JPH07122823B2 - Teaching and control method for robot / automatic machine with hand vision - Google Patents

Description

The present invention relates to a control method for a robot or an automatic machine,
In particular, the present invention relates to a teaching and controlling method of a robot / automatic machine with a fingertip vision, which is suitable for increasing the accuracy of a control method for correcting an operation path based on recognition data obtained by performing an image pickup / recognition process of a fingertip vision during operation.

[Prior Art] A conventional method for locating a robot / automatic machine on a work object by obtaining the position / orientation of the work object based on the recognition data obtained by recognizing the work object by hand vision is as follows. There is a way to explain. Note that this kind of visual robot and robot teaching method are described in, for example, JP-A-60-2.
It is described in JP-A-49583 and JP-A-60-262215.

FIG. 5 is an explanatory view showing a first example of a conventional positioning method for a robot with a visual sense of a hand. In Fig. 5, the TV is at your fingertips.
An example of a robot 2 with a camera (visual) 1 is shown. When positioning the hand of the robot 2 as in the field of view of the TV camera 1 add operation target 4, the vision coordinate system X _V -O _V -Y _V fixing the field of view of the TV camera 1 (visual images) 5th It will be in the position shown in the figure. Wherein a unique orthogonal coordinate system having the robot 2 (robot coordinate system) X _R -O _R -Y _R, the origin position O _V vision coordinate system is represented by a vector _1. Further, a vector ₂ representing the position of the work target 4 in the visual coordinate system is obtained by visual recognition processing. The vector ₂ is divided into coordinate axis components _2X and _2Y . These visual coordinate axis components
_By means of converting _2X , _2Y into robot coordinate axis components
_2X is converted to _2XX and _2XY components, _2Y is converted to _{2YX
Converted} to _2YY components. The _2XX and _2YX components are combined to obtain _2RX , and the _2XY and _2YY components are combined to obtain _2RY . These vectors _2RX and _2RY represent the robot coordinate axis components of vector ₂ . Thus the position in the robot coordinate system of the work object 4 is Motomari as long obtained vector ₁ and _2RX and _2Ry, thereby positioning the robot 2 hand to the position of the workpiece 4. The fifth
The coordinate system in the figure shows the relationship on a two-dimensional plane as a representative.

However, if one fixed coordinate system is set over the entire operation area of the robot 2 and the hand of the robot 2 is positioned according to the coordinate values as described above, it is difficult to improve the positioning accuracy. This is because the robot 2 has a mechanical error, and the error differs between individual robots. To match the actual position of the hand of the robot 2 with the theoretical calculated position within the allowable error, Requires a very detailed calibration process. Therefore, the following positioning method is adopted for the commercialized robot.

FIG. 6 is an explanatory view showing a second example of a conventional positioning method for a robot with a visual sense of hands. In Fig. 6, the size of the visual field of view for the assembly work of ordinary electronic devices is at most about 100 mm square, and a wider range can be recognized by moving the robot and setting multiple fields of view. Although it is possible, in practice, it is assumed that the position / orientation of the supplied work object is recognized in the visual field with a positional deviation error that is within one visual field. The conventional use example is in this range, and the reference work point is determined in advance in this partial area. That is, the robot is taught and stored as the reference work position P _W1 of either gripping or assembling the reference work object 3 in FIG. 6 with the robot hand, or performing another work. Next, the work object 3 is recognized as it is by the TV camera. The position (origin position O _V1 ) of the visual field (visual coordinate system X _V1 -O _V1 -Y _V1 ) at this time is represented by vector ₃ of the robot coordinate system X _R -O _R -Y _R , and the inclination of the visual field is θ ₁ . Further, the position vector ₄ of the reference work object 3 within this field of view is obtained, and these vectors ₃ and ₄ are stored. Next, the position (origin position O _V2 ) of the field of view (visual coordinate system X _V2- O _V2- Y _V2 ) set for recognizing the current work object 4 supplied during normal work is represented by vector _5. And the tilt of the field of view is θ ₂ .
Further, the current position vector ₆ of the work object 4 in this field of view is obtained by the recognition processing.

