Laser Tracker and 6DoF measurement strategies in industrial robot applications

Transcription

1 Laser Tracker and 6DoF measurement strategies in industrial robot applications Dr. M. Kleinkes / Espace2001 S.A. (Luxemburg) Dr. R. Loser / Leica Geosystems (Switzerland) Dr. M. Kleinkes, Espace2001 S.A., L-2449 Luxemburg, Luxemburg Phone , michael@kleinkes.de word count: xxxx words Number of images: xxxx images

2 Laser tracker and 6dof measurement strategies in industrial robot applications Abstract In this paper, the usage of the Leica Laser Tracker, including especially 6dof products is presented in the field of industrial robot applications. Hereby, a major need of modern production cells is to increase the efficiency of robot programming in quality and time consumption. Copying and mirroring of complex robot movement paths on symmetrical objects is a non-avoidable step in today s flexible automation processes. But for this, a set of critical information about the robot cell has to be acquired and calculated. Identified by special measurement procedures, this information is integrated into the duplication process of the programs to deliver the needed accuracy of the real application results. In this process, the robot and optional adapted linear tracks, which are used to increase the robots working range, can be analyzed separately. The robot for itself is used for passing the DIN EN ISO 9283, a test concerning the identification of robot accuracy. Also various transformation corrections for the integration of the robot into the working cell help to increase the robots final accuracy behavior. The linear track is analyzed by a special method using the laser tracker and the robot to identify a continuous 6-dimensional mathematical description of the whole linear track. After all identification and optimization procedures are completed, a set of created input data is used in a software tool to calculate the appropriate robot point corrections for each constructed robot tool position in the copied or mirrored robot program. A sample program in comparison to a non-corrected application presents the final resulting accuracy after duplication of a constructed robot program including all corrections. The main focus of the paper is on the various measurement strategies and also on the improvements and new possibilities by using 6dof measurement products. The introduction of 6dof measurements in the field of robotic applications gives more flexibility and efficiency, as shown in special examples. Introduction The need of high-performing industrial robots is proportional to the increasing demands on the flexible automation. Complex applications contain not only a robot working with a single program, but the composition of multiple peripheral systems, which are working simultaneous to support, clean or check the robot. All this peripherals will have their own influence on the resulting system accuracy, which demands are even increasing with new technology like offline programming using computer aided simulation environments. This results the need of proper installation and calibration of the robot and all involved peripherals. Different measurement strategies for identification of dedicated coordinate systems can be applied to mathematically integrate every component into the application system. A significant improvement of these measurements can be reached, if it is possible to use a lasertracking device. Additionally, if this laser tracker has 6D-measurement functionality, it gives a lot advantages for measuring shadowed positions, which are not measureable with common 3D laser trackers. The next sections describe, which identifications of the application components can be done using a laser tracker device and how the different measurement strategies are. Analyzing industrial robots Based on the tolerances of every component used for industrial robots the resulting real position and orientation of the robots tool center point (TCP) will never fit exactly to the theoretical programmed one. Beginning with tolerances in geometrical size of the arm parts and ending with limited resolution of the integrated measurement units, the robot control is always dealing with positional errors in calculation. Figure 1: identification of robot base frame using a laser tracker

