Method and device for calibrating a robot, method for determining the 3D coordinates of a measurement object, method for determining the position and orientation of a measurement object, robot, and computer program

Abstract

The present invention relates to a method for calibrating a robot kinematics system, wherein, in a first step, a 3D digitiser and / or a camera is / are brought, by means of at least one robot kinematics system, into different relative poses with respect to one another and relative to a reference object, and, in these poses, data of detectable features of the reference object are recorded by the 3D digitiser and / or the camera, and wherein, in a second step, measurement data are generated at least from part of the recorded data, said measurement data relating the recorded features of the at least one reference object to one another in a coordinate system, and calibration parameters for the at least one robot kinematics system are derived at least from part of the recorded data. The present invention also relates to a robot kinematics system and to a device for calibrating a robot kinematics system using a coordinate measuring device or as part of a coordinate measuring device, as well as to a method for determining the 3D coordinates of a measurement object and to a method for determining the position and orientation of a measurement object. The present invention also relates to a computer program comprising instructions which, when the program is executed by at least one processor, cause said at least one processor and / or further processors to carry out the methods according to the invention.

Description

Method and device for calibrating a robot, method for determining the 3D coordinates of a measuring object, method for determining the position and orientation of a measuring object, robot and computer program

[0001] The present invention relates to a robot, a method and a device for calibrating a robot with a coordinate measuring machine or as part of a coordinate measuring machine, as well as a method for determining the 3D coordinates of a measurement object and a method for determining the position and orientation of a measurement object.

[0002] The present invention further relates to a computer program with instructions which, when the program is executed by at least one processor, cause this at least one processor and / or further processors to carry out the methods according to the invention.

[0003] Robot kinematics (hereinafter also referred to as robots) are mechanical structures in which one or more axes are coupled to one another. The goal of robot kinematics is to bring an end effector into a specific position and orientation (pose) in relation to a reference system. An axis is understood as the movable connection between two structural components of the robot kinematics, whereby this connection can be movable in up to six degrees of freedom. Examples of this are a typical robot rotary axis that only allows rotation in one rotational degree of freedom, a typical linear axis that allows translation in one translational degree of freedom, or a spherical joint (ball joint) that allows rotation in three rotational degrees of freedom. Typical robot kinematics are six-axis robots. In particular, a rotary table can also be considered a robot kinematics.The robot axes can be equipped with an actuator or operated passively. The reference system can be located on the floor, for example. Within the scope of the present invention, mobile robot kinematics traveling on corresponding mobile platforms can therefore also be calibrated individually or relative to each other. In this case, the reference system can be located on the floor of the platform that is stationary during the measurement.

[0004] Robot kinematics, whether parallel, serial, or coupled parallel and serial, such as an articulated arm on a delta robot, are inaccurate for metrology purposes in many respects. This means that the nominal forward kinematics do not correspond to the actual forward kinematics. This is caused, for example, by geometric inaccuracies, elastic deformations, thermal expansion, and inaccuracies in the determination of the axis values. This is particularly problematic when the pose of the end effector or the tool mounted on the end effector (measuring device, tool, etc.) must be determined very precisely or is used for metrology. The end effector is the end interface along the link chain of the robot kinematics or the tool attached to the end interface.

[0005] In optical metrology, particularly 3D coordinate metrology, a measurement object can usually only be partially captured by an optical measuring system due to its limited measuring range. Therefore, the measuring system is positioned in various positions relative to the measurement object using a robot in order to capture the measurement object as completely as possible. The data available locally in the reference coordinate system of the measuring system is then combined in a global coordinate system. To do this, it is necessary to know the position and orientation of the measuring system in relation to the measurement object as precisely as possible. This accuracy is directly related to the positioning accuracy of the robot, or rather, the accuracy in determining the position of the end effector. The better a robot is calibrated, the more accurately the position of the measuring system can be determined and the more accurately the measurement object as a whole can be measured.

[0006] Measurement objects can be so large that even the reach of a robot is insufficient to capture the measurement object with the measuring system attached to the robot. In this case, the measurement object must be moved and realigned with the robot in order to be able to transfer the data before and after the measurement object is moved into a common global coordinate system. This is usually done using recognizable features attached to or around the measurement object, which are captured before and after the movement with the measuring system. The better the robot is calibrated, the more precise the alignment.

[0007] In 3D coordinate metrology, 3D digitizers are used to measure 3D data. The 3D data can be generated using various methods, such as triangulation with at least two cameras or one camera and a projector, time-of-flight (TOF) measurement, lidar, confocal measurement, etc. The 3D data can be present, for example, in the form of individual 3D points of discrete features or as a full-surface mesh of a large number of surface points. 3D data can also be generated using a 2D camera. However, a single 2D camera that cannot generate 3D data in a single pose without further prior knowledge of the measurement object, such as the distance between features of the measurement object, is not considered a 3D digitizer within the meaning of the invention. 2D cameras are also referred to simply as cameras below.

[0008] EP 2 489 977 B1 describes a method with a measuring system for determining the 3D coordinates of an object with a camera and a projector and a reference camera for recording reference marks of a field of reference marks for determining the position and orientation of the measuring system from one or more images of the reference camera and an industrial robot for positioning the measuring system, wherein the measuring system and the field of reference marks are used to calibrate the robot,

[0009] Since pose determination is inaccurate due to the acquisition of 2D data with a reference camera, and thus a 2D camera, particularly in the viewing direction of the 2D camera, EP 2 489 977 B1 further proposes using multiple reference cameras aligned in different viewing directions. However, using additional 2D cameras increases the cost of the system, as does the weight that must be attached to the end effector of a robot. This may mean that a robot with a correspondingly high payload must be used.

[0010] EP 3 377 948 B1 describes a method in which, at several points in time of a scanning plan, the positions of a camera system of a robot relative to a reference coordinate network frame are obtained by comparing three-dimensional images of a scene and the positions of the robot are obtained by using a second scanning plan. to determine the position of the reference coordinate network frame and the reference point of the camera system relative to the position of the robot. Several equations are created that are used to solve an optimization problem. It is described that the position of the reference coordinate network frame, the tool center point of the camera system, a possible time offset, and parameters of the robot model can be optimized in the optimization process. One input parameter of the optimization process is the position of the camera system, which results from a comparison of the three-dimensional images of the scene and is calculated in a step preceding the optimization process.

[0011] As mentioned in EP 3 377 948 B1, it is difficult to arrange the three-dimensional images using visual odometry, particularly in areas with few features.

[0012] This is especially true for low-resolution 3D digitizers with small measurement volumes. The larger the measurement volume, the higher the probability of capturing a feature, and the higher the resolution, the higher the probability of detecting a feature or accurately capturing its location.

[0013] In "Dynamic Photogrammetry Calibration of Industrial Robots," Videometrics V, SPIE Proceedings Series Vol. 3174, SPIE's 42nd Annual Meeting, San Diego, July 27 - August 1, 1997, a method for calibrating a robot is described. A reference object is attached to the robot, and 3D coordinates of marked points on the reference object are determined by recording with three synchronized cameras. The poses of the end effector are determined based on the 3D coordinates of the marked points. The determination of the 3D coordinates and the determination of the orientation and camera model parameters of the three-camera system occur simultaneously.

[0014] The disadvantage of the method is that inexpensive cameras have a relatively low resolution and therefore the position of the points of the reference object in the camera images and thus also in space cannot be determined very precisely.

[0015] The object of the present invention is to provide a better method and a better device for calibrating a robot kinematics as well as a better method for determining the 3D coordinates of a measurement object and a better method for determining the position and orientation of a measurement object as well as better robot kinematics and a computer program for carrying out the better method according to the invention.

[0016] Against this background, the invention proposes a method for calibrating at least one robot kinematics, which comprises at least: - a first step, a) in which at least one 3D digitizer attached to a first robot kinematics and / or at least one camera, the camera being attached to this first robot kinematics or to a further second robot kinematics, and at least one reference object located in the vicinity of one robot kinematics or in the vicinity of the two robot kinematics and stationary for the recordings, which has recognizable features, are brought into different relative poses to one another by the robot kinematics or the robot kinematics, and in which data of the recognizable features of the reference object are recorded in these different relative poses by the at least one 3D digitizer and / or by the at least one camera;or b) in which at least one 3D digitizer located and fixed in the environment of a first robot kinematics and / or at least one camera fixed in the environment and at least one reference object attached to the first robot kinematics, which has recognizable features, are brought into different relative poses to one another by the first robot kinematics and in which data of the recognizable features of the at least one reference object are recorded in these different relative poses by the at least one 3D digitizer and / or the at least one camera; or c) in which at least one reference object attached to a first robot kinematics, which has recognizable features, and at least one 3D digitizer; rer, wherein the 3D digitizer is attached to a further second robot kinematics or to a further third robot kinematics, and / or at least one camera, wherein the camera is attached to the second robot kinematics or to the third robot kinematics, are brought into different relative poses to one another by the first robot kinematics and at least one of the further robot kinematics, and in which data of the recognizable features of the at least one reference object are recorded in these different relative poses by the at least one 3D digitizer and / or the at least one camera;characterized in that in at least one further second step, at least from a part of the recorded data, both calibration data are generated which relate the recorded recognizable features of the at least one reference object to one another in a coordinate system, and calibration parameters for at least one of the robot kinematics are derived from at least from a part of the recorded data.;

[0017] It goes without saying that the camera can also be part of the 3D digitizer for all initial substeps of the method for calibrating robot kinematics. This naturally also applies to the subsequent method claims for measuring and aligning a measurement object, as well as to the corresponding device claims.

