FR2638115A1 - Method for setting a robot arm and robots using this method - Google Patents

Abstract

Method for setting a robot by determining the zero-adjustment errors in the angle encoder of a robot arm, at at least two segments which are articulated in series on parallel axes and each one including an angle encoder, comprising, in series, a shoulder joint, and arm segment, an elbow, a forearm, having a wrist, and a movable reference point on the forearm. According to the present invention the method comprises the following steps: - defining a fixed reference point R at a position which is fixed with respect to the axis O of the shoulder joint; - moving the movable reference point into coincidence with the fixed reference point R, the arm successively assuming two possible symmetrical configurations; - reading the two angles b1, b2 of the elbow, and, - determining the zero error in the angle encoder of the elbow as being equal to half the difference between the sum of the two angle readings and the theoretical sum.

Description

METHOD OF SETTING A ROBOT ARM AND
ROBOTS USING THIS PROCESS
The present invention relates to a method for calibrating the positioning of a robot arm and to robots implementing said method.

ENVIRONMENT OF THE INVENTION
The use of robots is widespread on the industry. Many robots have an arm, comprising a chain of at least two articulated segments, namely a “forearm” segment, pivotally mounted on a support by a “shoulder” articulation and a “forearm” segment. '' pivotally mounted on the "forearm" by an "elbow" joint, any tool being mounted at the end of the "wrist" of the "forearm" segment. For the sake of simplicity and consistency, we will denote by "robot arm" all the lever or connecting rod segments and their joints, by "arm" the segment connected to the shoulder, and by "forearm" the last segment of the chain carrying the tool in one "hand", via the "wrist".

It is common practice, both for the shoulder joint and for the elbow joint, to rotate these around parallel axes (which are usually vertical) and move the arm over a work surface (which is usually horizontal). The joints include angle encoders connected to inform a processor of the angles of the joints, and drive means are provided to modify the angles of the joints under the control of the processor, in order to move the "hand" relative to the work surface.

Although the prior multi-segment robot arms exhibit a high degree of repeatability when they execute the same movement sequences repeatedly, they suffer from a surprising lack of absolute precision.

In the past, this defect was corrected by the fact that robots were frequently "trained" in their movements by an experienced worker who physically moved the robot's hand into the required positions, the robot repeating the movements it had learned from worker. Either way, a high degree of repeatability is required.

However, there is also a need for a robot arm which is capable of directly executing instructions generated by so-called CAD (Computer Aided Design) systems, without an intermediate step, during which a worker holds the robot's hand. It is this application that requires a high degree of absolute precision.

The object of the present invention is to provide a method improving absolute precision, so as to obtain improved results when the arm is used under the direct control of a CAD system.

The invention is based on the fact that a significant source of errors affecting absolute precision (but not relative precision) lies in the zero settings from which the angle encoders operate, and on the fact that the The zeroing error of each angle encoder can be measured independently of the zero error on the other encoders of the same arm. Once the zeroing errors are known, various steps can be taken to correct the errors, or simply to take them into account.

SUMMARY OF THE INVENTION
The present invention provides a method for determining the error of zero in one or more of the angle encoders of a robot arm, comprising at least two segments, articulated in series around articulations having parallel axes and each comprising a angle encoder, the method being applicable to any chain of two consecutive segments comprising a shoulder joint, an arm segment, an elbow joint, a forearm segment and a reference point close to the wrist end of the forearm, the reference point being constituted by a fixed point on the longitudinal axis of the forearm segment, while being called the "mobile" reference point, since the arm is moves, the method comprising the following steps - definition of a fixed reference point, this point is
lying in a fixed position with respect to the axis of
the shoulder joint - displacement of the mobile reference point towards said point of
fixed reference, the arm taking each of the two
possible symmetrical configurations - reading of the two possible configurations; and - calculation of the error of zero in the angle encoder
the elbow joint, from the readings taken in
both configurations by exploiting the property of
symmetry of the two possible configurations, so that
any effect due to a possible zero error in the encoder
angle of the shoulder joint is canceled.

According to a first aspect of the invention, the mobile reference point is brought into exact coincidence with the fixed reference point, for example by physically engaging a ball carried by the mobile arm in a fixed seat; the angle read on the elbow angle encoder is read in each of the two configurations, and the error of zero in the elbow angle encoder is equal to half the difference between the sum of the two readings of angle and the expected sum (which is equal to 00 or 3600, depending on the conventions used to express the readings).