3 Defined by the application a minimum exactness in positioning of the robots TCP is needed to output sufficient quality of the application task. As an example for this, drilling a hole in a workobject using an industrial robot can be a very challenging task, if the allowed tolerances of the drilling result are strongly con strained. A central question coming up from these situations is how to identify the current robot accuracy in an objective and time-invariant way. One major approach to this was the definition of the international standard for identification of robot accuracy, published in the DIN EN ISO Using this method in combination with a laser tracker device for taking non-contacting 3D- and 6D-measurements is the first part of this paper. As shown in Figure 1, the robot moves to multiple positions in its working range, where a laser tracker device measures its TCP-position (mounted reflector). The resulting measurements can be used for calculation of the static positional and repeating accuracy corresponding to ISO9283. Using a laser tracker, there is an efficient and easy way to identify the robots base frame, which is needed to compare the measured and programmed coordinates. The robot TCP is moved to multiple different positions and is measured by the laser tracker, see Figure 2. After the measurement process, the transformation result between measurement device and robot can be calculated by an iterative solving of the over determined set of non-linear functions (e.g. Newton-Least-Squares). Figure 2: Robot program vs. measured points By this method, the robots base frame can be identified in a non-contact way (only mounting a reflector-tcp), which is quite useful in dirty or narrow application cells. Further, this identification process can be used for determination of the geometrical parameters (Denavit-Hartenberg {ref}) of the theoretical robot model. The standard parameter set included during manufacturing process will not fit exactly to the robots real geometry. The 6-dimensional measurement of the robots TCP-positions enhances the identification process for point database construction, which is used for input of the DH-parameter calculation. Taking the step from static accuracy to dynamic accuracy, there are multiple test methods described in the ISO 9283 standard. The core test for dynamic movements, the so-called optional path can be used for testing various kinds of path motions like circles, squares or different edges. Figure 3 shows an example deviation calculation of a straight-line movement. Figure 3: Straight-line movement deviations With the high measurement rate of modern laser trackers, it is possible to scan the TCP movement in a quasicontinuous way. Combined with the huge measurement range of up to 60 meters in radius, even large dimensions of robot path motions can be recorded and analyzed. This is core functionality for the identification of linear track geometries, which is mentioned later in {ref}. The automated software, which is using the laser tracker via its programming interface, is capable to synchronize the measurement process to the robot movement commands without manual interaction. To get an overview about this setup, figure {ref} displays the interaction between PC, robot and laser tracker. Assuring a real time process running on the PC, it is possible to construct a real-time measurement setup via a trigger interface of the laser tracker.

4 Therefore, start and stop commands are sent to the laser tracker synchronized to the start and stop triggers received from the robot. Dedicated switching test scenarios can be applied, to identify the robot timing delay between trigger and real movement. This can be used to adjust the application peripheral parameters, like spray gun trigger for example, to identified real switching timings to increase the interaction quality of peripheral components to the robot. Linear tracks One of the most important influences on the robot system accuracy is the use of an additional linear track. Common in the automobile industry they are used to stretch the working area of an industrial robot along large work pieces. Several aspects of the track geometry and position can be more or less compensated by the robot itself like gear factor setting or alignment orientation. These parameters can be identified, by using simple measurement methods during the installation process. A far more complex task is the identification of the linear tracks non-linear deformation. Due to constructional restrictions, the linear track is mounted on single equidistant, high-adjustable holders. The adjustment of the perfect height of a single holder to result a perfect straight line for the linear track is normally done during installation process without robot. Important in this step is the adjustment of the correct track height at every holder and of the exact position of the holder on the ground. Both are causing track non-linearity in different coordinates. Figure Figure 4 depicts robot deviations caused by in-accurate height adjustment and holder positioning. Figure 4 : Deviations caused by robot and track construction A new approach in the adjustment of linear tracks for robot systems can be done using a laser tracker device. It is measuring the robot TCP using the base frame identification method, mentioned in {ref}. At certain sampling positions on the linear track, the robot base frame for the current robot skid position is identified (Figure 5). After this, a 6-dimensional description of the linear track can be used to create a mathematical model of the track, to describe the nonlinearity. For this an interpolation in every degree of freedom is done to prepare this mathematical description, to calculate a correction value Figure 5 : Sampling of linear track by robot measurement for an arbitrary robot position on the track. Once constructed the model, for every programmed robot position, a 6-dimensional correction vector can be calculated, to compensate the current local track non-linearity. On of the great advantages of this correction method is the integration of the robot as physical influence factor into the track identification process. Different parameters like weight influence or forces on the track caused by robot arm configuration are integrated into the identified track nonlinearity.