[0018] According to a further aspect of the present invention, a method for determining the 3D data of a measurement object is proposed, wherein at least one 3D digitizer attached to a robot kinematics or several 3D digitizers attached to robot kinematics are brought into different poses for measuring the measurement object and the 3D digitizer(s) generate local 3D data in the reference coordinate systems of the 3D digitizers and wherein the local 3D data generated for the different poses are based on the method according to the invention derived calibration parameters of the robot kinematics are transferred into a global coordinate system.

[0019] According to a further aspect of the present invention, a method is proposed for aligning a measurement object with respect to a robot kinematics based on recognizable features that are distributed on or around the measurement object and are captured by at least one 3D digitizer attached to a robot kinematics and / or by at least one camera attached to a robot kinematics, wherein the calibration parameters of the robot kinematics were derived by means of the method according to the invention.

[0020] According to a further aspect of the present invention, a device for calibrating robot kinematics is proposed, which comprises at least the following components: at least one 3D digitizer attached to a first robot kinematics and / or at least one camera, wherein the camera is attached to this first robot kinematics or to a further second robot kinematics, and at least one reference object located in the vicinity of one robot kinematics or in the vicinity of both robot kinematics and stationary for the recordings, which has recognizable features, or at least one 3D digitizer located and fixed in the vicinity of a first robot kinematics and / or at least one camera fixed in the environment and at least one reference object attached to the first robot kinematics, which has recognizable features, or at least one reference object attached to a first robot kinematics,which has recognizable features, and at least one 3D digitizer, wherein the 3D digitizer is attached to a second robot kinematics or to a third robot kinematics, and / or at least one camera, wherein the camera is attached to a second robot kinematics or to a third robot kinematics, wherein for at least one of the robot kinematics, calibration parameters for this robot kinematics are derived by means of the method according to the invention.

[0021] According to a further aspect of the present invention, a robot kinematics whose calibration parameters were derived by means of the method according to the invention is proposed.

[0022] Furthermore, a computer program with instructions is proposed which, when the program is executed by at least one processor, cause this at least one processor and / or further processors to carry out the method according to the invention.

[0023] According to the invention, a method for calibrating robot kinematics is provided, wherein data for measuring a reference object and for deriving the calibration parameters of the robot kinematics or the robot kinematics are generated and at least part of the data is evaluated in a step which includes generating the measurement data and deriving the calibration parameters.

[0024] The generation of the measurement data and the derivation of the calibration parameters are carried out using an optimization process. Advantageously, for example, within the second step of the method according to the invention, it can be iteratively checked after each optimization iteration to determine which data is used for measuring a reference object and which for deriving the calibration parameters, or to weight these accordingly. Thus, for each data value of a recognizable feature, it can be checked to determine whether it is, for example, an outlier with regard to the measurement of the reference object or the derivation of the calibration parameters. This makes it possible to detect inconsistencies between the data, which could be caused, for example, by a temperature jump during data acquisition or by instability of the camera or 3D digitizer.Thus, data can be selected iteratively and selectively to obtain an ideal solution of the optimization procedure.

[0025] If an optimization were first performed to measure the recognizable features of the reference object, errors could arise that would propagate the errors in deriving the calibration parameters of the robot kinematics in a subsequent optimization. This is not the case with the method according to the invention, since both the errors in measuring the recognizable features of the reference object and the errors in deriving the calibration parameters of the robot kinematics are minimized. This means that errors in measuring the recognizable features that are identified by deriving the calibration parameters can still be corrected.

[0026] In addition, a possible combination of 2D data with 3D data can advantageously solve or even eliminate an initial value problem of the optimization procedure, which would exist when using only 2D data.

[0027] The data is recorded by at least one 3D digitizer attached to a first robot kinematics and / or at least one camera, the camera being attached to this first robot kinematics or to a further second robot kinematics, and at least one reference object located in the vicinity of one robot kinematics or the two robot kinematics and stationary for the recordings, which has recognizable features, being brought into different relative poses to one another by the robot kinematics or the robot kinematics, and in which data of the recognizable features of the reference object are recorded in these different relative poses by the at least one 3D digitizer and / or by the at least one camera.

[0028] Alternatively, the data is recorded by bringing a reference object attached to a robot kinematics into different relative poses to a 3D digitizer located and fixed in the environment of the robot kinematics and / or a camera located and fixed in the environment of the robot kinematics, and in these different relative poses, data of the recognizable features of the reference object are recorded.

[0029] Alternatively, data can also be recorded by placing a reference object attached to a first robot kinematics system, which has recognizable features, and at least one 3D digitizer, wherein the 3D digitizer is attached to a second robot kinematics system or to a third robot kinematics system, and / or at least one camera, wherein the camera is attached to a second robot kinematics system or to a third robot kinematics system, in different relative poses to one another, and recording data of the recognizable features of the reference object in these different relative poses. Thus, using a 3D digitizer and a camera, it is possible, for example, for the 3D digitizer and the camera to be attached to a second robot kinematics system, or for the 3D digitizer to be attached to a second robot kinematics system and the camera to be attached to a third robot kinematics system.The camera is attached to a second robot kinematics system, and the 3D digitizer is attached to a third robot kinematics system. For example, it is equally possible for only one 3D digitizer or only one camera to be attached to the second robot kinematics system.

[0030] In principle, it is possible to derive calibration parameters for multiple robot kinematics, which, depending on the subsequent task of the robot kinematics, acquire data using one of the aforementioned methods. The method for acquiring data for the respective robot kinematics can be different from the method for acquiring data using one of the other robot kinematics.

[0031] The recorded data can be 2D data captured with one or more cameras, including one or more cameras of the 3D digitizer. It can also be 3D data generated by the 3D digitizer.

[0032] The use of a camera, especially a high-resolution camera, is advantageous for measuring the recognizable features of the reference object. The use of a 3D digitizer is advantageous for determining the pose of the end effector, since depth data is particularly relevant for determining the pose. This makes it possible to use a high-quality, high-resolution camera to measure the recognizable features of the reference object based on 2D data, and a 3D digitizer for highly accurate determination of the end-effector's pose based on 3D data. However, a camera can also be used to determine the end-effector's pose, or a 3D digitizer can be used to measure the recognizable features of the reference object, or both can be combined.

[0033] Calibration data is generated from this or at least part of the recorded data, which relates the recorded features of the reference object to each other in a coordinate system. Furthermore, calibration parameters for the robot kinematics are derived from this or at least part of this recorded data.

[0034] In order to calibrate robot kinematics using a reference object, the reference object must be calibrated with high precision. This can be achieved, for example, by calibrating the reference object in advance in a calibration laboratory. If the reference object does not remain calibrated with long-term stability, regular calibration is necessary to ensure lasting accuracy. The advantage of the method according to the invention is that the reference object is calibrated with high precision during the method according to the invention by generating calibration data, thus eliminating the complexity of preliminary calibration and long-term stability.

[0035] Influencing factors such as the gravity orientation (substep b) or substep c)) of the reference object or the ambient temperature can also be taken into account. It is therefore possible, for example, to carry out the first step with different ambient temperatures and to generate the calibration data or the references of the recorded recognizable features of the at least one reference object in a coordinate system in a temperature-dependent manner, whereby it must be taken into account that sufficient recordings of the recognizable features of the at least one reference object are generated for the different ambient temperatures. Due to the existing temperature dependence of the references of the recorded recognizable features of the at least one reference object, this can lead to improved ambient temperature-dependent calibration of the robot kinematics.

[0036] Since the requirement for long-term stability of the reference object is reduced, a lightweight reference object mounted on a robot kinematics can be used, thereby reducing the payload on the robot kinematics, especially compared to a 3D digitizer attached to the robot kinematics. Furthermore, if the reference object is attached to the robot kinematics and not the 3D digitizer or camera, cabling along the robot kinematics can be eliminated.

[0037] The generation of measurement data and the derivation of calibration parameters takes place in one step, in that at least part of the recorded data is used in an optimization process, which creates both the measurement data and the calibration parameters in one result.

