CN111823221A - Robot polishing method based on multiple sensors - Google Patents
Robot polishing method based on multiple sensors Download PDFInfo
- Publication number
- CN111823221A CN111823221A CN201910236303.6A CN201910236303A CN111823221A CN 111823221 A CN111823221 A CN 111823221A CN 201910236303 A CN201910236303 A CN 201910236303A CN 111823221 A CN111823221 A CN 111823221A
- Authority
- CN
- China
- Prior art keywords
- robot
- polishing
- defect
- sensor
- grinding
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000005498 polishing Methods 0.000 title claims abstract description 63
- 238000000034 method Methods 0.000 title claims abstract description 25
- 230000007547 defect Effects 0.000 claims abstract description 68
- 238000010586 diagram Methods 0.000 claims abstract description 17
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 15
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 11
- 230000011218 segmentation Effects 0.000 claims abstract description 6
- 238000007781 pre-processing Methods 0.000 claims abstract description 5
- 238000001514 detection method Methods 0.000 claims description 10
- 238000003709 image segmentation Methods 0.000 claims description 7
- 230000000007 visual effect Effects 0.000 claims description 6
- 230000000694 effects Effects 0.000 claims description 4
- 238000012549 training Methods 0.000 claims description 4
- 230000002238 attenuated effect Effects 0.000 claims 1
- 230000005484 gravity Effects 0.000 abstract description 7
- 238000001914 filtration Methods 0.000 abstract description 3
- 230000003068 static effect Effects 0.000 abstract description 2
- 238000004519 manufacturing process Methods 0.000 abstract 3
- 239000011159 matrix material Substances 0.000 description 16
- 230000008569 process Effects 0.000 description 9
- 230000009466 transformation Effects 0.000 description 8
- 238000006243 chemical reaction Methods 0.000 description 7
- 238000013135 deep learning Methods 0.000 description 4
- 238000006073 displacement reaction Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000013519 translation Methods 0.000 description 3
- 230000001133 acceleration Effects 0.000 description 2
- 230000007797 corrosion Effects 0.000 description 2
- 238000005260 corrosion Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000007517 polishing process Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 239000004576 sand Substances 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 241000221535 Pucciniales Species 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000011217 control strategy Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 239000012636 effector Substances 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000003754 machining Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000004080 punching Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J11/00—Manipulators not otherwise provided for
- B25J11/005—Manipulators for mechanical processing tasks
- B25J11/0065—Polishing or grinding
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1694—Programme controls characterised by use of sensors other than normal servo-feedback from position, speed or acceleration sensors, perception control, multi-sensor controlled systems, sensor fusion
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1694—Programme controls characterised by use of sensors other than normal servo-feedback from position, speed or acceleration sensors, perception control, multi-sensor controlled systems, sensor fusion
- B25J9/1697—Vision controlled systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
Landscapes
- Engineering & Computer Science (AREA)
- Robotics (AREA)
- Mechanical Engineering (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Quality & Reliability (AREA)
- Finish Polishing, Edge Sharpening, And Grinding By Specific Grinding Devices (AREA)
- Manipulator (AREA)
Abstract
The invention discloses a robot polishing method based on multiple sensors, which comprises the following steps: (1) selecting the type of hardware of the robot polishing system, and calibrating the hardware composition of the vision sensor; (2) utilizing a vision sensor to collect defects on the surface of the automobile stamping part, manually marking the collected image data to obtain a true value diagram, and establishing a defect data set; (3) completing the segmentation of the surface defects of the workpiece by utilizing a PixelNet convolutional neural network; (4) preprocessing signals acquired by a torque sensor from three aspects of static calibration, filtering and gravity compensation; (5) the robot defect grinding system with the torque sensor utilizes known defect information and a force/position hybrid control algorithm and adopts an orthogonal path grinding scheme to grind the defects. The invention can be widely applied to the technical fields of production and manufacturing, such as aerospace, rail transit, medical appliances, furniture manufacturing and the like which need industrial robots.
Description
Technical Field
The invention belongs to the field of machine vision and sensors, and relates to a robot polishing method based on multiple sensors, which can be used for polishing industrial robots.
