AT406313B - Flexible sensor system for industrial robots - Google Patents

Abstract

The invention relates to a method which has the object of providing the robot having an external sensor system 2 with the point of incidence 5 of the sensor in the form of the absolute robot coordinates, although the distance sensor 2 is mounted in any desired way on the robot and the robot knows neither the mounting position, nor the mounting direction, nor the factor length unit per measured value. The offset between sensor zero point 14 and the TCP 3, the measuring direction of the sensor 4 and the factor length unit per measured value are determined by means of two search excursions over a reference workpiece which has two edges, and stored in the computer 9 as coordinates in the TCP coordinate system. This information is loaded when the measured value from the sensor is transformed into the absolute coordinates in the robot coordinate system of the point of incidence 5, and is sent to the robot via the serial interface 11. If the system is employed dynamically, that is to say a search excursion over a workpiece to be measured is programmed, then the result is a topographic image of the workpiece in the robot coordinate system, and the robot can be sent process-relevant values, such as gap, length, burr shape etc., in the form of robot coordinates which can easily be processed by the robot. <IMAGE>

Description

AT 406 313 B

FLEXIBLE SENSOR SYSTEM FOR INDUSTRIAL ROBOTS

This sensor system is used in the field of industrial robots and also affects welding robots and flame cutting robots and is flexible with regard to the assembly and the type of distance sensor. This system, consisting of a distance measuring head, an evaluation computer and an interface to the robot control, is not aware of the zero point of the distance measuring head in the robot coordinate system, the mounting angle and the length unit factor per output value (current or voltage) and is intended as an external system without knowledge of the robot geometry or number of axes.

The current sensor systems, roughly classified, are: - Simple systems for localizing the workpiece, which cannot take absolute measurements on the workpiece. These processes are carried out by the robot controller, which has stored all information relating to the geometry and the kinematics of the robot. - Laser scanner cameras, which are mounted parallel to the welding gun and only refer to the TCP with the Χ, Υ Z offset. This offset has to be determined by manual measurement and is time-consuming. These can make absolute measurements on the workpiece and z. B. track the seam during welding and thus correct the distortion, also determine process-relevant measurements such as seam volume. These systems have their own external control, which is tied to the type of sensor head and which the information regarding sensor head mounting and some robot data must be entered manually.

The only systems currently available that can take absolute measurements are the laser scanner cameras. In most applications, including welding, the workpiece does not warp during the process. A cross between a system that can only localize and a system that can also make absolute measurements on the workpiece would be required here. Since such a system is not necessarily, in contrast to a scanner camera, parallel to the z. B. welding gun of the robot must be installed, methods should be found which allow the sensor to be able to be attached to the robot at any point and in any orientation and search runs free of restrictions, such as programming the orientation of the axis, and still to be able to carry out absolute measurements. An external flexible system must know the sensor parameters length unit per measured value, offset vector sensor zero point to TCP and the measurement direction. Since there are seven parameters here (xyz offset, measuring direction in space and length unit per measured value) and not, as with a scanner camera mounted in parallel, only three values (xyz offset), an automatic determination is necessary, since a manual determination with a lot Time and expensive teaching is not useful. This procedure automatically determines these parameters in 30 seconds.

The determination of the sensor parameters only has to be carried out once, or after a new assembly or a robot collision. The determination of the sensor parameters can be automated if the robot is programmed so that it processes the program for determining the sensor parameters after each shift, for example, and thus compensates for long-term tolerances of the sensors and AD converters. The determination of the sensor parameters takes about 30s. If the sensor parameters have to be determined manually, as with the currently known systems, a time expenditure of about 6 hours can be expected and must be carried out by qualified personnel, since seven parameters have to be determined here. For systems that are not so flexible in the assembly and selection of the sensor head, only three parameters, x y z offset, need to be determined, which requires a reasonable amount of time for manual determination. Furthermore, there is currently no sensor that can carry out process-relevant measurements in a confined space, since the laser scanner cameras are too large for such tasks and too inflexible in assembly.

Patent WO 94/04968 A1 describes a method of how a laser flame cutter is moved over the surface of the workpiece and the angle of attack can be changed as desired. The coordinates of the focal beam focus are determined manually and transformed into the robot coordinate system via the angles of the 4th and 5th axes. The linear axes become 2

AT 406 313 B controlled that the point of incidence of the cutting beam focus remains constant in the event of a change in position, and thus a constant distance between the torch nozzle and the workpiece is maintained. A new coordinate system, the origin of which is the focus, is introduced and the cutting program is programmed in this new coordinate system. In this way the orientation of the workpiece can be compensated. The method now presented differs from the method described in patent WO 94/04968 in the following points:

Task:

The method described in patent WO 94/04968 A1 determines the orientation of the workpiece and keeps the distance constant during the process.