[0038] With the newly determined calibration parameters, the pose of a 3D digitizer can now be determined more precisely and thus a measurement object can be measured more accurately.

[0039] Furthermore, the newly determined calibration parameters also enable a more precise alignment of a measuring object to the calibrated robot kinematics.

[0040] The above-mentioned task is thus completely solved.

[0041] According to a further embodiment, the 3D digitizer is a triangulation sensor, in particular one with stripe light projection.

[0042] The advantage of triangulation sensors is that they can determine 3D data very precisely and can also be implemented relatively easily, for example by combining two cameras or a camera and a projector.

[0043] According to a further embodiment, the data for generating the measurement data and for deriving the calibration parameters can be different, identical or partially identical.

[0044] For example, it can be helpful to only use data recorded with a high-resolution camera to generate calibration data, and to only use data recorded with a 3D digitizer to derive the calibration parameters, so that these are completely different. However, it can also be helpful to select poses in such a way that the recorded data is suitable for both generating calibration data and deriving calibration parameters. The data sets are therefore suitable for generating calibration data and deriving calibration parameters, and these can therefore be completely or at least partially identical. This can advantageously reduce the time required to acquire the data.

[0045] According to a further embodiment, a scale-defining element is present on the reference object and / or a robot link itself is a scale-defining element

[0046] A scale-defining element is advantageous for determining the distances between recognizable features absolutely, rather than just relatively. Particularly when only 2D data is available, the distances between recognizable features of the reference object are scalable if there is no fixed reference to a scale-defining element. The scale-defining element could be, for example, a stable precision body. This is calibrated with high precision, usually using a tactile coordinate measuring machine, and is made of a material that expands little, ideally not at all, when the temperature changes. Precision bodies are usually simple linear scales, such as a ball-and-socket rod. The advantage is that by generating the calibration data, the accuracy of the precision body is transferred to the more complex reference object, without the reference object itself having to be calibrated with high precision (e.g., tactile).However, a scale-defining element can also be an exactly measured distance between two recognizable features of the reference object.

[0047] According to a further embodiment, the recognizable features of the reference object are absolutely related to one another by at least one absolute measurement with the 3D digitizer, whereby the 3D digitizer itself is the scale-determining element.

[0048] A calibrated 3D digitizer measures 3D data absolutely. A measurement of the recognizable features of the reference object with the 3D digitizer can therefore be used to determine the absolute distances between the recognizable features of the reference object, which may only be available in relative terms. This advantageously eliminates the need for a scale-defining element on the reference object.

[0049] According to a further embodiment, at least part of the local data of the recognizable features of the reference object present in the local reference coordinate system of the 3D digitizer and / or the local data of the recognizable features of the reference object present locally in the reference coordinate system of the camera are converted into a global reference coordinate system.

[0050] For the transfer or transformation of the local data of the recognizable features of the reference object (actual observations) into a global coordinate system, it is helpful to know the pose of the 3D digitizer attached to the end effector and / or the camera attached to the end effector or the reference object attached to the end effector. This results from the forward kinematics of the robot kinematics and the relationship between the 3D digitizer and / or camera or reference object and the end effector. The forward kinematics, or forward transformation, is the function that describes the position and orientation of the end effector, as well as individual robot links, in a given coordinate system, depending on the robot configuration given in the axis space (for example, angular positions of rotary joints or linear positions of linear axes).To do this, the coordinate transformations, for example of the individual robot links, must be known. Nominally, these coordinate transformations for a given robot are known from data sheets, technical drawings, CAD data, given Denavit-Hartenberg parameters, etc. The local data present in the local reference coordinate system of the 3D digitizer or camera relate, for example, to the position and orientation of the camera sensors. In the case of a triangulation sensor, this is specified, for example, by the calibration of the triangulation sensor. The lengths and angles required for the transformation between the local reference coordinate system of the camera sensor and the coordinate system of the end effector can therefore nominally be determined, for example, from the CAD data of the 3D digitizer and / or the camera and any necessary adapters between the 3D digitizer and. End effector can be read out. CAD data of the reference object can also be used, for example, to establish a relationship between the end effector and the recognizable features of a reference object attached to the end effector.

[0051] In the following, the calculation of the forward kinematics from these nominally known values ​​(i.e., values ​​not calibrated to reality) is referred to as "nominal forward kinematics." However, due to factors such as geometric inaccuracies, elastic or thermal deformations, or inaccuracies in determining the axis angles or lengths of the robot limbs, the accuracy of determining the position of the 3D digitizer and / or the camera or reference object on the robot kinematics is usually insufficient. The nominal forward kinematics therefore differs from the actual kinematics.

[0052] The data available after the transformation in the global coordinate system are then used as starting values ​​for a subsequent optimization procedure. By transforming to a global coordinate system using forward kinematics, it is more likely that a better extremum of the optimization procedure will be found.

[0053] According to a further embodiment, the generation of the measurement data of the recorded recognizable features of the reference object and the derivation of the calibration parameters of the robot kinematics and optionally an adjustment of the calibration parameters of the 3D digitizer or the camera are carried out in such a way that the deviations between the coordinates of the recorded recognizable features of the reference object and the coordinates of the recognizable features of the reference object calculated from a mathematical model are minimized.

[0054] A deviation can be, for example, a deviation in the position and / or a deviation in the orientation of the recognizable features of the reference object,

[0055] Within an optimization process, there are essentially two tasks. One is the generation of the calibration data using the data provided for the generation of the calibration data, and the second is the derivation of the calibration parameters of the Robot kinematics. Both are achieved simultaneously by simultaneously minimizing the deviations between the coordinates of the recognizable features of the reference object recorded for generating the calibration data and for deriving the calibration parameters, and the coordinates of the recognizable features of the reference object calculated from a mathematical model. The mathematical model can be a combination of a bundle block adjustment (see Luhmann, Robson, Kyle, Boehm, Close Range Photogrammetry and 3D Imaging, Second Edition, Chapter 4, ISBN 978-3-1110-2986-3) and a mathematical model that contains the calibration parameters of the robot kinematics and thus describes the robot kinematics, such as the forward kinematics mentioned above. This can be achieved, for example, using methods for solving a nonlinear optimization problem according to Nocedal, Wright, Numerical Optimization, Second Edition, ISBN 978-1-4939-3711-0.

[0056] If necessary, the calibration parameters of the 3D digitizer and / or the camera can also be adjusted, provided that a scale-defining element is captured by the 3D digitizer and / or the camera (see Luhmann, Robson, Kyle, Boehm, Close Range Photogrammetry and 3D Imaging, Second Edition, ISBN 978-3-1110-2986-3) or, if necessary, the calibration parameters of the 3D digitizer and / or the camera can also be adjusted without a scale-defining element being captured by the 3D digitizer and / or the camera, provided that one of the robot limbs and thus a part of the robot kinematics itself is the scale-defining element.Likewise, if necessary, scale-independent calibration parameters of the 3D digitizer and / or the camera can also be adjusted without capturing a scale-defining element, provided that the scale-dependent calibration parameters of the 3D digitizer and / or the camera were previously captured using a scale-defining element.

[0057] The input parameters for the optimization process are the calibration parameters of the robot kinematics obtained from the forward kinematics, the recorded data from the 3D digitizer or camera, if necessary with an indication as to whether these are to be used for generating the calibration data or for deriving the calibration parameters of the robot kinematics, and, if necessary, calibration parameters of the 3D digitizer or camera. The output parameters of the optimization process are the optimized calibration parameters of the robot kinematics, the generated calibration data of the recognizable Features of the reference object and, if applicable, the optimized calibration parameters of the 3D digitizer or camera.

[0058] The method according to the invention can therefore be used to calibrate a robot kinematics, measure the reference object and calibrate the 3D digitizer or camera.

[0059] By minimizing, an ideal solution to the optimization problem is advantageously found.

[0060] While the optimization procedure and the identification of the calibration parameters can be formulated using forward kinematics, it is equally possible to set up an equivalent optimization problem using inverse kinematics.

[0061] According to a further embodiment, the calibration parameters of the robot kinematics are geometric and / or elastic and / or thermal and / or calibration parameters of the robot kinematics that take the configuration of the robot kinematics into account, and / or calibration parameters that take the direction of approach into account, and / or calibration parameters of the characteristics of the or calibration parameters for determining the error curves of the length or angle measurement technology of the robot kinematics and / or calibration parameters that characterize the play of individual or all axes of the robot kinematics and / or calibration parameters that describe the dependence of the robot kinematics on other variables that can be recorded, for example, with sensors or can be calculated based on the available data.

[0062] Depending on the type of robot kinematics and the environmental conditions of the robot kinematics, as well as the type of 3D digitizer or camera and the reference object, it is useful to use different combinations of calibration parameters for optimization.