Background
With the gradual maturity of the related technologies of the industrial robot in the aspects of hardware and software, the precision of the robot body and the motion is greatly improved, and the cost is also greatly reduced. Aiming at the problem that workpieces with large sizes and complex shapes are difficult to polish, domestic and foreign experts and scholars begin to correspondingly research the polishing technology of the industrial robot on the basis of a large amount of research results of the industrial robot. Currently, the main research direction is developed according to factors affecting the polishing of industrial robots, including polishing tool materials, polishing path planning, robot speed control, polishing contact force, and the like.
To the problem that defect is polished inefficiency, and the quality is difficult to guarantee, this paper proposes a robot system of polishing based on multisensor, and this system includes two subsystems: based on visual sensor defect detecting system and based on moment sensor defect system of polishing. The visual sensor is used for detecting the surface defects of the workpiece to be detected, polishing operation is carried out on the defect area on the basis of detecting the defects, and polishing quality is guaranteed by means of the torque sensor in the polishing process. Efficiency and quality of industrial robot in automobile punching part surface is polished can be improved through this system.
Disclosure of Invention
The invention provides a robot polishing method based on multiple sensors, which introduces a visual sensor to collect defects on the surface of an automobile stamping part, adopts an image segmentation algorithm, and completes the segmentation of the defects on the surface of a workpiece by using a PixelNet convolutional neural network; the robot defect polishing system carrying the torque sensor utilizes known defect information and adopts an orthogonal path polishing scheme to polish the defects. The method of deep learning and sensors is well applied to robot polishing, and a good effect is achieved. The technical scheme for realizing the aim of the invention comprises the following steps:
step 1: selecting the type of hardware of the robot polishing system, and calibrating the hardware composition of the vision sensor;
step 2: utilizing a vision sensor to collect defects on the surface of the automobile stamping part, manually marking the collected image data to obtain a true value diagram, and establishing a defect data set;
and step 3: an image segmentation algorithm is adopted, and a PixelNet convolutional neural network is utilized to complete the segmentation of the surface defects of the workpiece;
and 4, step 4: preprocessing signals acquired by a torque sensor;
and 5: and defect polishing based on a torque sensor, wherein a robot defect polishing system performs polishing operation on the defects by adopting an orthogonal path polishing scheme by using a force/position hybrid control algorithm.
Compared with the prior art, the invention has the beneficial effects that:
1. a set of polishing system based on a multi-sensor robot is designed, and the system is divided into two subsystems: stamping workpiece surface defect detecting system based on vision sensor, defect system of polishing based on torque sensor have solved present and have used the artifical mode of polishing to give first place to, and the operational environment of polishing is abominable, wastes time and energy, and is inefficient, the problem that the quality of polishing can not guarantee moreover.
2. A stamping part surface defect detection method is provided based on machine vision and deep learning, the surface image of the stamping part is collected, the defects are marked as scratch and corrosion, the PixelNet convolutional neural network is adopted to carry out image segmentation on the surface defects to obtain a defect detection result, and the problems that the surface background of the automobile stamping part is complex, the defect shapes are various and the defect shapes are difficult to segment are solved.
3. The force/position control strategy is adopted, and the offset of each axis of the robot is adjusted in real time by means of torque information obtained by feedback of the torque sensor, so that the contact force is kept constant, constant-force polishing is realized, and the problem that the polishing quality of a complex curved surface is difficult to guarantee is solved.