However, the position vectors ₃ and ₅ above are the robot coordinate system X _R -O
_In the case of a robot, it is a vector represented by _R −Y _R and represents a theoretical position calculated by a geometric operation based on the length of each arm and the displacement amount of each joint. On the other hand, as described with reference to FIG. 5, the robot or the automatic machine always has a mechanical error (for example, arm length error or axis eccentricity), so that the actual position has a position error. Generally, the positional deviation error due to this mechanical error increases in proportion to the distance, and the error is distributed by such an error function. Therefore, the larger the absolute value of the position vectors ₃ and ₅ , the larger the degree of error. Accordingly vector _{3, 5} of the difference vector ₅ - Request _3. However, the positions of both fields of view should be sufficiently close. Then, the absolute value of the difference vector ₇ is small, and therefore the error of the relative positional deviation is also small. One of the position vectors ₄ and ₆ is represented by the respective visual coordinate system, but it can be transformed into a vector represented by the robot coordinate system by means of coordinate transformation.
A method for obtaining this conversion means (conversion coefficient) is conventionally known. Vector by combining the vectors _{7, 4, 6} obtained by the _{8 =} - ₄ + ₇ + ₆ When seeking, the vector ₈ actual work object from the work point P _W1 criterion 4
_Represents a relative displacement amount (a two-dimensional displacement is shown in FIG. 6) up to a work position P _W2 at which the work is applied to. If this relative displacement amount is obtained, the robot hand is used as the actual work object 4
Reference point that is already stored when positioning
It is possible to move the position relative to P _W1 relative to the target point P _W2 .

[Problems to be Solved by the Invention] In the above-mentioned conventional technique, the reference work point P _W1 explained in the second example.
Are stored with the coordinate values of the robot coordinate system side and the visual coordinate system side respectively, and after that, when the actual work object 4 is recognized and the data of its position / orientation is obtained, the relative displacement amount is calculated and the target is calculated. Control the robot motion to point P _W2 . However, the reference position P _W1 and the position at which the actual work object 4 is imaged by the TV camera are limited to the fixed point designated by the teaching or the imaging target stop point corrected to the situation. Therefore, there is a problem in that the temporary stop for taking the image delays the time of a series of work operations of the robot. As a method of reducing such loss time, a method of capturing an image of a work object during operation can be considered.

Fig. 4 shows images taken during the operation of the above-mentioned robot with hand vision.
It is a schematic explanatory drawing of the processing method of recognizing and positioning. In FIG. 4, since the TV camera 1 is attached to the wrist of the robot 2 with an offset, if the control of the operation is devised (for example, the TV camera 1 is swung forward in the operation direction (the direction of the broken arrow)), the robot It can reach above the reference work object 3 ahead of the second hand. Thus based on the data of visual coordinate system X _V -O _V -Y _V was subjected to recognition by capturing this moment, the operation path toward the reference working point P _W1 of the robot 2 in operation (indicated by a broken line) R _The correction is added to ₁ to control the robot 2 to the movement path R ₂ for positioning the recognized actual work object 4 to the work position P _W2 . That is, the robot 2
If it is possible to calculate the position of the TV camera 1 at the moment when the image is picked up during the operation of, the method for calculating the control amount for positioning the robot 2 on the actual recognition target 4 is the same as the conventional method. Can be realized.

However, when calculating the position of the hand of the robot 2, that is, the position of the TV camera 1, a slight error occurs in the calculation result depending on whether the robot 2 is in the stopped state or in the operating state. That is, the above-mentioned calculation of the position is a geometrical calculation based on the angular displacement of each joint and the size of each arm of the robot. In this case, even if the angular displacement of each joint has the same value, The position of the hand of the robot 2 may be due to slight bending of each part of the mechanism in the operating state compared to the stopped state, or due to changes in the clearances of the bearing, harmonic drive, gears, etc. It is estimated that an error will occur in (the position of the TV camera 1). Therefore, if the reference work position P _W1 is recognized, the robot 2 is taught to be stopped and recognized, and the actual work target 4 is recognized during operation based on the recognition data during the stop. When the relative displacement amount of is calculated, there is a problem that the above-mentioned error is included in the calculation result.

An object of the present invention is to accurately position a hand or tool by reducing an error component caused by a motion generated in a process of positioning a work target based on data obtained by performing a hand vision imaging / recognition process during the motion. (EN) Provided is a teaching and controlling method for a robot / automatic machine with a hand vision.