This first aspect can be used to discover a zero angle error of arbitrary magnitude. It is completely independent of any angle zero error from the shoulder encoder, and it does not use any additional information such as the length of the forearm segment.

In a second aspect of the invention, the arm is controlled to occupy the angles which should theoretically cause the displacement of the mobile reference point opposite the fixed reference point in the two possible configurations, and the difference is measured between the two positions effectively taken by the mobile reference point, the zero error of the elbow angle encoder then being equal to a trigonometric function of the difference measured between the two positions, divided by the length of the forearm segment, the specific trigonometric function depending on the conventions used to measure the difference in positions.

This second aspect is based on precise knowledge of the length of the forearm, but this length is generally known in practice, with sufficient precision.

The most advantageous characteristic of the above method, in one or other of its aspects, is that it makes it possible to measure the errors of possible zeroing of the elbow by completely eliminating the error of zero. of the shoulder angle encoder. Once the elbow error is known and taken into account, it becomes possible to estimate the error of zero on the shoulder in various conventional ways.

However, if the robot arm includes additional segments in series, not only can this method be applied to any couple of consecutive segments, but it can also be applied to the couple of the two extreme segments, which are then locked in a known configuration and treated together as a simple forearm segment, to measure the error of zero at the next articulation along the chain and so on.

This has the advantage of always using the same mobile reference point.

The invention also relates to a robot, in which said method is implemented in order to reduce or eliminate the effects of a zero error, at least in the bend angle encoder.

BRIEF DESCRIPTION OF THE DRAWINGS
Two embodiments of the invention will be described, by way of example, with reference to the figures of the accompanying drawings which represent geometric diagrams to describe the principle of the invention in which - Figure 1 shows schematically the first embodiment - FIG. 2 shows schematically the second embodiment,
DETAILED DESCRIPTION
FIG. 1 is a plan view, representing a robot arm with two segments, comprising an arm 10 and a forearm 12.

One end of the arm 10 is pivotally mounted on a fixed structure (not shown), by means of a shoulder articulation with a vertical axis at the point of origin 0. The opposite end of the arm 10 is pivotally mounted on one of the ends of the forearm 12, by means of an elbow joint with a vertical axis disposed at the first intermediate point li. The opposite end of the forearm segment 12 (that is to say his wrist), comprises a reference point which is brought into coincidence with a fixed reference point R, the position of which relative to the origin O is fixed, but does not require any known in absolute terms with a high degree of precision. For example, origin O can be defined relative to a work table, and the fixed reference point can between realized in a block which is temporarily fixed to the work table, only during the arm calibration.

For the sake of generality, the arm segment 10 has a length "L" different from the length "l" of the forearm segment, but the invention is also applicable to the case of two segments having an equal length.

Each of the articulations includes an angle encoder (not shown) which delivers a signal representative of the angle in this articulation. The angle at the point of origin O is measured with respect to an arbitrary fixed direction 14.

The arm segment is shown as forming an angle "a" with respect to direction 14. The angle "b" of the elbow joint is measured with respect to direction 16 in which the front segment extends -arms 12. By convention, the angle of the elbow is equal to zero when the two segments 10 and 12 are aligned, and also by convention, the two angles are measured in the direct trigonometric direction from the respective reference direction. Other conventions can be adopted without modifying the basis of the process.

The method consists in establishing a reference point R which is fixed with respect to the shoulder joint at the origin 0, and in causing the mobile reference point at the end of the wrist to coincide with the reference point. fixed R in each of the two possible configurations of the robot arm.

The final step of the method consists in comparing the sum of the two angles given by the elbow angle encoder for each of these positions, with the expected geometric sum, and in deducing the zeroing error from the difference resulting.

In FIGS. 1 and 2, a first configuration is drawn in solid lines (references 10 and 12). The other possible configuration is the specular image of the first around an imaginary line joining the point R of reference to the origin 0, and is drawn in broken lines. The arm segment and the forearm segment in the second configuration bear the references 20 and 22 respectively, and the second intermediate point bears the reference I2.

The second bend angle b2 is measured in the direct direction, starting from the direction 26 ′, in which the rear arm 20 extends.

