WO2011110279A1

WO2011110279A1 - Method for calibrating a robot

Info

Publication number: WO2011110279A1
Application number: PCT/EP2011/000726
Authority: WO
Inventors: Franz Ehrenleitner; Carsten Meyer; Robin Kaufmann
Original assignee: Eisenmann Ag
Priority date: 2010-03-10
Filing date: 2011-02-16
Publication date: 2011-09-15
Also published as: DE102010010920A1

Abstract

The invention relates to a method for calibrating a robot which is equipped by the manufacturer with an operating program, wherein under the control of said operating program the tool centre point (TCP) is guided on a trajectory defined by a plurality of support points. For this purpose, the support points (S_i) which are relevant for a specific application are determined. Before or after this, the errors of the robot are ascertained by measurement. Then it is calculated which real points (S_r) are actually actuated by the TCP if the ideal support points (S_i) are input into the operating program. The vectors (V₁, V₂) which connect the real support points (S_r) to the ideal support points (S_i) are computationally doubled via the ideal support points (S_i), whereby the coordinates of substitute support points (S_e) are obtained. If the coordinates of said substitute support points (S_e) are input into the operating program, in spite of the existing robot errors the TCP passes through the desired ideal trajectory.

Description

Method for calibrating a robot

The invention relates to a method for calibrating a robot which is equipped by the manufacturer with an operating program under whose regime its "Tool Center Point" (TCP) is guided on a trajectory defined by a plurality of interpolation points.

When using industrial robots, a distinction is made between direct and indirect programming. Direct programming, also called "on-line programming", mainly works according to teach-in procedures. Here, by means of a keyboard, for example, a handheld programmer, the industrial robot in a Cartesian coordinate system or axis-specific approached to the desired positions and the results are stored in the controller. The advantages of this direct programming method _. Are, _^ that the industrial robot by means of the workpiece can be positioned and that additional statements can be entered directly. The disadvantage is that the industrial robot fails during program creation for production.

Therefore, an attempt was made at an early stage to shift the programming to areas outside of production and to carry out indirect programming, also called "off-line programming". For about 20 years, there are systems that allow the graphical-interactive simulation and programming of industrial robots in a 3-dimensional representation. Here is one

Model of the industrial robot, the manufacturing cell in which the industrial robot is arranged, and the Workpieces created, with the help of which then programming tasks can be performed.

The off-line simulation systems allow the direct

Description of the nominal path in the basic coordinate system

of the industrial robot. However, direct adoption of simulation-based programs into manufacturing is often not possible because reality and ideal model differ widely due to inaccuracies and modeling errors.

For this reason, it is necessary to calibrate industrial robots in order to achieve the required high absolute accuracy. The application of automated measurement methods with sufficient resolution is essential.

For the determination of the robot parameters from the measured data, almost exclusively systems theory

Approaches used. Here the parameters of the

Overall model identified using numerical methods. In addition, the deviations of the robot from

the target position measured at a sufficient number of spatial positions. Thereafter, an attempt is made, the specific form of the Euclidean total deviation - starting from Idealmo ^¬ dell - by varying the model parameters to be minimized. The achievable limit accuracy depends on the one hand on the number and the position of the spatial positions and on the other hand on the result variables, for example the arm lengths, angular deviations, axle stiffnesses and so on. This known method of calibration requires

not only a very high measurement effort. In addition, the mathematical processing is extremely complex and also requires the ability to intervene in the provided by the manufacturer of the robot operating program. Object of the present invention is to provide a method of the type mentioned in such a way that it can be easily and basically performed without interference with the operating program provided by the manufacturer.

This object is achieved by a) the bases are determined in a concrete application; b) the errors of the robot determined by measurement

become; c) is calculated, which real bases (S) are actually approached by the TCP, if the ideal

Substitute points (p.) Are entered in the operating program; the vectors which connect the real vertices (S ^) to the ideal vertices (S ^) beyond which the ideal vertices (S ^) are computationally doubled, whereby the coordinates of replacement vertices (S _e ) are obtained, - for execution the movement of the TCP under operating conditions, the coordinates of spare bases (S _e) are inputted to the operation provided by the manufacturer ^¬ program.

In accordance with the invention, therefore, the entire working space which can be reached by the robot's TCP is to be calibrated in a certain way in advance. Rather, the calibration is done in view of a very specific Trajectory defined by concrete bases for a single application case. Taking into account the errors of the robot, it is determined which interpolation points would be approached by the robot if, in fact, the coordinates of the desired interpolation points were entered unchanged in the operating program. From the deviation of the real interpolation points from the ideal interpolation points determined in this way, the coordinates of such points, referred to here as "substitute interpolation points S _e ", can be determined, in the input of which in the operating program of the robot the ideal interpolation points are achieved despite the existing robot errors become.