The method now presented localizes a workpiece and compensates for the orientation and the displacement. Process-relevant values such as seam length, gap, ridge shape, etc. are also measured.

Type of sensor head:

The method described in patent WO 94/04968 A1 requires a sensor head whose measuring axis extends in the extension of the last robot axis.

The method now presented does not have this limitation. The measuring axis can run in any orientation to extend the axis on which the sensor is mounted. Reverse transformation:

The method described in patent WO 94 / 04968A1 uses the axis angles of the 4th and 5th axes to transform the focus and the program points into the robot coordinate system. Furthermore, to compensate for the orientation of the workpiece, a new coordinate system, the origin of which is in focus, must be defined manually.

The method now presented uses the rotation matrix of the TCP coordinate system, which is available with every standard robot, the sensor zero point and the orientation of the measuring line for the back transformation. The method now presented does not require a coordinate system to be defined specifically.

Determination of the sensor:

The method described in patent WO 94/04968 A1 requires a manual determination of the focus.

The method now presented automatically determines all the sensor parameters required for the reverse transformation.

The patent GB 2 232 504 A describes methods for increasing the programming convenience of an industrial robot by introducing new coordinate systems and the movement of the robot in these new coordinate systems. The points to be moved back are transformed using the rotation matrix of the new coordinate system. The calculation takes place in the robot controller, which knows all information regarding the robot geometry and kinematics. In addition to the orientation of the measuring line and the coordinates of the sensor zero point and other calculations, the method now presented also uses the rotation matrix of the TCP coordinate system, as described in patent GB 2 232 504 A. With this calculation method, however, completely different goals are pursued as described in patent GB 2 232 504 A. Furthermore, no protection is sought for the type of calculation

FIGURE OVERVIEW

1 shows the overall system, consisting of a distance sensor (2), mounted on the last robot axis with a welding gun (1) as an example, a sensor evaluation unit (7) which outputs an analog voltage proportional to the distance to an A / D converter (10) , a computer (9) for the mathematical evaluation and communication with the robot computer (12) via the serial interface (11) and a digital output (46) of the robot computer by means of a digital I / O interface (45). 3rd

AT 406 313 B

2 shows the distance sensor (2) with the vectors (13) (15) necessary for calculating the intersection (5) of the measuring direction (4) with the workpiece surface (16) and the zero point (14) of the sensor.

3 shows the process of determining the sensor assembly by means of search runs over the edges (33) of a reference workpiece. The TCP (3) moves on the straight lines (23) and (24) and the laser beam scans the workpiece along the lines (17) and (18).

4 shows the vectors for the calculation of the sensor assembly after the search run (18)

5 shows the vectors for the calculation of the sensor assembly after the search run (17)

6 shows the graphic representation of the sensor measured values and the digital one

Robot output on the screen (37) of the PC (9).

7 shows the graphic representation of the sensor measured values and the digital one

Robot output on the screen (37) of the PC (9) after a search run over a ridge of an edge. A laser distance measuring head (2) with an evaluation unit (7), which controls the measuring head via the sensor cable (6) and outputs an analog voltage, is used to verify the measuring method. The evaluation unit is connected with the cable (8) to an A / D converter PC plug-in card (10). The digital values are processed in the PC (9). The PC (9) communicates with the robot computer (12) via the serial interface (11). The digital output (46) of the robot computer (12) is connected to a digital input card (45).

Before the actual measurement of the position of the workpiece is started, the offset vector (13) of the sensor zero point (14), the measurement direction vector (15) and the factor length unit per output value (current or voltage) must be determined.

Procedure:

Determine the sensor parameters, these are the sensor measurement direction vector (15), the offset vector (13) of the sensor zero point (14) and the factor length unit per measured value:

Step 1:

The robot is programmed so that it can send the coordinates of the TCP (3) and the employment of the TCP (3) to the PC (9) at the beginning and at the end of the search movement. Furthermore, the robot is programmed so that it divides the search run into four routes of equal length and inverts a digital output (46) at the transition from one route to the other. If the output is set it is reset, if the output is reset it is set.

Step 2:

A reference workpiece, which has two edges that can be recognized by the sensor, is permanently attached in the robot work area

Step 3:

Definition of the edges of the reference workpiece:

The coordinates in the robot system of the corner point (34) and two points (35, 36) on the edges (33) themselves are entered into the PC. This saves the values.

Step 4:

A robot step program is created that moves the sensor over the reference workpiece using two different search runs. The search runs must be carried out parallel to the surface of the reference workpiece and the sensor measuring point must be able to detect both edges. The orientation of the measuring direction (4) of the sensor head (2) can be freely selected.