[Means for Solving the Problem] The above-mentioned object is to achieve good reproducibility of the position on the operation path when the same command value is given to the robot / automatic machine and the operation is repeated. Focusing on the point that the above error contained in the relative displacement can be canceled by performing the operation under the same condition and performing the imaging / recognition even in the teaching process of recognizing the work target (position), first, as a teaching of the robot, a manual operation is performed. The robot is moved to an appropriate position in advance and the position data is recorded (registered) in the memory. Next, as a visual instruction, the reference work object is imaged and recognized, and the position data in the visual field is stored in the memory. At the same time as recording (registering), the position of the field of view (position of the TV camera) is also recorded, and the robot should work according to the data calculated and registered (recorded) by these teachings. Specifically, after performing a process of registering a work sequence that is indispensable as a pre-process of the teaching process with a robot language program, in the teaching process of the robot, a process of registering a main position on an operation path and a reference work target are performed. On the other hand, the process of registering the reference work position P _W1 by aligning the robot's hand (tool), and the reference work object being in the field of view by manually operating the robot, and greatly deviating from the position where the movement path was predicted An image is picked up under the same conditions as in normal use (playback) of the robot based on the registered data in the above-mentioned visual teaching process, by selecting a non-existing position and registering the imaging position Pv. The process of performing a motion before and after passing the position Pv, and performing the image pickup / recognition when the image pickup position Pv is passed during the movement of the robot, the reference work object is set. Identify and process for determining the working position P _W1 of the reference, and performs a process of registering data of the reference working position P _W1 recognized in the process, by the robot in accordance with data calculated and registered (recorded) by the teachings processing This is achieved by the teaching and control method of the robot / automatic machine with the visual sense of the hand that is designed to perform the actual work.

[Operation] In the teaching and control method of the above-mentioned robot / automatic machine with fingertip vision, the robot teaching process manually positions the robot at a point of operation and stores the position in advance by the conventional method. The teaching of the robot with the mark requires not only to memorize the essential points of the robot's operation but also to visually recognize the reference position P _W1 and store the reference recognition data. In order to execute the imaging / recognition process during operation, the process of programming the sequence of the operation by teaching the essential points of the robot operation and the robot language is performed first, and the visual imaging position Pv is specified on the program. Declare visual instruction processing. After that, the visual teaching process performs an operation by executing the program and performs image pickup / recognition at the designated position Pv, but when the reference work object is placed at the predetermined position _PW1 in advance, the robot operation is performed at the image pickup position Pv.
You can reproduce the operation before and after the same as in normal work,
The data recognized at this time is registered as the reference position data P _W1 . The recognition data is handled by the method shown in the second example of the positioning method for the robot with a fingertip vision in FIG. 6, and the reference position data P _W1 includes an error component due to the robot motion. Then, when an actual work is performed by the robot according to the data by the teaching process, the error component due to the motion of the same tendency is also included in the position data recognized by imaging the current work target during the motion under the same condition at the time of the work. Therefore, if the relative displacement vector of both data is calculated there, the error component due to the movement can be canceled and the positioning accuracy of the robot can be improved.

[Embodiment] An embodiment of the present invention will be described below with reference to FIGS.
It will be explained with reference to FIGS.

FIG. 1 is a flowchart of a teaching processing procedure of a robot with a fingertip vision showing an embodiment of a teaching and control method for a robot / automatic machine with a fingertip vision according to the present invention. In FIG. 1, a process 101 is a process for registering a work sequence command in a robot language program, which is the same as the conventional procedure, and this process is a process different from the teaching process, but the visual teaching later. Since the robot will be moved by the process, it is included as a pre-process indispensable. A process 102 is a process by a conventional teaching method of teaching and registering a main position on an operation path, and conventionally teaching a robot that manually positions at a main point of the operation of the robot and stores the position in advance. In the mode, when the robot is to be operated later, the robot is controlled by connecting the taught points with a path by an interpolating method and following the path. The process 103 is a process of registering the reference work position P _W1 by aligning the robot hand (tool) with respect to the reference work object. in addition to storing the point, the process of storing the reference position P _W1 imaging · recognize and to advance at a reference work object in position P _W1 is required. Process 104 is to manually operate the robot so that the reference work object is within the field of view, and select a position that does not greatly deviate from the position where the motion path is predicted, and select the imaging position.
This is a process of registering as Pv, and the robot is controlled so that this imaging position Pv becomes an intermediate point on the movement path when the robot is moved later. The above processes 102 to 104 are the procedure of the teaching process of the robot. The next process 105 is a process for causing the robot to perform the same operation as during normal use (at the time of playback) based on the data registered in the above-mentioned processes 101 to 104 for the purpose of visual teaching. Only the movement before and after the vision passes the image pickup position Pv where the image is picked up is necessary, but if it is partially limited too much, it may not be the same condition as the operation during normal work, so continuous operation within the range possible. Try to do it. Process 106 is the above process 1
During 05 of the robot operation is an imaging-process of recognizing working position P _W1 of criteria when passing the imaging position Pv, determine the working position P _W1 of the reference so as to recognize the work object of this again reference. The process 107 is a process of registering the data of the reference work position P _W1 obtained by the recognition in the process 106.
The visual teaching process from 05 to 107 is completed.