The method according to the invention has the great advantage that the robot can be operated by the machine-operator in the manner familiar to him and that, in particular, an intervention in the operating program of the robot provided by the manufacturer is not required. Once the errors of a particular robot have been determined, an auxiliary program preceding the operating program of the robot can be set up, in which case it is sufficient to enter only the interpolation points desired in the individual application. This utility then automatically calculates the replacement bases and forwards them to the operating program of the computer.

Expediently, the model-specific errors and the individual errors of the robot are determined separately and taken into account mathematically. This has the advantage that the calibration of several robots, which belong to the same series and all have the same type-specific error, is simplified: Since the model-specific errors - separately - already known, it is sufficient for the other robots of the same series only still the individual errors to be determined and included.

It has been shown that in many cases it is sufficient to take into account only axial position and length errors in the individual errors. Experience has shown that these errors contribute the most to the individual errors. If one reduces the consideration of individual errors to axial misalignment and length errors or even only to axial position errors, the metrological effort is reduced significantly; the mathematical treatment can be closed in many cases, resulting in an increase in accuracy

and reducing the computational burden contributes.

In order to determine the axial position errors, it is sufficient if the robot arm is rotated about each axis and in each case three points lying on a circular arc are measured, from which the axial positions and orientations can be calculated. This means in total that with six axes only 18

Measuring points must be measured.

To achieve sufficient accuracy, it is desirable that the angular distance between the two farthest from each other be measured

Points at least 90 °, preferably at least 120 °.

In principle, it is conceivable that the measurement of an individual error is influenced by other individual errors, so that therefore the contributions of different individual errors can no longer be kept separate. This complicates the mathematical consideration of individual errors. Such a "crosstalk"

of other individual errors than the one who is about to be measured is therefore to be avoided as far as possible. When determining the axial position errors, this can be especially dere by the fact that the robot arm in the

Twist is held in such a Messpose in which other existing individual errors have no appreciable influence on the traversed circular path. The decisive factor is that the other individual errors, which are not known in detail, do not significantly distort the circular path as such

Position of the individual measuring points on the circular path

can change without any significant influence on the determined axle position.

An embodiment of the invention will be explained below with reference to the drawing; FIG. 1 shows a flowchart of the calibration method according to the invention;

FIGS. 2 and 3

Measuring poses, which a robot arm can take to measure errors of the robot to be calibrated;

Figure 4 shows the model of a faulty robot. The inventive and subsequently with reference to FIG

1 explained method is used to calibrate a

Robot, either new delivered by the manufacturer or made to the changes, so

that a calibration is required. The robot may in particular belong to a specific series, as is generally the case.

The robot will not be in the entire workspace,

from his "Tool Center Point" (TCP) approached

can be calibrated. Rather, in a procedural Step 1 defines a plurality of specific, "ideal" support points, which define a trajectory to be traversed by the TCP. The robot is regularly provided by the manufacturer with an operating program which interpolates a trajectory between the interpolation points and generates control signals which guide the TCP along this calculated trajectory.

First, the errors that the real robot makes when driving the ideal are determined as follows:

The multitude of mistakes robots afflict

are roughly divided into two groups: The first group of mistakes includes all those who are determined by the series, so all

Common to robots that belong to a specific series and that are design-related by this series. These include, for example, joint elastics with and without load or elasticities. These errors are referred to here as "model-specific".

The other group of errors concern individual errors of the individual robot, ie those errors in which

The considered robot also of robots of the same

Series differentiates. These include, in particular, axis position errors, arm length and angle errors, temperature influences, transmission errors and stochastic errors. In the method step represented by box 2, first of all all model-specific errors are determined by measuring. This can be done in a known manner, which will not be discussed in detail here. In a computing step, represented by the box 3, is

Determines where in real robot when controlling the desired support points S ^ the TCP would be in consideration of the model-specific errors. The tat ^¬ neuter location of these "real" points will

generally different from the desired position. This as well as the mathematical correction will be explained with reference to FIG. 4:

In this figure 4, the robot arm is schematically shown in FIG

represented three different positions. The middle

Position represented by solid lines represents the "ideal" robot as laid down in the manufacturer's model of the robot. Ideally, the TCP should reach the base S ^ at the end of the robot arm. In fact, the bill results

taking into account the model-specific errors that were determined in step 2 that when input

the values of the base S ^ in the program prepared by the manufacturer ^¬ set the base S is achieved.

From the real base, a vector V leads to

ideal base S ^. If the vector V ₁ is now doubled from S ₁ to S ₁ , one arrives at one

Base S _e , here called "substitute base"

becomes. The name "Ersatzstützpunkt" derives from the fact that when entering its coordinates in the operating program of

Computers the ideal base S ^ reached by the TCP

becomes.

