KR20230128862A - Method and system for auto calibration of robot workcells - Google Patents

Abstract

A disclosed method for automatically calibrating a robot workcell includes the following steps of: generating a robot program for robotic work of a target robot workcell; collecting scanned point-cloud data of the target robot workcell using a mobile scanner; extracting geometric information of a three-dimensional design model of the target robot workcell; initially aligning objects of the target robot workcell by comparing the extracted three-dimensional design model with the scanned point-cloud data; matching coordinate systems of the extracted three-dimensional design model and the scanned point-cloud data; and adjusting a tool center position (TCP) of the target robot workcell and modifying the robot program. Therefore, the present invention is capable of improving the reusability of an existing robot control program and enabling remote robot maintenance.

Description

Robot work cell automatic calibration method and automatic calibration system {METHOD AND SYSTEM FOR AUTO CALIBRATION OF ROBOT WORKCELLS}

The present disclosure relates to an automatic calibration method and an automatic calibration system of a robot work cell.

Dimensional errors that occur during the installation/manufacturing of the robot work cell need to be precisely measured for autonomous calibration of the robot. That is, if the coordinate system of the virtual model and the coordinate system of the real environment are accurately matched, it is possible to easily identify dimensional errors of surrounding objects such as parts, jigs/fixtures, and the like. For this purpose, various methods for matching the design reference and local coordinate system to the real environment have been proposed.

As a method for matching the reference and local coordinate system in the design to the real environment, touch probe-based shape measurement is a method of tracking and matching the shape and position using a touch sensor and probe, and marker-based coordinate tracking is a method of matching marker tracking based on image vision. It is a method of matching by estimating world and local coordinates through In addition, the laser tracker-based photogrammetry method is a method of matching by tracking the location of an object using an image and a laser tracker, and the object recognition method based on 3D point-cloud matching is a method using a 3D laser scanner or a depth sensor. A method of collecting and matching points.

The various methods presented above can measure the posture and shape of one robot, but additional requirements such as installation of measuring equipment and calibration of the sensor itself occur. For example, when a camera is used, calibration of the camera itself, and in the case of a sensor attached to a robot arm, correction of an installation position error between the robot arm and the sensor are required. This hassle leads to the problem of accurately matching the reference coordinate system of the real environment and the reference coordinate system of the design model, especially when the object of the automatic assembly system is changed, such as when a new robot work cell is introduced or the robot is changed. There is a problem that requires resetting the measurement system in use.

One aspect of the embodiment is to provide an automatic correction method and an automatic correction system for a robot work cell that automatically corrects a pre-generated program of a work robot by automatically detecting a dimensional error generated during manufacturing and installation using a movable laser scanner.

However, the problems to be solved by the embodiments of the present invention are not limited to the above problems and can be variously extended within the scope of the technical idea included in the present invention.

A method for automatically calibrating a robot work cell according to an embodiment includes generating a robot program for robot work of a target robot work cell, scanning a scanned point-cloud of the target robot work cell using a mobile scanner. ) Collecting data, extracting geometric information of the 3D design model of the target robot work cell, and comparing the extracted 3D design model with the scanned point-cloud data to obtain an object of the target robot work cell Initial alignment, matching the extracted 3D design model and the coordinate system of the scanned point-cloud data, and adjusting the tool center position (TCP) of the target robot work cell and modifying the robot program include

The geometric information of the 3D design model of the target robot work cell may be reference point-cloud data extracted from the 3D design model.

The initial aligning of the object may include a process of extracting and utilizing an edge and a plane of the object from point-cloud data scanned by the mobile scanner.

In the process of extracting the edge and plane of the object, surface normal vectors of a random point and surrounding points are derived from the point-cloud data, and the derived surface normal vectors are spheres having the same origin. and selecting an average of the sum of the distances between a point on the sphere where the surface normal vector corresponding to the arbitrary point is located and the surface normal vectors corresponding to the surrounding points as the enhanced edge feature. there is.

The initial aligning of the object may further include searching for a position of the object by comparing the divided plane shape of the 3D design model with the divided plane shape in the scanned and collected point-cloud data. .

The initial aligning of the objects may further include defining a plane relationship by using surface normal vectors, relative distances, and angles of the extracted planes.

The target robot work cell includes a robot base, a robot arm, and a gripper and is configured to perform an operation of picking up a product, and the step of matching the coordinate system includes the scanned point-cloud data and the 3D design model. The method may further include aligning the gripper, estimating coordinates of the robot base, and estimating regional coordinates of the gripper and the product.