Step 5:

Search trips:

Now the robot computer (12) sends the coordinates and the setting of the TCP (3) at the beginning and at the end of the search run via serial interface (11) to the PC (9), and the 4th

AT 406 313 B

Search run itself is started. The laser beam scans the reference workpiece along the line (17). To do this, the TCP (3) moves on the straight line (23) in space. The measured value of the sensor and the status of the digital output (46) are read in during the search run. The second search run on the straight line (24) immediately follows the first search run. This results in two line images (38) (39) of the reference workpiece. These line images can also be displayed graphically on the screen (37). 5 shows the graphical representation of the two samples (38) (39) and the value (44) of the digital robot output (46). The recording of the values begins with the rising edge, which is set at the start of the search run, and ends at the second falling edge of the digital output (46), which marks the end of the search run. The coordinates transmitted via the serial interface are the coordinates of the TCP (3) at those points where the digital output (46) switches the first rising edge and the second falling edge.

Step 6:

Calculation of points (25) (26) (27) (28):

The coordinates of the TCP (3) at those points at which the output (46) switches the first falling edge and the second rising edge are calculated. Since the coordinates of the first and last flanks are known, this is very easy. The PC looks for the edge points (40) (41) (42) (43) of the reference workpiece in the two line images (38) (39). By interpolating the coordinates of the edges of the signal (44), those coordinates (25) (26) (27) (28) that the TCP (3) had when the laser beam reached the edge points (19) (20) ( 21) (22) of the reference workpiece.

Step 7:

Calculation of the coordinates of the points (19) (20) (21) (22):

The distances between points (25) and (27) and points (26) (28) are the same length as the distances between points (19) and (21) and points (20) (22) when the search run was parallel to the surface of the reference workpiece. With the length and the points (34) (35) (36), the coordinates of the points (19) (20) (21) (22) can be calculated from the triangles, formed from the points (34) (19) (21 ) and (34) (20) (22).

Step 8:

Calculation of the offset vector (13), the measuring direction vector (15) and the factor length unit per measured value:

Fig. 4 shows the vector model for point C1 (19). Vector (32) points from the coordinate origin (30) to point T1 (25) and is identical to the calculated coordinates from step 6. Vector (31) points from the coordinate origin to point C1 (19) and is identical to the calculated coordinates from step 7 .

5 shows the vector model for point C2 (20). Vector (48) points from the coordinate origin (30) to point T2 (26) and is identical to the calculated coordinates from step 6. Vector (47) points from the coordinate origin to point C2 (20) and is identical to the calculated coordinates from step 7 .

With the coordinates calculated in steps 6 and 7, which vectors are from the robot zero point (30), vector A (13), measuring direction vector b (15) and the factor length unit per measured value V will now be calculated. The vectors (29) and (49) result from the multiplication of unit vector b (15) by the measurement value M times the factor V is equal to the length unit per measurement value.

Calculation of A, b and V: C1 (19) and C2 (20), as well as T1 (25) and T2 (26) are used for the calculation.

The approach A + VM b = C1 - T1 Fig. 4 A + VMb = C2-T2 Fig. 5 bx bx + by by + or bz = 1 b is unit vector M ....... The measured value of the sensor provides the solution : V = +/- SQRT (((C1x - Cx2 -Tx1 + Tx2) (C1x - Cx2 -Tx1 + Tx2) + (Cly - Cy2 -Ty1 + Ty2) + (C1y - Cy2 -Ty1 + Ty2) + 5

AT 406 313 B (C1z - Cy2 -Tz1 + Tz2) (C1z - Cz2 -Tz1 + Tz2)) / (M1-M2) (M1-M2)) bx = (C1x - Cx2 -Tx1 + Tx2) / V (M1 - M2) by = (C1y - Cy2 -Ty1 + Ty2) / V (M1 - M2) or = (C1z-Cz2-Tz1 + Tz2) / V (M1 - M2) A = C1 - TI -VMb

The sign is the main direction of the laser beam in the robot coordinate system and must be known. E.g. Direction + Z then sign plus, direction -Z then sign minus. Vector A (13) and b (15) are now vectors in the robot coordinate system. If a workpiece is now measured, the orientation of the TCP (3) must be the same as the orientation of the previous search runs. However, this is a limitation that makes the sensor system unsuitable for practical use. This problem is solved if the vectors A (13) and b (15) are transformed into the TCP (3) coordinate system. In step 5 the employment of the TCP (3) was already sent to the PC. This position is the rotation matrix R of the TCP coordinate system in the robot coordinate system or is calculated from the transmitted orientation.