The teaching processing of the above processing 101 to 107 can be regarded as a reference work when the robot is used, and the work by the robot is performed according to the data calculated and registered (recorded) by this teaching processing. At this time, the reference position data obtained by reproducing the motions before and after the imaging position of the robot in the visual teaching in the same way as during the work include an error component due to the robot motion, and the work target is the same during the work. Since the position data imaged and recognized during the operation of the condition also includes the error component due to the motion of the same tendency, if the relative displacement vector of both data is calculated by the method shown in FIG. The positioning accuracy of the robot can be improved by canceling out. Although the content of each item of the processing procedure has been described above, the order of the items of the above-described processing 101 to 104 does not necessarily have to be as shown in the figure, and the order can be replaced and reprocessing is also possible. The items of the next processing 105 to 107 are always succeeded. The method is applicable to general automatic machines as well.

The position data stored in the robot teaching process described above is labeled for identification, such as P ₁ and P ₂ . The above-mentioned position data is data of the angular displacement of each joint of the robot, and the position of the hand of the robot can be obtained by geometrical calculation based on this data and the dimension values of each part of the robot. The work sequence command registration processing 101 described above describes the work sequence of the robot in the robot language. For example, if you enter the operation command MOVE I, P ₁ ,
The point you taught the robot from its current position in advance P ₁
Up to the parameter I is specified to move along the interpolation path. The method of writing a program that represents an operation sequence in the robot language and the method of recording the main position of the robot by teaching may be in accordance with the conventional method. However, since the process of visually recognizing and the process of visually teaching during the above operation are newly added by the method according to the present invention, a language specification example describing these processes and an example of the process will be described below. First, there are the following three new items in the language specification.

(1) A label Pv that clearly indicates the imaging target position is provided.

(2) When the imaging target position is entered in the parameter of the motion command (MOVE), it is defined that the imaging / recognition process is started when the robot motion passes the imaging target position Pv.

(3) A command (VTEACH) for instructing visual teaching processing is newly provided.

Next, a processing example will be described. The reference work position P _{W1 in} the process 103 of FIG. 1 represents the position where the hand is first positioned when working with the robot hand. The reference position P _W1 is a position set to teach the position to the robot, and the actual work position P _W2 is a position having an error from the reference position P _W1 . This error is calculated from the visual recognition data, and the robot is positioned at a position shifted from the reference position P _W1 by the amount of the error to perform the work. This method is in accordance with the conventional method for controlling a robot with a vision. The process 104 shown in FIG. 1 is a teaching required for a robot with a visual hand attached to it. This is a TV camera attached to the hand to manually find a position where the work object is viewed (in the visual field), and an appropriate image is picked up. Teach the position Pv. Here, the conventional robot with a visual sense of the hand moves to the image pickup position Pv to recognize the work object, and then stops the image pickup / recognition, whereas the present invention aims at the robot on the image pickup position Pv. This is a method of capturing and recognizing the moment when the vehicle passes without stopping. The state of the hand of the robot at the moment when it passes over this imaging position Pv
The robot is controlled so that the TV camera has the same attitude as when the imaging position Pv was taught. The method for controlling the movement of the robot so that it passes over the imaging position Pv is based on the conventional method.
Setting by the teaching process is also necessary to check the visual recognition condition.