In the step of adjusting the TCP of the target robot workcell and modifying the robot program, the TCP of the target robot workcell is determined by using the transformation matrix of the gripper and the transformation matrix of the product derived in the step of matching the coordinate system. A change process may be further included.

The collecting of the scanned point-cloud data of the target robot workcell may include changing the location of the target robot workcell and scanning the target robot workcell multiple times, and matching the coordinate system may include performing the multiple scans. It may include being repeatedly performed as many times as one data number.

Meanwhile, a robot work cell automatic calibration system according to an embodiment includes a mobile scanner that scans a target robot work cell and collects scanned point-cloud data, and generates a robot program for robot work of the target robot work cell, , Extracting geometric information of the 3D design model of the target robot workcell, initial aligning objects of the target robot workcell through comparison of the extracted 3D design model and the actually scanned point-cloud data, and the extraction and a computer unit configured to match a coordinate system of the scanned 3D design model and the scanned point-cloud data, adjust a tool center position (TCP) of the target robot work cell, and correct the robot program.

The computer unit may be configured to extract and utilize an edge and a plane of the object from point-cloud data actually scanned by the mobile scanner.

The computer unit derives surface normal vectors of a random point and surrounding points from the point-cloud data in the process of extracting the edge and plane of the object, and the derived surface normal vectors have the same origin. The branch is expressed on a sphere, and the average of the sum of the distances between a point on the sphere where the surface normal vector corresponding to the arbitrary point is located and the surface normal vectors corresponding to the surrounding points is selected as the enhanced edge feature. It can be.

The computer unit may be configured to search for a location of the object by comparing the segmented planar shape of the 3D design model with the segmented planar shape in the scanned point-cloud data.

The computer unit may be configured to define a plane relationship by a surface normal vector, a relative distance, and an angle of the extracted planes.

The target robot work cell includes a robot base, a robot arm, and a gripper and is configured to perform an operation of picking up a product, and the computer unit aligns the scanned point-cloud data and the gripper of the 3D design model. and estimating coordinates of the robot base, and estimating local coordinates of the gripper and the product to match the coordinate system.

The computer unit may be configured to change the TCP of the target robot workcell by using the transformation matrix of the gripper and the transformation matrix of the product derived in the step of matching the coordinate systems.

According to the robot work cell automatic correction system according to the embodiment, it is possible to automatically correct a pre-generated program of the work robot by automatically detecting a dimensional error generated during manufacturing and installation using a movable laser scanner.

Since there is no need to fix the position of the laser scanner for point-cloud data collection and it can be used movably, the user does not require additional procedures for defining the relationship between the camera coordinate system and the world coordinate system.

It is possible to rapidly reconfigure the manufacturing system to utilize the robot work cell automatic calibration system for rapid calibration of the robot, uniform distribution of OLP-based robot control programs, and improved reusability of previously created robot control programs. Remote maintenance/repair of the robot is possible.

The burden on companies can be reduced by reducing the time and effort required for robot calibration work due to frequent environmental changes in the manufacturing system, such as new product production, manufacturing system layout change, and new assembly system development.

It can be applied to all industries where robots are used, and in the case of the robot-based service industry, which has recently been attracting attention, it is expected that robot re-teaching work will be required with a considerable frequency due to the very short release product cycle. A ripple effect is expected to occur.