Drawings
FIG. 1 is a diagram of an overall system of a multi-sensor grinding robot;
figure 2 industrial robot model IRB1410, produced by ABB corporation switzerland; FIGS. 3(a), 3(b) and 3(c) are MV-CE050-50GM industrial camera, MVL-MF1220M-5MP lens and light source manufactured by Haekwover corporation, respectively;
FIGS. 4(a) and 4(b) are respectively Delta IP 60F/T Senor torque sensor and sanding tool developed by ATI, Inc.;
FIG. 5(a) is a perspective projection model in which the camera coordinate system OC-XCYCZCWorld coordinate system OW-XWYWZWAn image plane coordinate system o' -xy and an image coordinate system o-uv;
FIG. 5(b) is a hand-eye relationship diagram and relative positions of coordinates, wherein CbaseAs a robot coordinate system, CtoolAs a robot end coordinate system, CcamAs a camera coordinate system, CcalAs a calibration plate coordinate system, fig. 5(c) is a hand-eye calibration process diagram;
FIGS. 5(d) and 5(e) are a tool coordinate diagram and a three-point method schematic diagram, respectively, wherein CbaseAs a robot coordinate system, CtoolAs a robot end coordinate system, CpolIs a grinding tool coordinate system;
FIG. 6 is a diagram of a PixelNet network architecture;
FIG. 7 shows scratches and corrosion on the surface of a stamped part of an automobile body;
FIG. 8 is an overall defect polishing process based on a torque sensor;
figure 9 is a force/position hybrid control algorithm diagram,
wherein, XD、
And
respectively the desired cartesian spatial displacement, velocity, acceleration; fDIs the desired force; xe、
And
actual cartesian spatial displacement, velocity, acceleration, respectively; feIs a cartesian space interaction force between the tool and the environment; s is a diagonal selection matrix, the elements on the diagonal are 1 or 0, and the matrix is determined by a constraint relation: when the constraint is natural force constraint and position control is required, taking 1 as a corresponding element, and otherwise, taking 0 as the corresponding element; I-S is also a diagonal selection matrix, and S and I-S are mutually orthogonal and determine each degree of freedom control mode; xp、Xp、Xf、XfAnd XcFor displacement offset, theta, of each axis of the robotpAnd thetapThe angular offset of each axis of the robot;
FIG. 10 is a graph showing the variation of the contact force of the sanding;
FIG. 11 is a partial surface comparison of a stamping before and after grinding.
Detailed Description
The invention realizes a robot polishing method based on multiple sensors through the following steps, which are specifically implemented as follows:
step 1: selecting the type of hardware of a robot polishing system, calibrating the hardware composition of a vision sensor, including camera calibration, hand-eye calibration and polishing tool calibration, and determining the coordinate conversion relation among the robot, the camera, the polishing tool and a polishing workpiece; when the camera is calibrated, a perspective projection model of the camera is adopted, coordinate conversion among all coordinate systems is completed, and finally a conversion formula from a world coordinate system to an image coordinate system is obtained:
wherein f isx＝f/sxAs an image coordinate system uiNormalized focal length of axis, fy＝f/syIs v isiScale factor of the axis, M1From fx、fy、uo、voFour parameters, which are only related to the camera structure, are called camera intrinsic parameters, M2The system consists of a rotation matrix and a translation vector from a world coordinate system to a camera coordinate system, and is called as a camera external parameter, and M is called as a projection matrix of a camera;
when Hand-Eye calibration is carried out, when an Eye-in-Hand mode is adopted, the camera is installed at the tail end of the robot, the position relation between the camera and the tail end of the robot is kept unchanged, the camera shoots a calibration plate at a known space position under different poses of the tail end of the industrial robot, and then the relative relation of the positions of the camera and the calibration plate is solved; the transformation relation matrix between the camera and the robot tip front-back position is denoted A, B, C, D, X, which describes the rotation matrix R and the translation vector t between the respective coordinate systems:
when the coordinate system of the grinding tool is calibrated, the grinding tool is installed on a flange plate at the tail end of the mechanical arm through a torque sensor, the pose of the robot is adjusted to enable the tail end of the grinding tool to point to the same point from three different directions, and in the calibration process, the homogeneous transformation matrix of the coordinate system at the tail end of the robot relative to the coordinate system of the robot body is obtained through the record of a robot control system.
