At a given immersion depth Below, based on the initial shape of the aspherical surface Remove function distribution The dwell time of the dwell point is solved iteratively by using the image deconvolution algorithm; S33: Finally, convert the obtained dwell time into feed rate. Generate the reference pose of the robot magnetorheological polishing machine With feed rate The CNC program.

8. The robot magnetorheological polishing pose closed-loop control method based on machine vision according to claim 7, characterized in that, In step S32, the formula for iteratively solving the dwell time of the dwell point using the image deconvolution algorithm is as follows: ； in, It is a smoothing factor, ranging from 0.01 to 1.

9. The robot magnetorheological polishing pose closed-loop control method based on machine vision according to claim 8, characterized in that, In step S33, before converting the obtained dwell time into feed rate, a bilateral constraint on the speed is required, and the constraint conditions are as follows: ； in, , , , These are machine tool acceleration, minimum speed, maximum speed, and speed node spacing, respectively.

10. The robot magnetorheological polishing pose closed-loop control method based on machine vision according to claim 1, characterized in that, In step S4, the immersion depth The expression is: ;in, This refers to the real-time ribbon width. The formula for calculating the deviation between the actual immersion depth and the given immersion depth is: 。

Images