In the calculation step 3, the replacement support points S _{g are} determined in a corresponding manner for all interpolation points S 1 which were determined in method step 1.

The described measurements (box 2) and calculations

(Box 3) need for all robots of a series

to be done only once. This means that after the first calibration of a robot from a

certain series, the corrections for a second or third robot of the same series are already known.

The next step, represented by box 4, is the determination of the individual errors of the respective robot to be calibrated. How this can be measured in detail, is in principle irrelevant in carrying out the OF INVENTION ^¬ dung. However, it is desirable that the number of measurements and the computationally processed quantities be kept as low as possible. Measurements and tests have shown that 80 to 90 percent of the axis position errors alone contribute to the individual errors of the robot, while arm length and angle errors contribute 5 to 10 percent of the total error. Therefore, it is generally possible to measure only the errors mentioned in step 4 in preparation for the mathematical rule ^¬ step. 5

When determining the individual error should be taken to ensure that by taking appropriate

Measuring poses of the robot arm possible only one type of error is detected, so that a mixing of different error influences in the measurement result is avoided. As

This can be done in the measurement of Achslagenfehlern, will become apparent from Figures 2 and 3:

First, the robot is the manufacturer of the

Provided system, for example, by the robot manufacturer set zero marks, brought to zero. Then the individual deviations of the robot are determined by single-axis measurement. This happens because the axes of the robot are moved individually and from three measuring points, which lie on a circular arc around the real axis, the respective real axis position is determined. The measuring techniques used here are not of interest in the present context. In order to achieve sufficient accuracy, the three measuring points measured for each axis lie at an angular distance of 60 ° from each other.

FIG. 2 shows the measurement of the first axis of the robot arm. The measuring pose is chosen by others

Errors as by deviation of the position of the first axis, although possibly the actual position of the circular path, which is traversed during the measurement, deviates from the expected circular path. Nevertheless, the traversed by the TCP path remains a circular path whose axis largely

coincides with the real position of the axis 1. Stochastic defects, for example in the individual joints, in this way have virtually no effect.

FIG. 3 shows how the axis 3 of the robot can be measured. Even with the measurement pose shown here, it is apparent that in the presence of others

Although the individual measuring points do not match errors as errors of the position of the axis 3 with the desired ideal measurement punk ^¬ th; the actual measured points but remain substantially lie on the desired circular path so that the detected position of the axis 3 with the tatsäch ^¬ union position of the axis 3 corresponds largely.

After in process step 4 the individual error

are determined, they are in the calculation step, which is represented by the box 5, taken into account.

The procedure is the same as explained above for the calculation step represented by box 3:

It is first calculated how the position of the ideal support point S. changes due to the individual error. Once again, a real interpolation point S results, which is connected by a vector V ₂ to the ideal interpolation point. By doubling this vector V ₂ , one obtains the replacement support point S.

Finally, the doubled vectors 2 ^ associated with the model-specific errors and the doubled vector V due to the individual errors of the robot are added together. It thus results

Coordinates for a total replacement replacement point that can be entered in step 6 in the manufacturer-provided operating program of the robot. Using these substitute bases S _e , the operating program of the robot generates the signals, which, despite the existing model-specific errors and individual errors, guide the TCP exactly along the desired trajectory.

Claims

claims

1. Method for calibrating a robot that is of the

Manufacturer is provided with an operating program under whose regime its "Tool Center Point" (TCP) is maintained on a trajectory defined by a plurality of vertices; characterized in that the bases (S ^) in the concrete application

be determined; the errors of the robot by measurement ermi

become; c) which real bases (S) are calculated

actually be hit by the TCP, though

iiddeeaallstatezupptots (p.) are entered into the operating program; d) the vectors (V ^ V ^), which connect the real vertices (S _r ) to the ideal vertices (S ^), are computationally doubled beyond the ideal vertices (S ^), whereby the coordinates of substitute vertices (S _e ) be won; e) to carry out the movement of the TCP under conditions of use, enter the coordinates of the replacement support points (S _e ) in the operating program provided by the manufacturer.

2. The method according to claim 1, characterized in that the model-specific errors and the individual errors of the robot are determined separately and taken into account mathematically.

3. The method according to claim 2, characterized

that in the individual errors only Achslagen- and

Length errors are taken into account.

4. The method according to any one of the preceding claims,

characterized in that for determining the

Axis position error of the robot arm is rotated about each axis and each three lying on a circular arc

Points are measured, from which the axial positions and the axial directions can be calculated.

5. The method according to claim 4, characterized

that the angular distance between the two furthest apart points to be measured is at least 90 °, preferably at least 120 °.

6. The method according to claim 4 or 5, characterized in that the robot arm is held in the rotation in such a measuring poses, in which other existing individual errors have no appreciable influence on the traversed by the "Tool Center Point" (TCP) circular path.