Asx = Ax R11 + Ay R21 + Az R31 Asy = Ax R12 + Ay R22 + Az R32 Asz = Ax R13 + Ay R23 + Az R33 bsx = bx R11 + by R21 + or R31 bsy = bx R12 + by R22 + or R32 bsz = bx R13 + by R23 + or R33

Vector As and bs are thus decoupled from the previous search runs and refer to the TCP (3). Vector As, bs and factor V are saved. With this process, the zero point offset and the translation factor (volt or mA) per bit of the A / D converter are also automatically determined. No other method currently offers such ease of use.

Determination of the coordinates of the laser impact point (5) in the robot coordinate system without a search run in order to further process points on the workpiece surface (16) in the robot:

These points can be used to carry out a level measurement or to determine the height of the workpiece, or to determine the orientation of a surface with three points.

Step 1:

The robot is programmed so that it sends the coordinates and the orientation of the TCP (3) to the PC (9) via the serial interface (11).

Step 2:

The PC (9) has now received the orientation R and the coordinates TCP.

Step 3:

The vectors As and bs and the factor V are loaded.

Step 4:

Pr (5) = As (13) + bs (15) V M Fig. 2 V ..... translation factor of the sensor (length unit / measured value M) M ...... measured value

Pr (5) is thus the point of incidence of the laser beam in the TCP coordinate system on the workpiece surface (16).

Step 5:

Now Pr (5) is transformed with the rotation matrix R into the robot coordinate system.

Px = TCPx R11 + TCPy R12 + TCPz R13 + TCPx Py = TCPx R21 + TCPy R22 + TCPz R23 + TCPy 6

Claims (4)

AT 406 313 B Pz = TCPx R31 + TCPy R32 + TCPz R33 + TCPz Step 6: The coordinates of P (5) are sent to the robot via the serial interface (11). Determination of the coordinates and the height of a burr by means of a search run: Step 1: The robot is programmed so that the sensor head is guided over the workpiece, the orientation of the measurement direction and the search travel direction being freely selectable. At the start of the search run, the robot sends the PC the coordinates and the orientation of the start and end point. step 2: After the search run, the coordinates of the points (50) (51) (52) in the line image are calculated with the aid of the signal from the robot output (46). The same method is used which was used in the determination of the sensor parameters step 6. The three coordinates are the coordinates of the TCP at the point where the laser beam hit the points (50) (51) (52) on the workpiece. step 3: These coordinates are now transformed into impact points. It is the same method used in determining the point of impact step 3 step 4 and step 5 4: The coordinates of the three points of impact of the laser beam, before and after the ridge, are sent to the robot. Claims: 1. Method for determining the absolute coordinates of the point of impact of a distance sensor, mounted on a robot, on a workpiece surface using a sensor system that communicates with the robot controller and can determine the position and orientation of the TCP (3) and coordinate transformations with the help of the rotation matrix, characterized in that a reference workpiece which has two edges recognizable for the sensor is fixedly attached in the robot work area and the coordinates in the robot system of the corner point (34), and two points (35) (36) on the edges (33) The sensor control can be entered and saved so that the distance sensor, which is mounted on the robot as required with regard to the offset of the sensor's zero point to the TCP (3) and with regard to the orientation of the measuring direction of the sensor, is then carried out using two different search runs, which are parallel to the reference workpiece surface , wel are also programmed so that the point of impact of the sensor sweeps over both edges (33) of the reference workpiece, is moved over the reference workpiece and the sensor values are simultaneously read in and stored by the sensor controller with the coordinates of the TCP (3) in order to search for the search to calculate those coordinates of the TCP (3) where the sensor point exactly hit the edge of the reference workpiece during the search runs, in order to use the length between the point (25) and the point (27), the point (34) and the point (35) and the point (36) to calculate the coordinates of the point (19) and, from the length between the point (26) and the point (28), the point (34), the point (35) and the point (36 ) to calculate the coordinates of the point (20), then the point (19) the point (20) the point (25) and the point (26) in the relationship vector (13) + unit of length per measured value * measured value * unit vector ( 15) = point (19) - point (25) and in the relationship vector (13) + length unit per measured value * measured value * unit vector (15) = point (20) - point (26) is used to derive the factor length unit per measured value, the vector (13) and the unit vector (15) in the sensor control calculate and with the rotation matrix of the TCP (3) transform the vectors (13) and (15) into the coordinate system of the TCP (3) and save them in order to determine the absolute coordinates of the point of impact (5) of the distance sensor (2) a workpiece surface (16), and that the measured value of the sensor is then used in the relationship point (5) = vector (13) + vector (15) * length unit per measured value * measured value, which is the point of impact point 7 AT 406 31 3 B (5) returns the result in the TCP (3) coordinate system and is now transformed into the robot coordinate system using the rotation matrix of the TCP (3) and sent to the robot. Including 3 sheets of drawings 8