[0063] However, the method according to the invention is not limited to these calibration parameters and can be supplemented and combined with other parameters.

[0064] According to a further embodiment, the calibration parameters are used to determine the pose of the end effector and / or to align the end effector based on the parameters.

[0065] The adjusted calibration parameters resulting from the optimization process can be used to position the end effector of the robot kinematics more precisely. This is useful if the measurement of a measurement object takes place after the robot kinematics have been calibrated. The calibration parameters can also be used to precisely determine the pose of the end effector. This makes it possible, for example, to use cost-effective robot kinematics whose approach accuracy to a certain pose is low for mechanical reasons, but the pose can still be precisely determined using the adjusted calibration parameters. Furthermore, it is also possible to subsequently correct the poses of a measurement. If a 3D digitizer orIf a camera is attached to the robot kinematics and is also used as a measurement sensor, the forward kinematics already includes the transformation from the global reference coordinate system to the local reference coordinate system of the 3D digitizer or camera. Thus, no additional hand-eye calibration is necessary.

[0066] According to a further embodiment, 3D digitizers and / or cameras and / or reference objects are mounted on several robot kinematics and at least one of the robot kinematics is calibrated according to the method according to the invention.

[0067] For example, a combination of two or more robot kinematics, each with a 3D digitizer and / or a camera, is possible. These are used to measure the same or different reference objects and their recognizable features in order to implement the method according to the invention.

[0068] For example, a combination with two or more robot kinematics, each with a reference object, is also possible. The recognizable features of the reference objects are recorded by a 3D digitizer and / or camera located and fixed in the vicinity of the robot kinematics, while the reference objects are brought into different poses relative to it.

[0069] Another exemplary combination can be that some robot kinematics are equipped with a 3D digitizer and some robot kinematics with a camera and a robot kinematics or some robot kinematics with a reference object, and these generate data, for example, after sub-step c).

[0070] A configuration of multiple robot kinematics is also possible, in which one, several, or all robot kinematics are equipped with a 3D digitizer and / or a camera as well as a reference object. The reference object can be attached anywhere on the robot kinematics.

[0071] With a combination containing multiple robot kinematics, it is possible to calibrate them simultaneously. It is also possible to calibrate an uncalibrated robot kinematics with a calibrated robot kinematics. The movement of the respective robot kinematics can be conveniently controlled by a controller that controls all robot kinematics.

[0072] In principle, it is possible to realize various combinations of robot kinematics with 3D digitizers, cameras and reference objects according to the method according to the invention.

[0073] According to a further embodiment, the robot kinematics or at least one of the robot kinematics is a turntable.

[0074] A turntable is particularly advantageous when 3D data is to be determined not only on the front of a measuring object, but also on the sides or back and the 3D digitizer is fixed stationary in the environment or another robot kinematics cannot reach the sides or back of the measuring object, which is made possible by the turntable.

[0075] The turntable can be calibrated individually or with another robot kinematics, for example a six-axis robot.

[0076] According to a further embodiment, the robot kinematics are measured relative to each other.

[0077] For example, if two or more 3D digitizers mounted on robot kinematics measure the same reference object fixed in the environment, this has the advantage that the coordinate systems of the end effectors of the respective robot kinematics can be converted into a common global reference coordinate system. This works equally well for the combination with two or more reference objects mounted on robot kinematics, as well as for the combination with a reference object mounted on a first robot kinematics and a 3D digitizer mounted on a second robot kinematics.

[0078] This means, for example, with regard to the measurement of a measuring object, that two or more robot kinematics, each with 3D digitizers attached to it, can be used and the surface data (3D data) of the measuring object of the respective 3D digitizers can be represented in a common global reference coordinate system.

[0079] Furthermore, it is possible to calibrate entire streets to robot kinematics and to transfer them into a common reference coordinate system.

[0080] According to a further embodiment, in addition to the 3D digitizer and / or camera and / or reference object, one or more further tools are attached to the robot kinematics and / or the 3D digitizer and / or the camera and / or the reference object on the robot kinematics are replaced by one or more tools.

[0081] This makes it possible for the method according to the invention to be used not only for 3D coordinate measurement technology, but also for other robot kinematics tasks.

[0082] Furthermore, it is also conceivable that, for example, the reference object attached to the robot kinematics for calibration is replaced by a measurement object and Subsequently, surface data of the measurement object in different poses can be generated using the 3D digitizer. This can be done by moving both the robot kinematics supporting the measurement object and, if the 3D digitizer is attached to a robot kinematics, the robot kinematics supporting the 3D digitizer.

[0083] According to a further embodiment, the reference object is designed such that the recognizable features are located on an independent body and / or a wall provided with recognizable features and / or on another robot kinematics and / or on one or more arbitrary members of a robot kinematics and / or on a 3D digitizer or a camera and / or on a tool and / or on a measurement object and / or around a measurement object.

[0084] Depending on the application, structure and environmental conditions of a robot kinematics, it is advantageous to create a reference object with a specific arrangement of the recognizable features for the method according to the invention.

[0085] The reference object can be an independent body. For example, it can be a wall with recognizable features, as described in EP 2 489 977 B1.

[0086] The reference object can also be located on another robot kinematic system; it can be attached to any robot link or extend over one or more links of the other robot kinematic system. It can also be located on the end effector of the other robot kinematic system or on a tool attached to the end effector. The reference object can consist of several partial reference objects. The partial reference objects can be spatially separated from one another. The reference object can also be part of the 3D digitizer. The reference object is normally a separate object from the measurement object. A combination of the two objects is also possible.

[0087] According to a further embodiment, the recognizable features are optical markers that themselves actively illuminate and / or are passively illuminated, and / or in the Two-dimensionally recognizable features with a unique 2D geometry and / or three-dimensionally recognizable features with a unique 3D geometry.

[0088] Depending on the type and application of a 3D digitizer or camera, it is advantageous to use corresponding recognizable features.

[0089] Features recognizable in two dimensions can be, for example, patterns, circles, holes, spheres, edges or corners; features recognizable in three dimensions can be, for example, spheres, cylinders, cones, etc.

[0090] According to a further embodiment, the camera is removed from the robot kinematics when it is not in use.

[0091] This has the advantage that the accessibility of the 3D digitizer z ference object and a measurement object. Thus, for example, if the measurement object is a car body, it is much easier to measure inside the car body if the camera has been removed from the robot kinematics for the measurement process.

[0092] It is understood that the above and the following The distinguishing features can be used not only in the combination specified, but also on their own or in other combinations without departing from the scope of the present invention.

[0093] In addition, it is understood that the features defined in the dependent claims for the method for calibrating the robot kinematics can also be used in the same or equivalent manner as device features for the method according to the invention, without these being listed again separately here as device features.

[0094] The indefinite term “a”, “an”, “anen” etc. is not to be understood as a number in the sense of the present invention, but as an indefinite term in the sense of “at least one”, "at least one," "at least one," etc. Thus, the features designated by "one element" also encompass several such elements, such as multiple robot kinematics or a robot kinematics with one or more 3D digitizers, cameras, or reference objects. Furthermore, other elements are not excluded.

[0095] Embodiments of the invention are illustrated in the drawings and explained in more detail in the following description. They show: Fig. 1 is a flowchart illustrating the embodiment of the method according to the invention; Fig. 2 is a schematic representation of a device of the method according to the invention in a first variant; Fig. 3 is a schematic representation of a device of the method according to the invention in a second variant; Fig. 4 is a schematic representation of a device of the method according to the invention in a third variant; Fig. 5 is a schematic representation of a device of the method according to the invention for measuring a measurement object; Fig. 6 shows a further schematic representation of a device of the method according to the invention with several robot kinematics; and Fig. 7 shows a further schematic representation of a device of the method according to the invention with a turntable and a six-axis robot.

[0096] In a first variant of the method according to the invention, in a sub-step 101a (Figure 1 ), a 3D digitizer 2 attached to a first robot kinematics 1a and I or a camera 4, wherein the camera 4 is attached to this first robot kinematics 1a or to a further second robot kinematics 1b, and a reference object 7 which is located in the vicinity of the robot kinematics 1a or in the vicinity of the robot kinematics 1a, 1b and is stationary for the recordings and which has recognizable features 8, is brought into different relative poses to one another by the robot kinematics 1a or the robot kinematics 1a, 1b and in these different relative poses data of the recognizable features 8 of the reference object 7 are recorded by the 3D digitizer 2 and / or by the camera 4.

[0097] An exemplary device of the first variant is shown in Figure 2. This device has a 3D digitizer 2 attached to a robot kinematics 1a with a measurement volume 3 and a camera 4 attached to a robot kinematics 1a and a reference object 7 located in the vicinity of the robot kinematics 1a, which reference object contains recognizable features 8. In this exemplary embodiment, the robot kinematics 1a is a six-axis robot. The axes are each marked with a hatched line. The device, as well as the exemplary embodiments mentioned below, can be controlled, for example, via a computer 5. The computer 5 contains a computer program with instructions which, when the program is executed by a processor, cause this one processor and / or further processors to carry out the method according to the invention.