Step 2: utilizing a vision sensor to collect defects on the surface of the automobile stamping part, manually marking the collected image data to obtain a true value diagram, and establishing a defect data set; when the vision sensor collects the surface defects of the workpiece, the data set is expanded, and different poses of the sample are collected at the same height by turning and rotating the simulation camera; simulating the poses of the cameras at different heights by scale transformation to acquire samples; simulating the influence of the field illumination environment on the sample by adjusting the brightness of the sample; through the superposition use of the modes and the elimination of partial data, the image data is finally expanded from the initial 3000 to 1.5 ten thousand;
and step 3: and (3) completing the segmentation of the surface defects of the workpiece by using a PixelNet convolutional neural network by adopting an image segmentation algorithm: 9000 images are selected from the data set as a training set, 3000 images are selected as a verification set, and 3000 images are selected as a test set to train and test the PixelNet network. The training process takes 90 images as one iteration step, wherein the initial learning rate is set to be 0.01, the final attenuation is to be 0.0001 along with the increase of the number of training rounds, and the number of iterations is 10 ten thousand. The PixelNet adopts pixel layered sampling, increases sample diversity in the batch random gradient descent updating process, and introduces a complex nonlinear predictor to improve the classification precision of pixels; the PixelNet establishes a predictor on the multi-scale features extracted by the multilayer convolutional neural network, uses a super-column descriptor to refer to the features extracted by a plurality of layers corresponding to the same pixel, and the calculation formula of the multi-scale super-column feature of the pixel point p is as follows: h isp(X)＝[c1(p)，c2(p)，…，cM(p)](ii) a Wherein, ci(p) is a feature vector corresponding to the i-th layer convolution with the pixel point p as the center, and X is an input image.
And 4, step 4: preprocessing signals acquired by a torque sensor from three aspects of static calibration, filtering and gravity compensation; wherein, the filtering is to design a Butterworth low-pass filter; the order of the digital low-pass filter obtained by bilinear transformation digitization is 7, 3dB down to the frequency omegac＝1.3539*103。
For measuring grinding tool gravity F during gravity compensationGThe influence on the acquisition value requires the action force F between the grinding tool and the workpiece to be groundE Set 0, thus adjusting the pose of the robot so that the end of the abrading tool does not contact the workpiece to be abraded, at which time F E0; the gravity direction of the grinding tool is opposite to the Z-axis direction of the robot body coordinate system, and the gravity of the grinding tool has the following coordinates under the robot body coordinate system:bFG＝[0 0 -GT]Tthrough the robot body coordinate system and the robot end coordinate system transformation matrix
Conversion matrix of robot tail end coordinate system and torque sensor coordinate system
Converting the gravity of the grinding tool into a torque sensor coordinate system, wherein the conversion relation is as follows:
SFGfor the amount of deviation of the grinding tool weight in the torque sensor coordinate system, in order to eliminate this deviation, the measured values of the torque sensor in the torque sensor coordinate system need to be measuredsFSAnd (3) correcting, namely:sFE＝sFS-sFG(ii) a Measured value of torque sensorsFSAnd after correction, carrying out coordinate transformation, and transferring to a robot body coordinate system:
and 5: the robot defect polishing system with the torque sensor utilizes the known defect information and adopts an orthogonal path polishing scheme to polish the defects: moving the polishing tool to the central point position of the defect to perform polishing tasks according to a planned path, wherein the polishing processing range is uniformly expanded, and each point in the defect range is uniformly stressed, so that the polishing effect is ensured; in the aspect of grinding force/position control, a force/position hybrid control algorithm is provided for realizing the constant-force grinding of the robot: the force/position hybrid control algorithm defines two spatial force and position spaces for force and position, respectively.