1 is a diagram schematically illustrating a system for automatically correcting a robot work cell according to an exemplary embodiment.
2 is a diagram for explaining a method for automatically correcting a robot work cell according to an exemplary embodiment.
3 is a diagram illustrating a step of extracting edge and plane features for 3D point-cloud matching in a method for automatically correcting a robot work cell according to an embodiment.
FIG. 4 is a diagram illustrating a step of searching for a position of an object by comparing the similarity between the shape of each plane and the shape of the plane in scan data in the method for automatically correcting a robot work cell according to an embodiment.
5 shows a set of interplanar relationship features.
6 shows an initial transformation matrix for object alignment based on planar relationships.
7 is a diagram for explaining a coordinate registration process in stages for coordinate matching between a 3D design model, an actual scanned point, and a cloud in a method for automatically correcting a robot work cell according to an embodiment.
8 is a diagram for explaining robot TCP correction through object matching in a method for automatically correcting a robot work cell according to an embodiment.
9 is a diagram illustrating a configuration of robot programming and simulation software used in a method for automatically correcting a robot work cell according to an embodiment.
10 is a view showing a point-cloud data collection software screen used in a method for automatically correcting a robot work cell according to an embodiment.
11 is a diagram showing a feature extraction unit screen of software for automatically calibrating a robot work cell used in a method for automatically calibrating a robot work cell according to an embodiment.
12 is a view showing a scan scene analysis unit screen of the robot work cell automatic correction software.
13 is a diagram showing a result display screen of the robot work cell automatic calibration software.
14 is a diagram illustrating an experimental environment for performing a method for automatically correcting a robot work cell according to an embodiment.
15 is a view showing the result of correcting a gripping work robot program by performing a method of automatically correcting a robot work cell according to an embodiment.
16 is a simulation result showing gripping pressure distribution acting as a part during gripping work.
17 illustrates a result of correcting a robot program for a pointing task by performing a method of automatically correcting a robot work cell according to an embodiment.
18 is a configuration diagram schematically illustrating a system for automatically correcting a robot work cell according to an exemplary embodiment.

Hereinafter, embodiments of the present invention will be described in detail so that those skilled in the art can easily practice with reference to the accompanying drawings. In order to clearly describe the present invention in the drawings, parts irrelevant to the description are omitted, and the same reference numerals are assigned to the same or similar components throughout the specification. In addition, in the accompanying drawings, some components are exaggerated, omitted, or schematically illustrated, and the size of each component does not entirely reflect the actual size.

The accompanying drawings are only for easy understanding of the embodiments disclosed in this specification, and the technical idea disclosed in this specification is not limited by the accompanying drawings, and all changes and equivalents included in the spirit and technical scope of the present invention It should be understood to include water or substitutes.

Throughout the specification, the terms "comprise" or "have" are intended to indicate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, but that one or more other features are present. It should be understood that the presence or addition of numbers, steps, operations, components, parts, or combinations thereof is not precluded. Therefore, when a part "includes" a certain component, it means that it may further include other components, not excluding other components unless otherwise stated.

1 is a diagram schematically illustrating a system for automatically correcting a robot work cell according to an exemplary embodiment, and FIG. 2 is a diagram illustrating a method for automatically correcting a robot work cell according to an exemplary embodiment.

Referring to FIGS. 1 and 2 , the method for automatically correcting a robot work cell according to the present embodiment includes extraction of geometric features of the robot work cell, object alignment and 3D coordinate matching of the robot work cell, and robot TCP (Tool Center) Position) can be performed in steps of correction and update.

Specifically, a method for automatically calibrating a robot work cell includes generating a program for robot work in a virtual environment and collecting point-cloud data from a mobile laser scanner to calibrate the generated robot program. do. At this time, the creation of the robot job may be performed using robot programming and simulation software, and the collection of point-cloud data may be performed using point-cloud data collection software.

Then, the geometric information of the 3D (three-dimensional) design model is extracted using the robot work cell automatic calibration software, and the steps of initial alignment of the object through 3D model-scan data comparison, coordinate system matching, robot TCP adjustment, and program correction are performed. can be done

Hereinafter, each step is described in more detail.

3 is a diagram illustrating a step of extracting edge and plane features for 3D point-cloud matching in a method for automatically correcting a robot work cell according to an embodiment.

The various automation devices used in robotic workcells are usually designed through a combination of various aspects. For example, the shapes of a gripper for handling a product and a jig or fixture for fixing the product firmly support the product and at the same time prevent damage to the product surface by contacting the product. Parts are often designed flat. Based on this point, in this embodiment, the edges and planes of devices and products are distinguished from the collected point-cloud data and used for object alignment and coordinate matching.

The method of extracting the edge and plane of an object from point-cloud data is as follows.

First, surface normal vectors of an arbitrary point and surrounding points are derived.

Next, the derived surface normal vectors are expressed on a sphere having the same origin.

Next, an average of the sum of distances between a point on the sphere where the surface normal vector corresponding to the arbitrary point is located and surface normal vectors corresponding to neighboring points is selected as the enhanced edge feature. Here, the distance between the surface normal vectors is the distance through the surface of the sphere.