[0098] The camera 4 can be disassembled after the images required by the camera 4 have been taken. This has the advantage of increasing the accessibility of the 3D digitizer 2 to the reference object 7 or to a measurement object 9 (Figure 5).

[0099] In a second variant of the method according to the invention, in a sub-step 101 b, a 3D digitizer 2 located and fixed in the environment of a robot kinematics 1a and / or a camera 4 fixed in the environment and a reference object 7 attached to the robot kinematics 1a, which reference object has recognizable features 8, are brought into different relative poses to one another by the robot kinematics 1a, and in these different relative poses, data of the recognizable features 8 of the reference object 7 are recorded by the at least one 3D digitizer 2 and / or the camera 4.

[0100] An exemplary device of the second variant is shown in Figure 3. This device comprises a reference object 7 attached to a robot kinematics 1a, which contains recognizable features 8, and a 3D digitizer 2 located and fixed in the vicinity of the robot kinematics 1a, and a camera 4 located and fixed in the vicinity of the robot kinematics 1a.

[0101] In a third variant of the method according to the invention, in a sub-step 101c, a reference object 7 which is attached to a first robot kinematics 1a and has recognizable features 8, and a 3D digitizer 2, wherein the 3D digitizer 2 is attached to a second robot kinematics 1b or to a third robot kinematics 1c, and / or a camera 4, wherein the camera 4 is attached to the second robot kinematics 1b or to the third robot kinematics 1c, are brought into different relative poses to one another by the robot kinematics 1a and the robot kinematics 1b and / or 1c, and in these different relative poses, data of the recognizable features 8 of the reference object 7 are recorded by the 3D digitizer 2 and / or the camera 4.

[0102] An exemplary device of the third variant is shown in Figure 4. This device has a reference object 7 attached to a robot kinematics 1a and a 3D digitizer 2 attached to a robot kinematics 1b with a measurement volume 3 and a camera 4 attached to the robot kinematics 1b.

[0103] The substeps 101a, 101b, 101c, which belong to a first step, are used to record data that will be evaluated in the next step.

[0104] In a further second step 102 (Figure 1), calibration data are generated from at least part of the recorded data, which relate the recorded features 8 of the reference object 7 to one another in a coordinate system, and calibration parameters of the robot kinematics 1a, 1b, 1c are derived.

[0105] In order to calibrate a robot kinematics 1a, 1b, 1c on the reference object 7, the reference object 7 must be measured with high precision. This can be achieved, for example, by It should be noted that the reference object 7 has been calibrated in advance in a calibration laboratory. If the reference object 7 does not remain calibrated with long-term stability, regular calibration is necessary to ensure lasting accuracy. The advantage of the method according to the invention is that the reference object 7 is calibrated with high precision by generating calibration data during the method according to the invention, thus eliminating the complexity of a preliminary calibration and long-term stability.

[0106] The reference object 7 can be designed such that the recognizable features 8 are located on an independent body and / or a wall provided with recognizable features 8 and / or on another robot kinematic system 1a, 1b, 1c and / or on one or more arbitrary members of a robot kinematic system 1a, 1b, 1c and / or on a 3D digitizer 2 or a camera 4 and / or on a tool and / or on a measurement object 9 and / or around a measurement object 9.

[0107] The recognizable features 8 can be optical markers that themselves actively illuminate and / or are passively illuminated, and / or features 8 that are recognizable in two dimensions with a unique 2D geometry and / or features 8 that are recognizable in three dimensions with a unique 3D geometry.

[0108] A scale-determining element 6 can be attached to the reference object 7 in order to be able to determine the distances of the recognizable features 8 absolutely and not only relatively. The scale-defining element 6 is shown in Figure 2, but could just as easily be included in Figure 3 or 4 for the device of the second or third variant of the method according to the invention. The scale-defining element 6 is, for example, a stable precision body. This is calibrated with high precision, usually using a tactile coordinate measuring machine, and consists of a material that hardly expands, ideally not at all, when the temperature changes. The precision bodies are usually simple forms of length scales, such as a ball rod. The advantage is that by generating the calibration data, the accuracy of the precision body is transferred to the rather complex reference object 7, without the reference object 7 itself having to be calibrated with high precision (e.g., tactilely).

[0109] A scale-defining element 6 can also be provided by the or one of the 3D digitizers 2, as shown in Figures 3, 4, 6 or Figure 7, in that the 3D digitizer 2 provides at least one recording of 3D data. An additional scale-defining element 6, as shown in Figure 2, is therefore not absolutely necessary. However, the additional scale-defining element 6 could also be omitted in Figure 2, i.e., the device for the first variant of the method according to the invention, if the 3D digitizer 2 is used as the scale-defining element 6. The 3D digitizer 2 can be the scale-defining element 6, since it absolutely measures the distances between the recognizable features 8 of the reference object 7 through at least one 3D data recording based on the calibration of the 3D digitizer 2.The recognizable features 8 of the reference object 7, which may be related to one another only from 2D data when generating the calibration data, can therefore be absolutely related to one another by combining them with at least one recording of 3D data from the 3D digitizer 2. It would also be conceivable for one of the robot limbs itself to be the scale-defining element. In this case, no further scale-defining element would be required for the method according to the invention.

[0110] The data recorded to generate the calibration data can come from the 3D digitizer 2 and / or the camera 4.

[0111] Preferably, a high-resolution digital camera is used to generate the calibration data, wherein said camera is brought into many different relative poses to the reference object 7 in order to record images of the recognizable features 8 of a reference object 7.

[0112] Typically, 2D data is used to generate the calibration data. However, it is also possible to use 3D data from a 3D digitizer.

[0113] The calibration parameters of the robot kinematics 1a and / or 1b and / or 1c are derived simultaneously with the generation of the calibration data. To derive the calibration parameters of the robot kinematics 1a, 1b, 1c, 2D data can be used, which are recorded, for example, with camera 4. Advantageously, 3D data recorded by a 3D digitizer 2, in particular a triangulation sensor, can also be used, since triangulation sensors can determine 3D data very precisely and can also be implemented relatively easily, for example by combining two cameras or a camera and a projector.

[0114] A stripe projection system, for example, can be used as a triangulation sensor. This consists, for example, of two cameras and a projector for projecting stripe images. Homologous image points—points that depict the same object point on a surface in both cameras—are triangulated, thus determining the object point in space. This generates a 3D coordinate in the local reference coordinate system (DCS) of the stripe projection system for the object point. Since a stripe projection system records a large number of homologous image points, it is considered an area sensor. This also allows 3D data of multiple recognizable features 8 of a reference object 7 to be generated in one pose of the stripe projection system.Due to the high accuracy of the 3D data of the fringe projection system in all three spatial dimensions, the pose of the fringe projection system relative to the reference object 7 can be determined very precisely, thus allowing the calibration of the robot kinematics 1a, 1b, 1c to be carried out very accurately. (See also Luhmann, Robson, Kyle, Boehm, Close Range Photogrammetry and 3D Imaging, Second Edition, Chapter 6, ISBN 978-3-1110-2986-3).

[0115] In this case, it may be useful to move to poses suitable for calibrating the robot kinematics 1a, 1b, 1c to record the recognizable features 8 of the reference object 7 with the 3D digitizer 2 and / or the camera 4. For example, poses are used that cover as many axis angles as possible or poses that generate different gravitational effects on the robot limbs or axes of the robot kinematics 1a, 1b, 1c. For generating calibration data, it may be useful to move to poses so that the recognizable features 8 of the reference object 7 are precisely measured.

[0116] However, it is also possible that poses are used which are suitable for both the generation of calibration data and the derivation of calibration parameters of the Robot kinematics 1a, 1b, 1c are suitable. The data recorded in this case can then be identical for generating the calibration data as well as for deriving the calibration parameters.

[0117] In general, the data used to generate the measurement data and to derive the calibration parameters can be different, identical, or partially identical. Furthermore, the chronological order in which the data are recorded is irrelevant.

[0118] A portion of the data of the recognizable features 8 of the reference object 7 present in the local reference coordinate system DCS of the 3D digitizer 2 or in the local reference coordinate system CCS of the camera 4 can be transformed into a global reference coordinate system, for example, RCS, RCS1 or RCS2, RCS3 or RCS4, whereby other reference coordinate systems, including non-global reference coordinate systems, are also conceivable. This particularly applies to the portion of the data used to derive the calibration parameters.