5 Figure 6 : Robot accuracy with/without linear track correction The final correction of the robot TCP can be done in a offline mode using a software to correct existing robot programs, or an online correction using a data set integrated into the robot control which is using a lookup-table to pick the skid position-depending 6-dimensional correction values. Figure 6 is showing test results for a robot system with and without track compensation. Copy & Mirror process of robot programs Modern robot program construction is focused on computer aided design systems. Complex simulation software tools are able to create a virtual application cell on the computer and to simulate nearly every occurring component of the robot system. Forced by the increasing demands on the application strategy, robot programs are getting more and more complex. Against this, the supported time-slot for installation and setup of the application is more and more reduced to increase the productivity of the system. Complex robot programs are now constructed in a virtual simulation and tested with complex mathematical models of the real robots to get as close as possible to the real system. With this, the time for setting up the robots is significantly decreased, as all robot programs can be prepared, even before the robots are mounted into the application cell. Unfortunately, the usage of an offline-generated robot program without any correction is a task, which is depending on many predictions to the robot and the application. In nearly all cases, the offline robot programs need to be manually adjusted by a robot programmer on-site. This is caused by the gap between real world and theoretical simulation based on perfect conditions. One of these conditions, the linear track geometrical structure, is mentioned in {ref}. Most components of the application can be identified and measured using a laser tracker to increase the resulting accuracy in identification. Identification of robot work object The transformation between robot and current object is used to transform all coordinates into the object system, which is used for easier program construction. To identify this transformation, the 6- dimensional object position has to be measured. Using a laser tracker, references at the car body with known coordinates can be measured with high accuracy. Figure 7 : Symmetrical robot cell This differs from common methods for integration like using robot coordinates in tactile measured TCP positions. In comparison to other measurement systems, e.g. stereovision devices, the work piece size can be much larger, which specially can be used in wind craft industry. The laser tracker, especially with 6D-functionality enables measurements of the work pieces even at hidden positions, which increases the number of possible points for object coordinate system identification. Due to this, the quality of the identified frame is higher because of more stability by more used points for solution of the non-linear equation set for transformation calculation.

6 Vision system calibration The advantage of more flexibility in measurement provided by laser tracking devices, enables new techniques for vision system calibration and integration. Multi-camera vision systems can be used, to identify the 6-dimensional shift of an object relative to its calibrated position. Calculated shifts are sent to the robots to adapt the program. Different calibration methods for vision system include the identification of the camera coordinate system, see figure {ref}. Taking camera pictures of referenced calibration bodies or markers can identify this coordinate system. The direct measurement of the camera body is another possibility of camera coordinate system identification, but it is depending on the flexibility of the used measurements system. Using 6D laser tracker, even difficult measurements on the mounted cameras can be taken and used for integration. Figure 8 : Applcation cell with vision system Laser tracker as inline measurement system Thinking about pose correction of robots, an ideal solution is to measure the robots TCP permanently during application. In this case, all measurements and corrections can be done in the coordinate system of the current work object, which avoids inaccuracies in identification of robot base frame. Also, every influencing component of the robot system is integrated into the error chain affecting the TCP. By controlling the TCP movement during application, all errors can be compensated by one single correction vector, calculated by using the far more exact measurement result delivered by the laser tracker. As an example for one crucial peripheral robot component, the linear track should be mentioned. As seen in section {ref}, the accuracy influence of a linear track can be a deciding influence on the resulting application. Given by the 6-dimensional correction possibility provided by modern laser trackers, the non-linearity and with this error influences caused by linear tracks can be nearly completely compensated. The resulting accuracy of the robot system is no longer dominated by the accuracy of the robot itself, the exactness in setup or the influence of peripheral devices. Depending on the capability of process even small correction vectors, it is possible to reach application accuracy up to 5-times better than robot accuracy. First tests provided results of an industrial sealing robot (16 kg maximum tool weight), which was used with accuracy better than 0,1 mm in absolute static positioning. The strategy, how to provide or process this correction value varies with several parameters. The most important of these parameters are robot interface, laser tracker availability and needed application accuracy. In this section, there are three different scenarios presented, how to use the laser tracker as inline measurement system for controlling the robots TCP movement. Static positioning correction Applying static positioning control, single TCP positions are measured and corrected with a 6- dimensional shift vector. This can be done without, or if possible directly on the work object. The needed measurements are taken only at one time, which is reducing the needed usage time of the laser tracker. After correction, the robot is using the corrected program, which consists of new created TCPpositions or dedicated correction offset for every point of the original program. A prediction for this correction method is a robot time-invariant behavior. Therefore, the resulting accuracy is dependent on the robots physical model changing like fading or aging. For guaranty of constant exactness in positioning the correction vector identification must be repeated periodically or after modifications of the robot system.