Compared to the case where the edge is represented by curvature based on covariance, the normal vector can be used to display the characteristics of the edge more prominently. After enhancing the features for all points in the point-cloud data, the feature values of each point are normalized to have a value between 0 and 1, so that planes and edges can be quantitatively distinguished. For example, if the experimentally normalized e _pi value is between 0.1 and 0.5, it can be classified as an edge, and if it is smaller than 0.1, it can be classified as a plane.

The preparation of the “enhanced edge” feature can be first calculated for 3D design models, and in particular, the plane and edge of the gripper can be distinguished. Then, when the point-cloud collection for the actual work cell is executed through the mobile scanner, the feature enhancement process is performed on the scanned point-cloud, and then the plane and the edge are separated and the reference point-cloud extracted from the 3D design model A matching algorithm using a planar relationship can be driven by comparing a reference point-cloud and a scanned point-cloud.

Figure 3 shows the result of extracting the edge and plane of the object from the 3D 　 design 　 model. That is, separated edges and separated surface planes extraction results were shown, and highlighted edges extraction results were also shown.

In order to obtain object edge and plane information from the 3D design model, the 3D design model is converted into a dense (approximately 1mm interval) mesh as a preprocessing process, and only point information is extracted from it and stored as a point-cloud data set. Then, the feature enhancement process is performed. In this way, reference point-cloud data can be extracted from the 3D design model.

In this way, the reference point-cloud data obtained from the 3D design model can then be compared with the edge and plane information identified in the scanned point-cloud data of the actual work cell to perform matching.

4 is a view to explain a step of searching for a location of an object by comparing the shape of each plane and the similarity of the plane shape in the scan data in the method for automatically correcting a robot work cell according to an embodiment; 5 shows a set of inter-plane relationship features, and FIG. 6 shows an initial transformation matrix for object alignment based on plane relationships.

In order to align the reference point-cloud data of the 3D design model in the actual scanned point-cloud data, it is first necessary to find the position of the object to be aligned. To this end, the approximate position of the object may be searched by comparing the shape of each divided plane shown in FIG. 3 with the divided plane shape in the scan data.

Referring to FIG. 4 , boundary information of a reference plane A may be extracted from a 3D design model (upper left corner), and boundary information of a scan plane B corresponding to the reference plane A may be extracted from scan data (upper right corner). Each of the extracted reference plane A and scan plane B has a surface normal vector, and a plane B' obtained by transforming the scan plane B based on these surface normal vectors can be derived. The transformed plane B' may be compared with the reference plane A to measure shape similarity.

If the face is divided using the enhanced features for both the reference data and the scanned data of the 3D design model, the most similar plane can be selected by measuring the shape similarity between the plane existing in the 3D design model and the plane existing in the scanned data. Thereafter, planar relationships within the maximum distance between faces of the target object are extracted based on the searched approximate object position. Each of the extracted planes has a surface normal vector, and their relationship can be defined by a relative distance and angle with the planes they are in contact with.

Referring to FIG. 5, when there are two arbitrary planes (plane1, plane2) constituting an object, the surface normal vector they have ( , ) and the distance vector between the surface normal vector ( ) can be found. At this time, the cross product vector of the two surface normal vectors ( ) and the angle between each surface normal vector ( , ) is determined.

The plane of the reference point-cloud obtained from the 3D design model and the actual scanned point-cloud by storing the surface normal vector pair, the cross product vector of the two surface normal vectors, and the interplane relation feature between the surface normal vector and the cross product vector. Initial sorting can be performed by comparing the .

That is, each plane in the reference data and actual scanned data of the selected 3D design model has a relationship with other planes in the point-cloud in which they are included. For example, a relationship between a similar surface of a 3D design model and another surface of the 3D design model, and a relationship between a similar surface of scan data and another surface of the scan data may be defined as defined in FIG. 5 .

In addition, an initial transformation matrix can be calculated as shown in FIG. 6 by comparing these relationships. If the search for similar faces fails, a comparison of all facet relationships can be performed.

In FIG. 6, if the normal vector, distance vector, and angle between the two faces in the 3D design model are defined, and the normal vector, distance vector, and angle between the two faces in the scan data are defined, the left three vectors in FIG. 6 ( ( ) ^ref ) to the right of the three vectors ( ( ) ^scan ) can be moved (transformation).

To this end, the 3D design model (ref) and scan data (scan) can be expressed as a feature vector matrix on the right. Here, through the process of singular value decomposition (SVD) (see (1) and (2) on the right side of FIG. 6), a rotation matrix R capable of rotating scan data into a 3D design model can be obtained.