The whole process is described in detail with reference to the attached drawings of the specification:
1. hardware selection for multi-sensor based robotic vision sensor
Presented herein is a multi-sensor based robotic polishing system comprising two subsystems: based on visual sensor defect detecting system and based on moment sensor defect system of polishing. FIG. 1 is a diagram of an overall system of a multi-sensor grinding robot; the robot is an industrial robot model IRB1410, which is produced by ABB of Swiss and is shown in fig. 2, and has six joints which can independently operate, and the relative independence of the six joints ensures that the industrial robot has a large working space. The industrial robots of the type have 1, 2 and 3 shafts which are mainly used for adjusting the overall space position of the robot, and 4, 5 and 6 shafts which are mainly used for adjusting the space pose of a tool at the tail end of the robot, and have larger rotation range, so that the end effector has higher flexibility;
the size of the defect detection view field is 100 multiplied by 75mm2And the theoretical precision is 0.05mm, the unidirectional resolution of the camera is 2000pix and 1600pix respectively, and the resolution of the camera is at least 320 ten thousand. Based on the above consideration, the MV-CE050-50GM industrial camera manufactured by Hai Congwortz corporation and shown in FIG. 3(a) is selected, the resolution of the camera reaches 500 ten thousand, and the acquisition frame rate is 14fps, so that the system design requirement is met. According to the object distance, the size of a CCD target surface and the size of a view field, an MVL-MF1220M-5MP lens shown in fig. 3(b) is selected, and according to the characteristics of a workpiece to be detected, a low-angle annular light source is adopted to polish right above the workpiece, wherein the adopted light source is shown in fig. 3 (c);
higher accuracy and frequency are required in robotic sanding operations, so the Delta IP 60F/T Senor torque sensor and sanding tool developed by ATI corporation, shown in FIGS. 4(a) (b), were selected.
2. Calibration of robot vision sensor based on multiple sensors
The calibration comprises camera calibration, hand-eye calibration and polishing tool calibration. A perspective projection model of the camera is adopted during camera calibration to complete coordinate conversion between coordinate systems, and fig. 5(a) is a perspective projection model in which a camera coordinate system OC-XCYCZCWorld coordinate system OW-XWYWZWAn image plane coordinate system o' -xy and an image coordinate o-uv; the camera calibration method mainly comprises the following steps: multiple moving plane checkerboard calibration in camera view field rangeShooting the image of the calibration plate, and ensuring that the calibration plate changes in the depth direction of the camera when moving; extracting image coordinates of checkerboard corner points of the calibration plate; camera internal parameters are calculated from the camera model.
As shown in the left diagram of fig. 5(b), the camera is mounted at the robot end while the positional relationship between the camera and the robot end is maintained, in the Eye-in-Hand manner. The camera shoots a calibration plate with a known spatial position under different poses of the tail end of the industrial robot, and then the relative relation of the position of the calibration plate and the position of the calibration plate is solved. As shown in the right diagram of fig. 5(b), the transformation relation matrix between the front and rear positions of the camera and the robot end is represented by A, B, C, D, X, which describes the rotation matrix R and the translation vector t between the respective coordinate systems:
and then adjusting the pose of the robot, and keeping the calibration plate within the visual field of the camera. According to the hand-eye calibration principle, the calibration steps are as shown in the hand-eye calibration process diagram of fig. 5(c), and firstly, the internal parameters of the camera are calibrated: then adjusting the pose of the robot twice, respectively calibrating external parameters of the camera and recording the tail end position information of the robot before and after adjustment; respectively calculating a camera coordinate system change matrix C and a robot tail end position change matrix D according to the camera external parameters recorded before and after adjustment and the robot tail end position information; and repeating the steps for multiple times to obtain multiple groups of transformation matrix data, and obtaining the hand-eye calibration matrix X in a simultaneous manner.
In the force sensor based defect grinding system, the grinding tool is mounted on the end flange of the robot arm by means of a torque sensor, as shown in fig. 5 (d). In order to obtain a conversion relationship between the coordinate system of the grinding tool and the coordinate system of the robot tip, it is necessary to adjust the robot pose so that the grinding tool tip points to the same point from three different directions, as shown in fig. 5 (e). In order to improve the calibration accuracy, the tip of the grinding tool can be used as a fixed reference point in a world coordinate system.
3. Image segmentation for multisensor-based robots
Aiming at the problems of complex surface background and various defect shapes and difficulty in segmentation of automobile stamping parts, a stamping part surface defect detection method is provided based on machine vision and deep learning, a PixelNet convolutional neural network is adopted to carry out image segmentation on surface defects to obtain defect detection, a PixelNet network structure diagram is shown in figure 6, pixel hierarchical sampling is adopted by PixelNet, sample diversity is increased in the batch random gradient descent updating process, a complex nonlinear predictor is introduced to improve the classification accuracy of pixels, the network carries out multi-scale feature extraction behind a VGG-16 basic network, meanwhile, the PixelNet establishes a predictor on the multi-scale features extracted by the multilayer convolutional neural network, and a 'supercolumn' descriptor is used for referring to the features extracted by a plurality of layers corresponding to the same pixel.