[0119] For the transformation of local data (actual observations) into a global coordinate system, for example, RCS, RCS1 or RCS2, RCS3 or RCS4, it is helpful to know the pose of the 3D digitizer 2 and / or the camera 4 attached to the end effector or the reference object 7 attached to the end effector. This results from the forward kinematics of the robot kinematics 1a, 1b, 1c and the relationship between the 3D digitizer 2 and / or camera 4 or reference object 7 and the end effector. The forward kinematics or forward transformation is the function that describes the position and orientation of the end effector or its coordinate system ECS as well as individual robot links in a given coordinate system, for example RCS, RCS1 or RCS2, RCS3 or RCS4, depending on the robot configuration given in the axis space (e.g. angular positions of rotary joints or linear positions of linear axes).To do this, the coordinate transformations of the individual robot elements must be known. These coordinate transformations are nominally known for a given robot, for example, from data sheets, technical drawings, CAD data, given Denavit-Hartenberg parameters, or similar sources. The data available in the local reference coordinate system of the 3D digitizer 2 or the camera 4 relate, for example, to the position and orientation. their camera sensors. In the case of a triangulation sensor, this is specified by the calibration of the triangulation sensor. The lengths and angles required for the transformation between the local reference coordinate system (DCS, CCS) of the camera sensor and the coordinate system of the end effector (ECS) can therefore nominally be read out, for example, from the CAD data of the 3D digitizer 2 and / or the camera 4 and any necessary adapters between the 3D digitizer 2 and / or camera 4 and the end effector. CAD data of the reference object 7 can also be used to establish a relationship between the end effector and the recognizable features 8 of a reference object 7 attached to the end effector. The relationship between RCS1 and RCS2 or between RCS3 and RCS4 can, for example, also be derived from CAD data or have been measured.

[0120] For information on how to perform coordinate transformations, see Luhmann, Robson, Kyle, Boehm, Close Range Photogrammetry and 3D Imaging, Second Edition, Chapter 2, ISBN 978-3-1110-2986-3.

[0121] In the following, the calculation of the forward kinematics from these nominally known values ​​(i.e., values ​​not calibrated to reality by any type of calibration) is referred to as "nominal forward kinematics." However, due to, for example, geometric inaccuracies, elastic or thermal deformations, or inaccuracies in determining the axis angles or lengths of the robot links, the accuracy of the position determination of the 3D digitizer 2 and / or the camera 4 or the reference object 7 on the robot kinematics 1a, 1b, 1c is usually insufficient. The nominal forward kinematics therefore differs from the real ones.

[0122] The data available after transformation into a global coordinate system, for example, RCS, RCS1 or RCS2, RCS3 or RCS4, are then used as starting values ​​for a subsequent optimization process. By transforming into a global coordinate system, RCS, RCS1 or RCS2, RCS3 or RCS4, using forward kinematics, it is more likely that a better extremum of the optimization process will be found.

[0123] In the optimization process, the measurement data of the recorded recognizable features 8 of the reference object 7 are generated and the calibration parameters of the robot kinematics 1a, 1b, 1c are derived and, optionally, the calibration parameters of the 3D digitizer 2 or the camera 4 are adjusted in such a way that the deviations between the coordinates of the recorded recognizable features 8 of the reference object 7 and the coordinates of the recognizable features 8 of the reference object 7 calculated from a mathematical model are minimized.

[0124] A deviation can be, for example, a deviation in the position and / or a deviation in the orientation of the recognizable features 8 of the reference object 7.

[0125] Within an optimization process, there are essentially two tasks. One is the generation of the calibration data using the data provided for generating the calibration data, and the second is the derivation of the calibration parameters for the robot kinematics 1a, 1b, 1c. Both are achieved simultaneously by simultaneously minimizing the deviations between the coordinates of the recognizable features 8 of the reference object 7 recorded for generating the calibration data and for deriving the calibration parameters, and the coordinates of the recognizable features 8 of the reference object 7 calculated from a mathematical model.The mathematical model can be a combination of a bundle block adjustment (see Luhmann, Robson, Kyle, Boehm, Close Range Photogrammetry and 3D Imaging, Second Edition, Chapter 4, ISBN 978-3-1110-2986-3) and a mathematical model that contains the calibration parameters of the robot kinematics 1a, 1b, 1c and thus describes the robot kinematics 1a, 1b, 1c, such as the forward kinematics mentioned above. This can be achieved, for example, using methods for solving a nonlinear optimization problem according to Nocedal, Wright, Numerical Optimization, Second Edition, ISBN 978-1-4939-3711-0.

[0126] If necessary, the calibration parameters of the 3D digitizer 2 and / or the camera 4 can also be adjusted, provided that a scale-defining element 6 is captured by the 3D digitizer 2 and / or the camera 4 (see Luhmann, Robson, Kyle, Boehm, Close Range Photogrammetry and 3D Imaging, Second Edition, ISBN 978-3-1110-2986-3) or, if necessary, the calibration parameters of the 3D digitizer 2 and / or the camera 4 can be adjusted without a scale-defining element 6 being detected by the 3D digitizer 2 and / or the camera 4, provided that one of the robot links and thus one of the robot kinematics 1a, 1b, 1c itself is the scale-defining element 6. Likewise, if required, scale-independent calibration parameters of the 3D digitizer 2 and / or the camera 4 can also be adjusted without a scale-defining element 6 being detected, provided that the scale-dependent parameters of the calibration of the 3D digitizer 2 and / or the camera 4 were adjusted using a scale-defining element.

[0127] The input parameters for the optimization process are the calibration parameters of the robot kinematics 1a, 1b, 1c obtained from the forward kinematics, the recorded data of the calibration data or of the camera > v. with an indication of whether these are to be used for generating the calibration data or for deriving the calibration parameters of the robot kinematics 1a, 1b, 1c, and, if required, calibration parameters of the 3D digitizer 2 or camera 4. The output parameters of the optimization process are the optimized calibration parameters of the robot kinematics 1a, 1b, 1c, the generated calibration data of the recognizable features 8 of the reference object 7 and, if required, the optimized calibration parameters of the 3D digitizer w. of camera 4.

[0128] With the method according to the invention, the robot kinematics 1a, 1b, 1c can be calibrated, the reference object 7 can be measured and the 3D digitizer 2 or the camera 4 can be calibrated.

[0129] By minimizing, an ideal solution to the optimization problem is advantageously found.

[0130] As an example, an optimization problem to be solved is explained for sub-step A 101a, in which a 3D digitizer 2 and a camera 4 are attached to the robot kinematics 1a. Pics.ij contains the coordinates of the i-th recognizable feature 8 of the reference object 7 for the j-th pose of the robot kinematics 1a in the local reference coordinate system DCS of the 3D digitizer 2 or CCS of the camera 4. The forward kinematics is described by the axis values ​​Jj belonging to the j-th pose. Axis values ​​can For example, an angle or a displacement between structural components of the robot kinematics 1a. The coordinates PGCS.IJ transformed into a global reference coordinate system, for example RCS, dependence of the calibration parameters p are obtained by applying a transformation function F p, which can result, for example, from the sequential execution of several transformations for the individual robot links. (1)

[0131] Thus, an optimization problem (2) (2) where QGCSXQ) are the coordinates of the i-th recognizable feature of the recognizable features 8 of the reference object 7, and these can also be adjusted using the parameters q. This makes it possible, as mentioned at the beginning, to correct errors in the measurement of the recognizable features 8, which are detected by deriving the calibration parameters. The parameters q can, for example, contain the coordinates of the features 8 of the reference object 7.

[0132] The optimization procedure and the identification of the calibration parameters can be formulated using forward kinematics. However, it is also possible to formulate an equivalent optimization problem using inverse kinematics.

[0133] The calibration parameters of the robot kinematics 1a, 1b, 1c can be, for example, geometric and / or elastic and / or thermal and / or calibration parameters of the at least one robot kinematics 1a, 1b, 1c, which take into account the configuration of the at least one robot kinematics 1a, 1b, 1c, and / or calibration parameters which take into account the direction of travel, and / or calibration parameters of the characteristics of the or calibration parameters for determining the error curve of the Length or angle measuring technology of the at least one robot kinematics 1a, 1b, 1c and / or calibration parameters that characterize the play of individual or all axes of the at least one robot kinematics 1a, 1b, 1c and / or calibration parameters that describe the dependence of the robot kinematics 1a, 1b, 1c on other variables that can be recorded, for example, with sensors or calculated based on the available data.

[0134] The distance and angular position of the two coordinate systems RCS1 and RCS2 or RCS3 and RCS4 to each other can also be regarded as a parameter of the robot kinematics 1a, 1b, 1c,

[0135] Depending on the type of robot kinematics 1a, 1b, 1c and the environmental conditions of the robot kinematics 1a, 1b, 1c as well as the type of 3D digitizer 2 or camera 4 and reference object 7, it is useful to use different combinations of calibration parameters for optimization.