After obtaining the rotation matrix, the difference between the distance vectors ( ) using a translation vector (translation vector, ) is obtained, and the final transformation matrix ^S T _S' can be composed of a combination of R (3x3 matrix) and T (3x1 matrix) as shown in (4), and this process is the process of obtaining the initial transformation matrix.

7 is a diagram for explaining a coordinate registration process in stages for coordinate matching between a 3D design model, an actual scanned point, and a cloud in a method for automatically correcting a robot work cell according to an embodiment.

Matching of 3D data may be performed by equalizing the reference coordinates of the 3D modeling CAD environment and the reference coordinates of the real environment, and performing matching of surrounding objects based on the reference coordinates of the 3D modeling CAD environment.

In this embodiment, since matching is performed only with the relative positional relationship of objects, additional work such as marker installation can be excluded without performing the definition of a laser scanner for 3D point-cloud data collection. To this end, a gripper that is in common contact with the robot and the product is first searched and matched, and then other objects at a relative position and distance may be matched.

Matching of coordinates may proceed in the order of i) aligning the gripper with the scanned data, ii) estimating the coordinates of the robot base, and iii) estimating the regional coordinates of the gripper and the product. Matching of these coordinates may be repeatedly performed as many times as the number of scanned data of the target work cell. When obtaining other scanned data, it can be scanned by changing the location of the robot.

Once coordinate matching is completed, the initial transformation matrix and robot base coordinates obtained at this time can be used as initial input values for the next scanned data. In addition, more precise coordinate estimation can be performed by considering the distance between points of each object.

8 is a diagram for explaining robot TCP correction through object matching in a method for automatically correcting a robot work cell according to an embodiment.

The robot's TCP can be expressed as [x, y, z, roll, pitch, yaw], which can be expressed as a transformation matrix combined with rotation and movement. Through this, the attitude transformation of the TCP is the transformation matrix of the gripper derived from the coordinate system matching process ( ) and the transformation matrix of the product ( ) can be used to change the robot TCP to fit the actual measured environment.

That is, the calibrated robot TCP ( ) can be calculated by the following Equation 1.

[Equation 1]

Hereinafter, an example of a system implementing a method for automatically guaranteeing a robot work cell according to the present embodiment will be described together with drawings.

9 is a diagram illustrating a configuration of robot programming and simulation software used in a method for automatically correcting a robot work cell according to an embodiment.

The robot programming and simulation SW can be implemented for generating a robot program driven in a robot work cell and simulating a calibrated robot program if necessary. The drive sequence may include i) interworking with the robot, ii) settings for robot control and simulation, iii) electric gripper operation settings, and iv) robot drive simulation and control program update.

10 is a view showing a point-cloud data collection software screen used in a method for automatically correcting a robot work cell according to an embodiment.

The first step to calibrate the robot program created through the robot programming and simulation software is point-cloud data collection of the actual robot work cell. Referring to FIG. 10, for this purpose, a point-cloud data collection SW compatible with a movable laser scanner (PhoXi® 3D Scanner) was implemented. The Photoneo PhoXi® Control API was utilized for laser scanner connection and data acquisition. The collected point-cloud data is stored in PLY format, and has a function of converting and storing mesh information of a 3D CAD model stored as an STL file into point-cloud data. The position of the laser scanner during point-cloud collection is not predefined.

11 is a view showing a screen of a feature extraction unit of software for automatically calibrating a robot work cell used in a method for automatically calibrating a robot work cell according to an embodiment, and FIG. 12 is a scan scene analysis unit of the automatic calibration software for a robot work cell. 13 is a diagram showing a screen of the result display unit of the robot work cell automatic calibration software.

As a core function of the robot work cell automatic calibration system, edge and plane information extraction from point-cloud data extracted from actual work cell collected and point-cloud data extracted from 3D CAD model, object initial alignment based on surface relationship, design-real coordinate system matching etc., and finally the robot TCP and program are modified. Then, the modified robot program is transmitted to the robot programming and simulation SW to update the new robot program to the controller.

Robot work cell automatic calibration SW i) Loads reference point-clouds of robots, grippers, products, etc., and extracts edge, plane, and relationship information between planes (Feature Extraction) unit (see FIG. 11), ii) Scan scene analysis unit for loading point-cloud data of actual robot work cells, inputting robot joint values corresponding to scanned scenes (acquired from OLP), aligning objects, and performing coordinate system matching (see FIG. 12) ), iii) It consists of a result display unit (see FIG. 13) for checking the modified robot TCP and control program and saving the control program.