The model is trained by using a single GPU through a Matlab development interface of a deep learning framework Caffe under a Windows 7 operating system. The surface image of the stamping part is collected, and the defects are marked as scratches and rusts as shown in figure 7.
4. Robot constant force polishing based on multiple sensors
The robot constant-force polishing method based on the torque sensor reduces errors through a series of preprocessing on signals of the torque sensor, and then controls the robot to implement a constant-force polishing scheme through formulating a polishing strategy. The overall defect grinding process based on the torque sensor is shown in fig. 8. For the contact force information measured by the torque sensor, this information cannot be used directly. In addition to the zero drift of the torque sensor in the no-load state, the zero drift may also be affected by other factors, such as high-frequency noise, distortion point, etc. In order to eliminate the interference factors, the designed filter is a digital filter Butterworth low-pass filter; a force/position hybrid control algorithm is provided for realizing constant-force grinding of the robot, a force/position hybrid control algorithm diagram is shown in FIG. 9, and grinding contact force is controlled by converting control quantity into displacement of each shaft joint of the robot according to the force/position hybrid control algorithm. Contact force F acquired by torque sensorEWith a desired force FDThe difference between them is converted into joint position by inverse kinematicsAnd (4) moving amount to realize constant-force polishing.
5. Analysis of results
The contact force change during the grinding process is shown in fig. 10, and the surface pairs of the automobile stamping parts before and after grinding are shown in fig. 11, so that the grinding force is finally controlled to be about 15N and the grinding effect meets the actual machining requirement.
Claims (4)
1. A multi-sensor based robot polishing method comprises the following steps:
step 1: selecting the type of hardware of the robot polishing system, and calibrating the hardware composition of the vision sensor;
step 2: utilizing a vision sensor to collect defects on the surface of the automobile stamping part, manually marking the collected image data to obtain a true value diagram, and establishing a defect data set;
and step 3: an image segmentation algorithm is adopted, and a PixelNet convolutional neural network is utilized to complete the segmentation of the surface defects of the workpiece;
and 4, step 4: preprocessing signals acquired by a torque sensor;
and 5: and defect polishing based on a torque sensor, wherein a robot defect polishing system performs polishing operation on the defects by adopting an orthogonal path polishing scheme by using a force/position hybrid control algorithm.
2. The multi-sensor based robot polishing method of claim 1, wherein a multi-sensor based robot polishing system is designed, and the system is divided into two subsystems: the surface defect detection system of the stamping part based on the vision sensor and the defect grinding system based on the torque sensor, wherein the defect detection part in the step 2 is to detect the surface defects of the stamping part to be ground by using the vision sensor; the method comprises the steps that a defect detection path is set according to the surface shape of a stamping part, then an industrial robot moves according to the set path, meanwhile, a vision acquisition system acquires images of the surface of the stamping part according to set acquisition parameters, and a computer detects whether the surface of the stamping part has defects or not in real time according to the acquired images by using a defect detection algorithm.
3. The multi-sensor based robot grinding method of claim 1, wherein in the step 3, the workpiece surface defect is divided by using a PixelNet convolutional neural network for the first time, the initial learning rate is set to be 0.01, the initial learning rate is finally attenuated to be 0.0001 along with the increase of the number of training rounds, the PixelNet establishes a predictor on the multi-scale features extracted by the multi-layer convolutional neural network, and the features extracted by a plurality of layers corresponding to the same pixel are indicated by using a 'super column' descriptor.