[0136] However, the method according to the invention is not limited to these calibration parameters and can be supplemented and combined with other parameters.

[0137] The calibration parameters resulting from the optimization procedure can be used to determine the pose of the end effector and / or to align the end effector based on the parameters.

[0138] The adjusted calibration parameters resulting from the optimization process can be used to position the end effector of the robot kinematics 1a, 1b, 1c more precisely. This is useful if the measurement of a measurement object 9 takes place after the calibration of the robot kinematics 1a, 1b, 1c. The calibration parameters can also be used to precisely determine a pose of the end effector. This makes it possible, for example, to use a cost-effective robot kinematics 1a, 1b, 1c whose approach accuracy to a specific pose is low due to mechanical reasons, but the pose can still be precisely determined using the adjusted parameters. Furthermore, it is also possible to subsequently correct the poses of a measurement. Therefore, only data from the measuring object the 3D digitizer 2 in different poses of the 3D digitizer (taken, then the calibration parameters i joterkinematics land of the method according to the invention and, based on the calibration parameters derived thereafter, the recorded data of the measurement object 9 are converted into a global coordinate system, for example RCS, RCS1 or RCS2, RCS3 or »r.

[0139] As shown in Figure 5, a robot kinematics 1a, strengthened 3D digitizer 2 can be brought into different poses for measuring the measuring object 9 and the 3D digitizer 2 in the reference coordinate system the 3D digitizer 2 generates local 3D data and the local 3D data generated for the various poses based on the robot kinematics calibrated according to the method according to the invention a global coordinate system, for example RCS. Thus, during the measurement, the 3D data of a measuring object 9, in which a 3D digitizer 2 only captures partial areas 10 of the measuring object its measuring volume« can be determined more accurately overall by precisely converting the local 3D data into the global coordinate system RCS.

[0140] Furthermore, a measurement object 9 can be related to the robot kinematics 1a, 1b, 1c based on recognizable features 8' that are distributed on or around the measurement object 9 and by a 3D digitizer 2 ui attached to a robot kinematics 1a, 1b, 1c Camera 4 is captured, wherein the robot kinematics 1a, 1b, 1c have been calibrated according to the method according to the invention. This is useful, for example, if the measurement object 9 cannot be completely captured due to its size and the limited arm length of the robot kinematics 1a, 1b, 1c. As shown in Figure 5, the measurement object 9 can, for example, be located on a mobile platform 11 that has recognizable features 8', wherein the recognizable features 8' can differ from the recognizable features 8. A measurement sequence can, for example, proceed as follows: Providing a measurement object 9 attached to a mobile platform 11; First measuring the 3D data of the recognizable features 8' with the 3D digitizer 2 in the local reference coordinate system DCS of the 3D digitizer 2 and transforming the 3D data into a global reference coordinate system RCS; First measurement of the measurement object 9 with the 3D digitizer 2 attached to the robot kinematics 1a, 1b, 1c in different poses and for each measurement in the corresponding pose. Carrying out a transformation of the 3D data measured in the local reference coordinate system DCS of the 3D digitizer 2 into the global reference coordinate system RCS; Moving the measuring object 9 with the mobile platform 11 in order to be able to measure the previously inaccessible areas of the measuring object 9; Second measurement of the 3D data of the recognizable features 8' with the 3D digitizer 2 in the local reference coordinate system DCS of the 3D digitizer 2 and transformation of the 3D data into a global reference coordinate system RCS; Determining the displacement and rotation of the mobile platform 11 in the global reference coordinate system RCS based on the 3D data of the recognizable features 8' recorded during the first and second measurements; Second measurement of the measurement object 9 with the 3D digitizer 2 attached to the robot kinematics 1a, 1b, 1c in different poses and for each measurement in the corresponding pose. Carrying out a transformation of the 3D data measured in the local reference coordinate system DCS of the 3D digitizer 2 into the global reference coordinate system RCS and shifting and rotating the 3D data based on the determined shift and rotation of the mobile platform 11.

[0141] As an alternative to moving the measuring object 9 on the mobile platform 11, the robot kinematics 1a, 1b, 1c or only the robot kinematics 1a, 1b, 1c can also be moved.

[0142] A mobile platform 11 can also, if z For example, position data of the mobile platform 11 can be read out via the control of the mobile platform 11, even a robot kinematics 1a, and these according to the inventive Procedure can be calibrated so that the measured 3D data of the measuring object 9 are converted into a global reference coordinate system.

[0143] The methods and embodiments mentioned so far can also be applied to more bot kinematics For example, it is possible, as in Figure 6 shows that two 3D digitizers 2 with their reference coordinate systems DCS1 and DCS2 are connected to the end effectors with reference coordinate systems ECS3 and ECS4 each of a robot kinematic If the same reference object 7 is measured, as shown in Figure 6, it is possible to measure the robot kinematics 1a, 1b, 1c relative to each other, thus they have a common coordinate system The coordinate system RCS3 can therefore be transferred to the coordinate system RCS4 and vice versa, since the pose of the reference object 7 in RCS3 and RCS4 is derived from the calibration of the robot kinematics srproceeds.

[0144] In general, the setups shown in the application examples are not limited to one or two robot kinematics It is also possible to restrict entire streets to Robot kinematics u calibrate and convert them into a common reference coordinate system.

[0145] However, this also means that a measurement object 9 can be measured with several 3D digitizers 2 on robot kinematics 1a, 1b, 1c, and the measurement data can be transferred into a common reference coordinate system, for example RCS1 or RCS2 or RCS3 etc.

[0146] Another device example is shown in Figure 7. This shows a rotary table as the first robot kinematics 1a and a six-axis robot as the second robot kinematics 1b. The robot kinematics 1a, 1b can be controlled separately or by a controller located, for example, on a computer 5, as shown in Figure 7. If one of the robot kinematics 1a, 1b is already sufficiently well calibrated, it can be sufficient to to calibrate the robot kinematics 1 b, 1a according to the present procedure, i and orientation between the end effectors ECS1 and ECS2 or the position and orientation between CCS or DCS and ECS1. However, both robot kinematics 1a, 1b can also be calibrated simultaneously according to the present method and measured relative to each other, so that all calibration parameters of the robot kinematics 1a, 1b are determined at once and from this the position of the two end effectors ECS1, ECS2 relative to each other or the position and orientation of CCS, DCS relative to ECS1 can be determined. To carry out the method according to the invention, in this example the of the 3D digitizer 2. In its dual function, it provides 2D data, in particular for measuring the reference object Here, it is represented by two partial reference objects 7' located on the rotating plate. On the other hand, it generates data for the 3D digitizer to calculate 3D data, which is used in particular to derive calibration parameters. The coordinate systems CCS and DCS can also be identical. [0 For simultaneous calibration of both robot kinematics, i.e. c : ; :htisch and the . ' isachsroboters, the optimization problem can be supplemented accordingly. PLCS.IJ contains the coordinates of the i-th recognizable feature 8 of the reference object 7 for the j-th pose between the robot kinematics 1a and 1b in the local reference coordinate system DCS of the 3D digitizer 2 or CCS of the camera 4 and Q Lcs,i the coordinates of the i-th recognizable feature 8 of the reference object 7 in the local reference coordinate system of the reference object 7. The forward kinematics of the first robot kinematics described by the axis values ​​belonging to the j-th pose The forward kinematics of the second robot kinematics si described by the axis values ​​belonging to the j-th pose J2j- axis values ​​can, for example, be angles between structural components of the respective of 1b. The coordinate system integrated into a global reference Example RCS1 , transformed coordinates Q G cs,i,j depending on the calibration parameters s for robot kinematics 1a are obtained by applying a transformation function F liS . The global reference coordinate system RCS1 Coordinates PGCS.IJ > n Dependence of the calibration parameters p for Robot kinematics 1b are obtained by applying a transformation function F 2)P , which can result, for example, from the sequential execution of several transformations for the individual robot elements of robot kinematics 1b. This then also includes, for example, the transformation from RCS2 to RCS1. (3) (4)

[0148] Thus, an optimization problem (5) (5) to solve.

[0149] It should be noted at this point that the method according to the invention can be applied not only to 3D measurements. Thus, in addition to the 3D digitizer 2 and / or camera 4 and / or reference object 7, one or more additional tools can be attached to the robot kinematics 1a, 1b, 1c, and / or the 3D digitizer 2 and / or camera 4 and / or reference object 7 on the robot kinematics 1a, 1b, 1c can be replaced by one or more tools, so that other tasks can also be performed with the robot kinematics 1a, 1b, 1c.

[0150] It is also possible, for example, to replace the reference object 7 attached to the robot kinematics 1a, 1b, 1c for calibration with a measurement object 9, and then to generate surface data of the measurement object 9 in different poses using the 3D digitizer 2. In this case, both the robot kinematics 1a, 1b, 1c supporting the measurement object 9 and, if the 3D digitizer 2 is attached to a robot kinematics 1a, 1b, 1c, the robot kinematics 1a, 1b, 1c supporting the 3D digitizer 2 can be moved.