14 is a diagram illustrating an experimental environment for performing a method for automatically correcting a robot work cell according to an embodiment.

The test robot work cell consists of a collaborative robot, a 2-finger electric gripper, a pressure sensor attached part, a turntable to generate dimensional errors, and a laser scanner for point-cloud collection of the work cell. The robot control program was created using robot programming and simulation SW, and the point-cloud of the robot work cell was secured using point-cloud data collection SW.

15 is a view showing the result of calibrating a gripping work robot program by performing a method for automatically correcting a robot work cell according to an embodiment, and FIG. 16 is a simulation result showing a gripping pressure distribution acting as a part during gripping work. am.

When operating the robot program generated through OLP, the space between the part and the gripper shows a difference as shown in the upper left of FIG. you can see what it did When the robot program is corrected through the robot work cell automatic correction SW, it can be confirmed that the gripper is positioned as originally planned, as shown on the right side of FIG. 15 .

16 shows the distribution of pressure generated when the gripper grips the product after calibration of the robot program, and it can be seen that the pressure is evenly distributed in the center of the two surfaces of the product.

17 illustrates a result of correcting a robot program for a pointing task by performing a method of automatically correcting a robot work cell according to an embodiment.

To verify the accuracy of the robot work cell automatic calibration system, a calibration experiment of the robot program for pointing work was performed. In the same environment as in FIG. 14, a pen capable of marking the gripper was mounted, and a program capable of marking 5 dots on the top of the product was created through OLP. When the planned program was immediately executed, the robot failed to take all the points, and after correcting the robot program, the results shown in FIG. 17 were shown. The actual stamped points show an average error of 0.0817mm from the planned points.

18 is a configuration diagram schematically illustrating a system for automatically correcting a robot work cell according to an exemplary embodiment.

Referring to FIG. 18 , the robot work cell automatic calibration system 100 according to the present embodiment includes a mobile scanner 120 and a computer unit 130 .

The mobile scanner 120 may be a mobile laser scanner, and is portable that can be moved freely and is not fixed to a specific location. The mobile scanner 120 scans the target robot work cell 50 to collect point-cloud data. In this case, the robot work cell 50 may include a robot base, a robot arm, and a gripper to perform a task of picking up a product.

The computer unit 130 generates a robot program for robot operation of the target robot work cell 50, extracts geometric information of the 3D design model of the target robot work cell 50, and converts the extracted 3D design model and the Objects of the target robot work cell 50 are initially aligned through comparison of actual scanned point-cloud data, matching the coordinate system of the extracted 3D design model and the actually scanned point-cloud data, and target robot work cell It may be configured to adjust the TCP (Tool Center Position) and correct the robot program. That is, the method of automatically calibrating the robot work cell described above with reference to FIGS. 1 to 17 may be performed through the computer unit 130 .

Software may include a computer program, code, instructions, or a combination of one or more of these, which may configure a processing device to operate as desired, or may independently or in combination direct a processing device. Software and/or data may be permanently or temporarily stored in any tangible machine, component, physical device, virtual device, computer storage medium or device for interpretation by a processing device or to provide instructions or data to a processing device. can be typified. Software may be distributed on networked computer systems and stored or executed in a distributed manner. Software and data may be stored on one or more computer readable media.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer readable medium.

Computer readable media may include program instructions, data files, data structures, etc. alone or in combination. The medium may be specially designed and constructed or may be known and usable to those skilled in computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical recording media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and ROMs, RAMs, flash memories, and the like. A hardware device specially configured to store and execute the same program instructions is included. Here, the medium may be a transmission medium such as light including a carrier wave for transmitting a signal specifying a program command, data structure, or the like, or a metal wire or a waveguide. Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter or the like as well as machine language codes such as those produced by a compiler. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

Although the preferred embodiments of the present invention have been described above, the present invention is not limited thereto, and various modifications and implementations are possible within the scope of the claims and the description and the accompanying drawings, and this is also the purpose of the present invention. It is natural to fall within the scope.