7 Even after a dedicated robot program has been corrected, additional points which are added after correction process can only be corrected in the next identification and correction run and must be used until this with normal robot system accuracy. A major advantage of this method is the efficient usage of the laser tracker device for many robot cells and can be used for many applications with constant robot program structure. First identified accuracy results using a common sealing robot for automotive industry (end effector maximum weight 16 kg) is up to 5 times better than robot system accuracy with results better than 0.1 mm absolute static positional accuracy in its complete working range. Dynamic correction control loop If the application deals with complex movements paths, which need high positioning accuracy the robot can be put in a correction control loop together with the laser tracker. In this, the laser tracker is the indicator and the robot the actuator. The 6-dimensional position of the robots TCP is measured in a quasi-continuous way using a measurement rate of 1000 measurements per second or faster. Using special correction interfaces, which are changing depending on the robot manufacturer, it is possible to send a 6D correction offset to the robot control, which is processed with an update rate between 2 ms and 16 ms. Provided via the high update rate for control offset values, it is possible to react not only on dynamic robot movement parameter like flexibilities or linear track deforming. Even process feedback forces can be detected as disturbance of the control loop and compensated by the offset vector calculation unit (see Figure 9). Figure 9 : Dynamic correction control loop The influences that can be detected and compensated are strongly depending on the TCP movement speed and the maximum processing rate for the correction offsets inside of the robot control. Side effects are swinging of the control loop, processing delays or beam breaks because of shadowing of the reflector. Even if the laser tracker is blocked as always-on inline measurement device, all time-variant effects and changings of the robot system are integrated as disturbance in the control loop and the result will be of constant TCP movement accuracy.

8 MAP Metrology Assisted Production As multi-iterative static correction procedure the MAP-process is an example for many assembly or guided-movement applications. It bases on the separation of a complex non-linear movement into controlled sub-movements. For this, the original movement path is supplied with special correction positions, which amount and position is depending on the most important movement positions. During a correction cycle, the robot is stopping or slowing down for getting the current TCP position, calculation and processing of a correction offset for the next following part of the movement. It delivers a compromise between movement distance and accuracy, because the positional deviation between practical and theoretical point will increase proportional with TCP distance form the last corrected position. The final step of the MAP-process is the blind step, which is done after last correction value and final movement Figure 10 : MAP process demonstration (shown in the top picture of Figure 10). This can be conferred to many practical applications where the correction positions can be set for example before and after critical movement paths or reorientations. The easy implementation of the MAP structure without need of special high-speed correction interfaces is especially adequate for robots with limited correction value processing units. Result Movement correction of industrial robots defines widely structured area reaching from simple offline pose correction to complex dynamic path planning. The central issue of the used correction method is given by the complexity and needed accuracy of the application. Due to missing standard-interfaces of common industrial robots, the online pose-correction is currently a method for special applications with unsolvable accuracy restrictions to the robot system itself. In general, the step to external pose correction can be used to keep all robot depending accuracy influences out of the application system and to use the robot like theoretically defined in CAD world. The flexibility in measurement provided by long range contact-less measurement of laser trackers is one of the basic demands for setting up an efficient dynamic robot correction control loop. Additionally, the 6-dimensional measuring of the robot tool provides direct corrections in all degrees of freedom. No static multi-measurement procedure for identification of the robots tool frame is needed, so the provided measurement rate of up to 1 khz can be directly used for correction. In combination with robust and intelligent measurement components, which can easily be adapted on the robot via automated tool changers, the needed flexibility for the integration of laser tracker systems as inline measurement devices into the robot cell is prepared.

9 References [1] International Organization for Standardization, DIN EN ISO9283:1998, Manipulating industrial robots Performance criteria and related test methods, European Standard, 05/1999. [2] Leica Geosystems AG, [3] Spong, M.W., Vidyasagar, M.; Robot dynamics and control, John Wiley & Sons, [4] Paul, R.P., Shimano, B.E., Mayer, G.; Kinematic control equations for simple manipulators, IEEE transactions on systems and cybernetics, Vol. SMC-11 no. 6, [5] Daemi-Avval, M.; Modeling and identification of the dynamic of industrial robots for the use in closed loop controls, Progress-Reports VDI, Series 8, Volume 719, Germany, [6] Maas, H. G.; Dynamic photogrammetric calibration of industrial robots, SPIE proceedings series Vol. 3174, [7] Kleinkes, M., Lilienthal, A., Neddermeyer, W.; Highly accurate integration of track motions International conference on Informatic in Control, Barcelona, [8] Kleinkes, M., Neddermeyer, W., Schnell, M.; An automated quick accuracy and output signal check for industrial robots, WSEAS Int. Conf. on ROBOTICS, CONTROL and MANUFACTURING TECHNOLOGY (ROCOM 06), Hanghzou (China), 2006.