In the normal teaching process, the robot is manually moved to teach the main points (positions) and work positions on the work route. This operation is performed in the teaching process such as the process 102 in FIG. However, the different point is that the robot performs the work operation in the visual teaching process thereafter. That is, the process 105 in FIG. 1 represents a process for executing a motion of the robot for visual teaching, and the motion may pass through the imaging position Pv at the same speed and route as during actual work (during playback). It is a condition.
Therefore, in the visual teaching process, a program that uses the same operation command (robot language) as when working is executed,
Move on the imaging position Pv under the same conditions as when working. And the first
In the process 106 in the figure, the reference work object (the same as the reference work object used in the item of the process 103) is imaged at the moment when it passes over the imaging position Pv, and the reference work position P _W1 is recognized, In the next process 107, the reference work position P _W1 recognized in process 106 is registered, and the visual teaching process ends. The above is the visual teaching process, but the route in which the robot needs to move is only before and after passing the imaging position Pv. Therefore, the robot language program executed during the teaching process is the imaging position Pv.
An operation command to pass above, an operation command before and after it, an instruction to receive the recognition data of the reference work position P _W1 after the motion, and the data representing the reference work position P _W1 from the visual recognition data and the visual field It is assumed to include a command for instructing visual teaching processing for calculating and registering position data.

2 (a) and 2 (b) are explanatory diagrams of an example of the visual teaching processing program and the operation path of FIG. Figure 2 (a),
In (b), PROGRAM TEAC of (1) in FIG. 2 (a)
Define the program identification name TEACH in H, and (2) SPEED8
00 defines the operation speed as 800 mm / sec. The next MOVE is a motion command, and the first parameter I specifies to operate by calculating a linear path. The current position (the position where the robot is currently stopped) is determined by MOVE I, P1 in (3). )
To the point P1 of the second parameter (FIG. 2 (b)), it moves in a straight path and stops at the point P1. Then (4) MOVE T, Pv
The straight line path from the point P1 to the point Pv (Fig. 2 (b)) is calculated according to Second
A path that is continuous with the path leading to the diagram (b)) is obtained, that is, a continuous path from the point P1 to the point P2 via the point Pv is calculated to move the robot. Here, since the point Pv represents the imaging position, when the robot passes over the imaging position Pv, the robot sends an imaging command to the visual sense, and the visual sense follows the reference work position P _W1 (Fig. 2 (b)). ) Image pickup / recognition processing is performed. Further, since the first parameter of (5) is also T, the path from point Pv to point P2 and the path from point P2 to _PW represented by the MOVE command of (6) described below are continuous (smooth path). The solid line in FIG. 2B) is calculated. Here, the second parameter of (6) is P _W1 + <0,0, -10>, which is -10 mm in the Z direction at the taught reference work position P _W1.
Represents the position where the shift is given (the position 10 mm vertically above). The third parameter, S = 200, has a vertical speed of 2
00 mm / sec is specified (the first parameter is I). Therefore, the robot moves from point P1 to point P _W1 + <0,0, -1 in Fig. 6 (b).
It moves continuously to 0> and stops at point P _W1 + <0,0, -10>. Next, the visual recognition data is received by the RECV command of (7), the visual teaching process is executed by the VTEACH command of (8), and the process ends with END of (9). The robot operation here is the same as a part of the operation during the actual work (during playback).

FIG. 3 is a block diagram of a control device for performing the teaching and control processing of FIG. In FIG. 3, reference numeral 5 is a robot controller, which has a multi-CPU configuration in this example. Of these, CPU3 (18) is dedicated to serial communication control. Reference numeral 6 is an image processing device (visual). A program written in robot language (sequence program) is input from the operation panel 7 and stored in the memory 1 (9).
In the teaching process of the robot 2, after the teaching box 8 is operated to position the robot 2 and then teaching of the position is instructed, the servo counter 14 of the robot controller 5 reads the joint angle indicated by the pulse encoder 25 of the robot 2. It is stored in the memory 9. Here, the robot language program is started by a key input on the operation panel 7, and the CPU 1 (10) of the robot control device 5 sequentially reads out the corresponding program from the memory 9 one by one and interprets it, and prepares for each instruction. The executed processing program. Of these, the operation instruction is to read the position data of the parameter from the memory 9, then temporarily store it in the memory 2 (11) and execute it sequentially by the CPU 2 (12). In the processing, the robot 2 is operated by calculating an operation path between teaching points (position data) by a specified interpolation method, and following the calculated path data every predetermined sampling time (for example, 10 msec).
To control. The control command value controls the servo motor 26 of the robot 2 via the servo amplifier 15 and the power amplifier 16. At this time, when the robot 2 reaches the imaging position (imaging target point) Pv on the operation path (position of the TV camera 1 in FIG. 4), the CPU 12 issues an imaging start command to the image processing device (visual) 302. It transmits via the parallel I / O interfaces 14 and 20. When the image processing device 6 receives this command signal, its camera control unit 21 controls the TV camera (visual) 1 to
The image data captured by the camera 1 is input to the image memory 22, the CPU 24 reads it and performs recognition processing, and sends the recognition result to the robot controller 5 via the serial interfaces 23 and 19. The CPU 18 dedicated to communication control receives the recognition data sent from the image processing device 6 and stores it in the memory 3 (17). Next, the robot control device 5 may use this recognition data to perform visual teaching processing, or to position the robot on a work target and perform work. Here, the teaching process relating to the teaching method of the present invention will be described below.