50: robot work cell
100: Robot work cell automatic calibration system
120: mobile scanner
130: computer unit

Claims (16)

generating a robot program for robot operation of a target robot work cell;
Collecting scanned point-cloud data of the target robot work cell using a mobile scanner;
extracting geometric information of a 3D design model of the target robot work cell;
initially aligning objects of the target robot work cell through comparison of the extracted 3D design model and the scanned point-cloud data;
Matching coordinate systems of the extracted 3D design model and the scanned point-cloud data; and
Adjusting the TCP (Tool Center Position) of the target robot work cell and correcting the robot program
Robot work cell automatic calibration method comprising a. According to claim 1,
The method of automatically correcting the robot work cell, wherein the geometric information of the 3D design model of the target robot work cell is reference point-cloud data extracted from the 3D design model. According to claim 1,
In the initial sorting of the objects,
A method for automatically correcting a robot work cell, comprising a process of extracting and utilizing an edge and a plane of the object from point-cloud data scanned by the mobile scanner. According to claim 3,
The process of extracting the edge and plane of the object,
Derive surface normal vectors of an arbitrary point and surrounding points in the point-cloud data;
The derived surface normal vectors are expressed on a sphere having the same origin,
Selecting the average of the sum of the distances between a point on the sphere where the surface normal vector corresponding to the arbitrary point is located and the surface normal vectors corresponding to the surrounding points is selected as the enhanced edge feature
A method for automatically calibrating a robot work cell, including a process. According to claim 3,
In the initial sorting of the objects,
Further comprising the step of searching for the position of the object by comparing the divided plane shapes of the 3D design model and the divided plane shapes in the scanned and collected point-cloud data. According to claim 3,
In the initial sorting of the objects,
The method of automatically correcting a robot work cell, further comprising defining a plane relationship by a surface normal vector, a relative distance, and an angle of the extracted planes. According to claim 1,
The target robot work cell includes a robot base, a robot arm, and a gripper, and is configured to perform a task of picking up a product,
The step of matching the coordinate system,
Aligning the gripper of the 3D design model with the scanned point-cloud data,
Estimating the coordinates of the robot base,
Estimating the local coordinates of the gripper and the product
A method for automatically calibrating a robot work cell, including a process. According to claim 7,
The step of adjusting the TCP of the target robot work cell and modifying the robot program,
and changing the TCP of the target robot workcell by using the transformation matrix of the gripper and the transformation matrix of the product derived in the matching of the coordinate systems. According to claim 1,
The step of collecting the scanned point-cloud data of the target robot work cell,
It is performed by changing the position of the target robot work cell and scanning it a plurality of times,
The step of matching the coordinate system,
A method for automatically correcting a robot work cell, comprising repeatedly performing as many times as the number of data scanned a plurality of times. a mobile scanner that scans a target robot work cell and collects scanned point-cloud data; and
Creating a robot program for robot operation of the target robot work cell, extracting geometric information of a 3D design model of the target robot work cell, and comparing the extracted 3D design model with the actually scanned point-cloud data The object of the target robot work cell is initially aligned, the coordinate system of the extracted 3D design model and the actually scanned point-cloud data is matched, the TCP (Tool Center Position) of the target robot work cell is adjusted, and the A computer unit configured to modify the robot program
Robotic work cell automatic calibration system comprising a. According to claim 10,
The computer unit is configured to extract and utilize an edge and a plane of the object from point-cloud data scanned by the mobile scanner. According to claim 11,
The computer unit is
In the process of extracting the edge and plane of the object, surface normal vectors of a random point and surrounding points are derived from the point-cloud data, and the derived surface normal vectors are a sphere having the same origin. And configured to select the average of the sum of the distances between a point on the sphere where the surface normal vector corresponding to the arbitrary point is located and the surface normal vectors corresponding to the surrounding points as the enhanced edge feature, the robot work cell automatically correction system. According to claim 11,
The computer unit is configured to search for the position of the object by comparing the divided plane shape of the 3D design model and the divided plane shape in the scanned point-cloud data. According to claim 11,
The computer unit is configured to define a plane relationship by a surface normal vector, a relative distance, and an angle of the extracted planes. According to claim 10,
The target robot work cell includes a robot base, a robot arm, and a gripper, and is configured to perform a task of picking up a product,
The computer unit is configured to align the coordinate system by aligning the gripper of the 3D design model with the scanned point-cloud data, estimating coordinates of the robot base, and estimating local coordinates of the gripper and the product. , robot work cell automatic calibration system. According to claim 15,
Wherein the computer unit is configured to change the TCP of the target robot workcell by using the transformation matrix of the gripper and the transformation matrix of the product derived in the step of matching the coordinate system.