[0151] It is understood that the exemplary embodiments shown here serve merely to schematically illustrate the principle and embodiment of the present invention. Various functional and structural modifications are possible without departing from the scope of the present invention.

Claims

Patent claims 1. Method for calibrating at least one robot kinema^ 1b; 1c) comprising at least: - a first step comprising: A sub-step A (101a), in which at least one of a first robot kinematics fixed 3D digitizer (2) and / or at least one carnage where the on this first robot cinema sr on a further second robot kinematics ^is fixed, and at least one in the environment of the robot kinematics in the environment of robot kinematics nitive and stationary reference object (7) for the recordings, which has a recognizable feature jfweist, through the robot kinematics ler the robot kinematics (1 in different relative poses to each other and in which in these different relative ;en data of the recognizable characteristics reference object (7) by the at least one 3D digitizer (2) and / or by the at least one channel ' (taken; or A partial step which at least one in the vicinity of a first Robot kinematic individual and fixed : vitalize he has at least one fixed in the environment and at least one robot kinema established reference object (7), which has recognizable Features (8), by the first robot kinema in different relative poses to each other and in which in these different relative - Data of recognizable characteristics . ss at least one reference object (7) by the at least one 3D digitizer (2) and the at least one camera be taken; or a partial step at least one of a first robot kinematics fixed reference object (7) which has recognizable features (8), and at least one 3D digitizer (2), wherein the 3D digitizer (2) is connected to a further second robot kinematics or to a further third Robot kinematics (1c) is attached, and / or at least one camera (4), wherein the camera (4) is attached to the second robot kinematics (1b) or to the third robot kinematics (1c), are brought into different relative poses to one another by the first robot kinematics (1a) and at least one of the further robot kinematics (1b; 1c), and in which data of the recognizable features (8) of the at least one reference object (7) are recorded in these different relative poses by the at least one 3D digitizer (2) and / or the at least one camera (4);characterized in that in a second step (102), at least from a part of the recorded data, both calibration data are generated which relate the recorded features (8) of the at least one reference object (7) to one another in a coordinate system, and calibration parameters for the at least one robot kinematics (1a; 1b; 1c) are derived from at least a part of the recorded data.; 2. The method according to claim 1, wherein the at least one 3D digitizer (2) is a triangulation sensor, in particular one with stripe light projection 3. Method according to one of claims 1-2, wherein a scale-defining element (6) is present on the at least one reference object (7) and / or a robot member itself is a scale-defining element (6).

4. Method according to one of claims 1-2, wherein the recognizable features (8) of the reference object (7) are absolutely related to one another by at least one absolute measurement with the 3D digitizer and thus the 3D digitizer itself is the scale-determining element.

5. Method according to one of claims 1-4, wherein at least a part of the at least one 3D digitizer (2) in the local reference coordinate system and / or the data of the recognizable features (8) of the at least one reference object (7) present locally in the reference coordinate system (CCS) of the at least one camera (4) are transferred into a global reference coordinate system (RCS).

6. The method according to any one of claims 1-5, wherein the generation of the calibration data of the recorded features (8) of the reference object (7) and the derivation of the calibration parameters of the at least one robot kinematics (1a; 1b; 1c) and optionally an adaptation of the calibration parameters of the at least one 3D digitizer (2) or the at least one camera (4) are carried out in such a way that the deviations between the coordinates of the recorded recognizable features (8) of the reference object (7) and the coordinates of the recognizable features (8) of the reference object (7) calculated from a mathematical model are minimized.

7. The method according to claim 6, wherein the calibration parameters of the robot kinematics (1a; 1b; 1c) are geometric and / or elastic and / or thermal and / or calibration parameters of the at least one robot kinematics (1a; 1b; 1c) which take into account the configuration of the at least one robot kinematics (1a; 1b; 1c), and / or calibration parameters which take into account the direction of approach, and / or calibration parameters of the characteristic of the orcalibration parameters for determining the error curves of the length or angle measuring technology of the at least one robot kinematics (1a; 1b; 1c) and / or calibration parameters that characterize the play of individual or all axes of the at least one robot kinematics (1a; 1b; 1c) and / or calibration parameters that describe the dependence of the robot kinematics (1a; 1b; 1c) on other variables that can be recorded, for example, with sensors or calculated based on the available data.

8. Method according to one of claims 6-7, wherein the calibration parameters are used to determine the pose of the at least one end effector of the at least one robot kinematics (1a; 1b; 1c) and / or to determine the at least one End effector of the at least one robot kinematics (1a; 1b; 1c) is to be aligned using the calibration parameters.

9. Method according to one of claims 1-8, wherein at least one 3D digitizer (2) and / or at least one camera (4) and / or at least one reference object (7) are mounted on a plurality of robot kinematics (1a; 1b; 1c) and wherein at least one of the robot kinematics (1a; 1b; 1c) is calibrated according to one of claims 1-8.

10. The method according to claim 9, wherein the robot kinematics (1a; 1b; 1c) are calibrated relative to one another.

11. Method according to one of claims 1-10, wherein the robot kinematics (1a) or at least one of the robot kinematics (1a, 1b, 1c) is a turntable.

12. The method according to any one of claims 1-11, wherein in addition to the at least one 3D digitizer (2) and / or the at least one camera (4) and / or the at least one reference object (7), one or more further tools are attached to at least one robot kinematics (1a; 1b; 1c) and / or wherein the at least one 3D digitizer (2) and / or the at least one camera (4) and / or the at least one reference object (7) are replaced by one or more tools on at least one robot kinematics (1a; 1b; 1c).

13. Method according to one of claims 1-12, wherein the at least one reference object (7) is designed such that the recognizable features on an independent body and / or a wall provided with recognizable features (8) and / or on another robot kinematics (1a; 1b; 1c) and / or on one or more arbitrary members of a robot kinematics (1a; 1b; 1c) and / or on a 3D digitizer between a camera (4) and / or on a tool and / or on a measuring object (9) and / or around a measuring object (9).

14. Method according to one of the claims where the recognizable features (8, 8') optical markers that actively glow are passively illuminated, features (8, 8') recognizable in two dimensions with a clear 2D geometry and / or features (8, 8') recognizable in three dimensions with a clear 3 metry are.

15. Method for determining the 3D data of a measuring object where at least one connected to a robot kinematics (1s 3D digitizer (2) or several attached to robot kinematics (1a; fixed 3D digitizers (2) are brought into different poses for measuring the measuring object (9) and the 3D digitizer(s) (2) in the reference coordinate systems (DCS, the 3D digitizer (2) generates local 3D data and wherein the local 3D data generated for the various poses are converted into a global coordinate system (R( , RCS2, RCS3, RCS4).

16. Method for the relative alignment of a measuring object (9) in relation to a robot kinematics (1a; based on recognizable features (8') that are distributed on or around the measuring object (9) and are determined by at least one robot kinematics (1c .italizer (2) and / or at least one on a robot kinematics, ©fixed camera wherein the calibration parameters of the at least one robot kinematics (1a; 1b; 1c) were derived by means of a method according to one of claims 1-14.

17. Device for calibrating at least one robot kinematics (1a; 1b; 1c) comprising: at least one on a first robot kinematics solidified 3D digitizer (2) i 1 ler at least one camera (4), wherein the camera; on this first robot kinematics on a further second robot kinematics is maintained, and at least one in the vicinity of the Robot kinematics (1a) or in the environment of the robot kinematics (1a;1b) and stationary for the recordings, which has recognizable features (8), or at least one 3D digitizer (2) located and fixed in the environment of a first robot kinematics (1a) and / or at least one camera (4) fixed in the environment and at least one reference object (7) attached to the first robot kinematics (1a), which has recognizable features (8), or at least one reference object (7) attached to a first robot kinematics (1a), which has recognizable features (8), and at least one 3D digitizer (2), wherein the 3D digitizer (2) is attached to a further second robot kinematics (1b) or to a further third robot kinematics (1c), and / or at least one camera (4), wherein the camera (4) is attached to the second robot kinematics (1b) or to the third robot kinematics (1c), wherein for at least one of the robot kinematics (1a; 1b;1c) calibration parameters for said robot kinematics (1a; 1b; 1c) are derived by means of a method according to any one of claims 1-14; 18. Device according to claim 17 comprising at least one scale-defining element (6).

19. Robot kinematics (1a; 1b; 1c), the calibration parameters of which were derived by means of a method according to one of claims 1-14.

20. A computer program comprising instructions which, when the program is executed by at least one processor, cause this at least one processor and / or further processors to carry out the method according to one of claims 1 to 16.

Images