When the CPU 12 of the robot control device 5 executes the visual teaching program shown in FIG. 2 (a), the robot 2 moves on the path passing through the point P1 shown by the solid line in FIG. An image pickup start command is sent at the point of passing. Immediately after sending this imaging start command, the CPU 12 issues a command to the servo counter 14 to count the pulses of the pulse encoder 25 of each joint and latch them in the register. The number of encoder pulses represents the joint angle displacement of each joint of the robot 2 when the imaging command is issued, and if geometric calculation is performed based on these joint angle displacements, the robot 2 when the imaging command is issued.
It is possible to obtain the position of the hand of the player, that is, the position of the TV camera (visual) 1. After instructing the latch of the encoder pulse, the CPU 12 controls the normal servo motor 26, and thereafter reads the latched encoder pulse number from the register of the servo counter 14 and records it in the memory 11. Then, the CPU 12 executes the visual teaching process by the RECV command ((7) in FIG. 2A) after the operation of the robot 2 is completed. This processing is executed after the CPU 18 receives the visual recognition data and stores it in the memory 17. From the number of encoder pulses in the memory 11, the position of the TV camera 1, that is, the position of the visual field at the moment when the image is picked up is obtained. This position is represented by robot coordinate system vector ₃ (FIG. 6). Note that FIG. 6 shows the positional relationship in the horizontal plane. Here, an example is shown in which the TV camera 1 is capturing an image in the vertical direction. Then, the visual recognition data stored in the memory 17 is read out and the vector ₄ representing the position within the visual field is obtained in the robot coordinate system. These vectors ₃ and ₄ and the inclination θ ₁ of the visual field are stored in the memory 9, and the visual teaching process is completed. The position 3 (FIG. 6) visually recognized in this process is the reference work position P _W1 .

Next, when the robot 2 is to perform a work after the above teaching process, a robot language program (already registered) in which a work sequence is described is started. The robot 2 passes over the image pickup position Pv during operation, and the robot control device 5 similarly sends an image pickup command to the image processing device (visual) 6.
The image processing device 6 picks up an image with the TV camera (visual) 1 to recognize the work position, and recognizes the recognition data.
Send to. The work position P _{W2 at} this time is the reference work position P
There is an error with _W1 . While the CPU 12 of the robot control device 5 controls the operation of the robot 2, it monitors the CPU 18 for receiving the recognition data of the image processing device 6 and storing it in the memory 17 after transmitting the imaging command. Then, immediately after being stored in the memory 17, the position vector ₅ (FIG. 6) of the visual field at the time of imaging and the position vector ₆ (6th position) of the current work object 4 are read together with the number of encoder pulses stored in the memory 17. Figure) is calculated. Current working position P
_W2 is represented by a relative displacement vector ₈ based on the working position P _W1 of the reference, the vector _{8 =} - ₄ + ₅ -
It can be calculated by ₃ + ₆ (Fig. 6). From this, the CPU 12 reads out ₃ and ₄ stored in the teaching process from the memory 9 to calculate the vector ₈ , positions the hand of the robot 2 at the current work position P _W2 , and then performs work.