4. The multi-sensor based robot polishing method of claim 1, wherein a multi-sensor based robot polishing system is designed, and the system is divided into two subsystems: the method comprises the following steps that 5, a stamping part surface defect detection system based on a visual sensor and a defect grinding system based on a torque sensor are adopted, grinding is started when a grinding tool reaches a corresponding grinding position based on the torque sensor, the contact force between the grinding tool and a workpiece is collected in real time by the torque sensor arranged between an execution arm of the industrial robot and the grinding tool, the collected contact force information is fed back to a computer, and the computer judges whether the terminal pose of the industrial robot is adjusted or not to ensure that the force between the grinding tool and the workpiece is relatively constant; an orthogonal path polishing scheme is designed during polishing path planning, and each point in a defect range is uniformly stressed, so that the polishing effect is guaranteed; in the aspect of grinding force/position control, a force/position hybrid control algorithm is provided for realizing constant-force grinding of the robot.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910236303.6A CN111823221A (en) | 2019-03-27 | 2019-03-27 | Robot polishing method based on multiple sensors |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910236303.6A CN111823221A (en) | 2019-03-27 | 2019-03-27 | Robot polishing method based on multiple sensors |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111823221A true CN111823221A (en) | 2020-10-27 |
Family
ID=72914132
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910236303.6A Pending CN111823221A (en) | 2019-03-27 | 2019-03-27 | Robot polishing method based on multiple sensors |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111823221A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113199491A (en) * | 2021-05-26 | 2021-08-03 | 广西大学 | Industrial robot constant-force polishing method |
| CN114770502A (en) * | 2022-04-25 | 2022-07-22 | 深圳市超准视觉科技有限公司 | Quick calibration method for tail end pose of mechanical arm tool |
| CN114833647A (en) * | 2022-05-25 | 2022-08-02 | 安徽安步轴承有限公司 | Grinding equipment and method for high-precision processing of bearing surface |
| CN115301573A (en) * | 2022-07-12 | 2022-11-08 | 成都合创视达科技有限公司 | Automatic detection device system for surface quality of automobile stamping parts |
| CN115847398A (en) * | 2022-11-09 | 2023-03-28 | 武汉理工大学 | Robot intelligent processing method and system based on region |
| CN116945195A (en) * | 2023-09-19 | 2023-10-27 | 成都飞机工业（集团）有限责任公司 | Omnidirectional measurement device system arrangement, registration method, electronic device and storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107052950A (en) * | 2017-05-25 | 2017-08-18 | 上海莫亭机器人科技有限公司 | A kind of complex-curved sanding and polishing system and method |
| CN107866717A (en) * | 2017-11-09 | 2018-04-03 | 四川工程职业技术学院 | A kind of turbine blade automatically grinding system |
| CN108908120A (en) * | 2018-08-07 | 2018-11-30 | 东南大学 | Robot grinding device and polishing process based on six-dimension force sensor and binocular vision |
| CN108942940A (en) * | 2018-08-01 | 2018-12-07 | 东南大学 | A kind of teleoperation robot polishing control system based on Multi-sensor Fusion |
| CN108972573A (en) * | 2018-06-12 | 2018-12-11 | 浙江大学 | A kind of industrial robot automation wheel hub polishing system and method based on characteristics of image identification |
| CN109118471A (en) * | 2018-06-26 | 2019-01-01 | 广东工业大学 | A kind of polishing workpiece, defect detection method suitable under complex environment |
-
2019
- 2019-03-27 CN CN201910236303.6A patent/CN111823221A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107052950A (en) * | 2017-05-25 | 2017-08-18 | 上海莫亭机器人科技有限公司 | A kind of complex-curved sanding and polishing system and method |
| CN107866717A (en) * | 2017-11-09 | 2018-04-03 | 四川工程职业技术学院 | A kind of turbine blade automatically grinding system |
| CN108972573A (en) * | 2018-06-12 | 2018-12-11 | 浙江大学 | A kind of industrial robot automation wheel hub polishing system and method based on characteristics of image identification |
| CN109118471A (en) * | 2018-06-26 | 2019-01-01 | 广东工业大学 | A kind of polishing workpiece, defect detection method suitable under complex environment |