As described above, the teaching method for the robot / automatic machine with the visual sense of the hand according to the present invention is a method of registering the stop position after the conventional manual positioning of the robot and recognizing and recording the visual teaching at the stop position. Instead of performing the same operation (at the time of playback) (however, only the operation before and after the imaging position), teaching is performed based on the data taken and recognized under the same conditions. There is. The reason for implementing the teaching method according to the present invention is that if the method of imaging / recognizing during operation as described above is adopted, the absolute position at the moment of imaging cannot be accurately determined, that is, the displacement of each joint at the moment of imaging. When the position of the TV camera (the position of the field of view) is calculated based on this displacement and an error is generated, an error that depends on the operation speed and the operation direction occurs. The cause of this error is the mechanism error of the robot or the error of the displacement measurement system. However, there are many unclear points.Therefore, it is assumed that these errors will be reproduced at the same point when the robot is moved in the same condition speed and path, and the actual measurement results. This is also because it was found that the reproducibility of the error exists. Therefore, in the teaching process according to the present invention, the reference work position is recognized during the operation under the same condition, and the recognized work position is recognized. By obtaining the relative displacement between the working positions of the reference, it is possible to offset the inherent errors due to the operation can be improved positioning accuracy, it can be applied to an automatic machine robot or general.

[Effect of the Invention] According to the present invention, an error component caused by an operation that occurs in a process of positioning a work object according to data captured and recognized during operation of a robot with a visual sense of a hand / automatic machine is reduced. Accurate positioning,
From the result of the measurement experiment of the positioning accuracy of the robot, it is expected that the method of teaching by performing this operation can improve the positioning accuracy, for example, more than twice as compared with the conventional stationary teaching method.

[Brief description of drawings]

FIG. 1 is a flowchart showing the teaching process of a robot with a fingertip vision showing an embodiment of a teaching and control method for a robot / automatic machine with a fingertip vision according to the present invention, FIG. 2 (a),
(B) is an explanatory view of an example of a visual teaching processing program and an operation path of FIG. 1, FIG. 3 is a block diagram of a control device for performing the teaching and control processing of FIG. 1, and FIG. FIG. 5 is an explanatory diagram showing an outline of a processing method for imaging / recognizing and positioning during operation, FIG. 5 is an explanatory diagram showing a first example of a positioning method for a conventional robot with a hand tip, and FIG. 6 is a diagram for a conventional robot with a hand tip. It is explanatory drawing which shows the 2nd example of the positioning method. 101 …… Registration of work sequence command (robot language program), 102 …… Registration of main position of movement path, 103 ……
Registration of the reference work position P _W1 , 104 ... Imaging position of the work target
Registration of Pv, 105 ... Reference work position P after passing through the imaging position Pv
Execution of robot operation leading to _W1, 106 imaging and recognition of the working position P _W1 criteria ...... imaging position Pv during the passage, 107 the recognition result registration of the ...... reference working position P _W1 (visual teaching), 1 ... … TV
Camera (visual), 2 ... Robot, 3 ... Reference work object, 4 ... Actual work object, 5 ... Robot control device, 6 ... Image processing device (visual), 7 ... Operation panel , 8 ・ ・ ・ Teaching box, Pv …… Imaging position, P _W1 …… Reference work position, P _W2 …… Actual work position.

Claims (2)

[Claims] 1. In a robot or an automatic machine controlled by a program of a robot or an automatic machine having a hand vision, the position / orientation of a work object placed in a work area is visually imaged and recognized, and the recognition is performed. When performing control to position the hand or work tool according to the data, as a preparatory step, work to register the operation sequence of the robot / automatic machine with a program such as robot language, and teach the main position on the operation path After performing preparatory operations such as work and manually aligning the robot or automatic machine hand or work tool with the reference work object placed at the reference position and registering the position as the reference work position As a visual instruction, after manually registering and registering the imaging target position for imaging, the robot / automatic machine is passed through. Reproduce the motion before and after the imaging target position under the same conditions as when using playback, and visually recognize the position / orientation of the reference work object when the imaging target position is passed during the motion by visualizing with a fingertip. A method of teaching a robot / automatic machine with a fingertip vision, characterized in that the recognition result of the reference and the data of the reference visual position are registered. 2. When the robot or the automatic machine is played back after performing the teaching process of the robot / automatic machine with the hand tip according to claim 1, the current supply in the work area is performed. By recognizing a work object by imaging with a fingertip during operation and performing recognition, and obtaining a relative displacement amount between the recognition result and visual position data and the reference recognition result and reference visual position data, the current A robot with a hand tip that is characterized by determining the relative displacement of the work object from the reference work object, and performing control to position-correct the robot / automatic machine hand or work tool on the current work object. Automatic machine control method.