| CN108942940A (en) * | 2018-08-01 | 2018-12-07 | 东南大学 | A kind of teleoperation robot polishing control system based on Multi-sensor Fusion |
| CN108908120A (en) * | 2018-08-07 | 2018-11-30 | 东南大学 | Robot grinding device and polishing process based on six-dimension force sensor and binocular vision |
Non-Patent Citations (1)
| Title |
|---|
| AMERICAN SCIENTIFIC RESEARCH JOURNAL FOR ENGINEERING, TECHNOLOGY: "Machine Vision Detection Method for Surface Defects of Automobile Stamping Parts", 《AMERICAN SCIENTIFIC RESEARCH JOURNAL FOR ENGINEERING, TECHNOLOGY, AND SCIENCES》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113199491A (en) * | 2021-05-26 | 2021-08-03 | 广西大学 | Industrial robot constant-force polishing method |
| CN114770502A (en) * | 2022-04-25 | 2022-07-22 | 深圳市超准视觉科技有限公司 | Quick calibration method for tail end pose of mechanical arm tool |
| CN114833647A (en) * | 2022-05-25 | 2022-08-02 | 安徽安步轴承有限公司 | Grinding equipment and method for high-precision processing of bearing surface |
| CN115301573A (en) * | 2022-07-12 | 2022-11-08 | 成都合创视达科技有限公司 | Automatic detection device system for surface quality of automobile stamping parts |
| CN115847398A (en) * | 2022-11-09 | 2023-03-28 | 武汉理工大学 | Robot intelligent processing method and system based on region |
| CN116945195A (en) * | 2023-09-19 | 2023-10-27 | 成都飞机工业（集团）有限责任公司 | Omnidirectional measurement device system arrangement, registration method, electronic device and storage medium |
| CN116945195B (en) * | 2023-09-19 | 2024-01-12 | 成都飞机工业（集团）有限责任公司 | Omnidirectional measurement device system arrangement, registration method, electronic device and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111823221A (en) | Robot polishing method based on multiple sensors | |
| CN110370286B (en) | Method for identifying rigid body space position of dead axle motion based on industrial robot and monocular camera | |
| CN109822574B (en) | Industrial robot end six-dimensional force sensor calibration method | |
| CN107367229B (en) | Free binocular stereo vision rotating shaft parameter calibration method | |
| CN110103217B (en) | Industrial robot hand-eye calibration method | |
| CN111486802B (en) | Rotating shaft calibration method based on self-adaptive distance weighting | |
| JP6180087B2 (en) | Information processing apparatus and information processing method | |
| CN111366084B (en) | Part size detection platform based on information fusion, detection method and fusion method | |
| CN111127568B (en) | Camera pose calibration method based on spatial point location information | |
| CN109015110B (en) | Machine tool rotary shaft position error modeling and identifying method | |
| KR20150129701A (en) | Self-contained holonomic tracking method and apparatus for non-destructive inspection | |
| JP2013036987A (en) | Information processing device and information processing method | |
| CN107817682A (en) | A kind of space manipulator on-orbit calibration method and system based on trick camera | |
| CN113870366B (en) | Calibration method and calibration system of three-dimensional scanning system based on pose sensor | |
| CN111862238A (en) | Full-space monocular light pen type vision measurement method | |
| CN109059755B (en) | High-precision hand-eye calibration method for robot | |
| CN113702384A (en) | Surface defect detection device, detection method and calibration method for rotary component | |
| CN113814870A (en) | Method for measuring and calculating pose of magnetorheological polished workpiece and polishing method | |
| CN111521204B (en) | Angular displacement visual measurement method based on absolute position rotary encoder | |
| CN110490934B (en) | Monocular camera and robot-based mixer vertical type blade attitude detection method | |
| CN114346538B (en) | Welding gun pose recognition method based on multi-axis linear Hall effect sensor | |
| CN113781581B (en) | Depth of field distortion model calibration method based on target loose attitude constraint | |
| CN110645916B (en) | Free-form surface measuring method and device based on reference calibration plate correction pose | |
| Liyanage et al. | A single time scale visual servoing system for a high speed SCARA type robotic arm | |
| CN114926531A (en) | Binocular vision based method and system for autonomously positioning welding line of workpiece under large visual field |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20201027 |
|
| WD01 | Invention patent application deemed withdrawn after publication |