Alternatively, another convention is to measure the second bend angle b2 clockwise, and give it a negative sign to get bt2. Since the lengths of the segments do not change, we can see by symmetry that bl has the same value and is of opposite sign to b'2, so that the sum bl + b'2 is equal to zero, or that, alternatively, the sum bl + b2 is equal to 3600 or 2 T radians. It will be observed that this applies independently of the values of the angles a1 and a2 of the shoulder joint.

The geometric discussion above relates to an ideal situation. In a real robot arm, the angles specified above are known only with finite precision. A common type of angle encoder includes a coding disc with multiple radial marks and an optical up and down counting system. Angles can be measured in this way with a resolution of around 6000 divisions per full revolution. This resolution sets a limit on repeatability.

An independent problem lies in the precision on zeroing. It often happens that the direction in which a robot angle encoder indicates a zero angle differs, in fact, from the true direction of zero by a deviation which is greater than the resolution limit. Such a deviation is a source of error in determining the absolute position of the robot arm.

Since the deviation from zero of the elbow angle encoder is a small angle d in the positive direction, as shown in an exaggerated manner at the intermediate points 11 and 12 with respect to the erroneous directions of reference 16 'and 26', the first reading given by the elbow angle encoder is
(1) al = bl - d.

Similarly, the second reading given by the same encoder is
(2) a2 = b2 - d.

The addition of equations (1) and (2) gives
(3) al + a2 = bl i b2 - 2d.

So, using the first convention mentioned above
(4) d = g360 (al + a2) / 2, or using the second convention
(5) d = - (al + a2) / 2, we can observe that, in this first embodiment, the precision of the result obtained depends only on the precision with which the reference points coincide, and on the reading precision of the elbow angle. In particular, the length 'll't of the forearm has no effect on the accuracy of the final result. As a result, the movable reference point can be mounted by means temporarily attached to the forearm. The fixation only needs to be located on the longitudinal axis of the forearm segment and to be ensured only for the time necessary for the implementation of the method, but it does not need to be positioned precisely on the 'axis.

Such a temporary means for achieving the mobile reference point can be constituted by a hard steel ball at the end of a rod and the corresponding fixed reference point can be produced by a V-groove or a pyramid imprint constituting a seat in a metal block temporarily fixed on the work surface which, in turn, is fixed relative to the origin of point 0. A coincidence is ensured by pressing the ball in the groove or in the impression.

In the second embodiment, the fixed reference point R is a purely theoretical point Ro, the coordinates of which in X and Y, XO and YO, are given by the equations
(6) X = - L cos al - l cos cl.

 (7) YO = L sin al + 1 sin cl.

or as follows
(8) XO = - L cos a2 - 1 cos c2.

 (9) YO = L sin a2 + 1 sin c2.

in which: ci = ai + bi, the Y axis being considered as positive going downwards from the origin 0, while the X axis is taken as being positive from the line of zero conventionally.

Two attempts are made to move the mobile reference point, so that it coincides with the theoretical point R ,.

Unfortunately, the theoretical values of a and b cannot be obtained in practice. By using the convention that the zero error on b is a small positive angle d, and disregarding a possible zero error on a, it is clear that the moving reference point actually takes two different positions R1 and R2, so that all calculations made
d = 4 / lf (c), where # is the difference of the coordinates d R1 and R2, and the trigonometric function f (c) is a function of theoretical angles only, which can therefore be calculated with any desired precision. more, Ro can be deliberately selected either to maximize the measurement accuracy (by making the modulus of f (c) as large as possible, that is to say 2), or to simplify the calculations. The calculations are simplified, in particular when all the angles ai and bi are chosen - to be multiples of 900, so that the value of f (c) is equal to t 1.

Various trogonometric functions f (c) can be used depending on whether one chooses d as representing the difference between R1 and R2 along the X axis, or along the Y axis, or the distance itself between R1 and R2. This choice is guided by the requirements of the measuring instruments used. In the case where # represents the distance between R1 and R2, and applying the law of cosines and trigonometric identities concerning the half angles, we obtain
sin (d / 2) = ## / 41sin C2 - C1 2 or, for an angle d / 2
d = # / # sin C2 -C1
d = + / 21sin ()
As the angles ci are purely theoretical, we can choose the point R0 such that:
C2 C1 \
2 # ---- # = 1, giving: d = ## / 21.

2
In a geometric configuration as shown in FIG. 2, the + or - sign of d is the same as the + or - sign of Y2 - Y1.

Once the zeroing error has been determined according to the method, various measures can be taken to improve the precision of the robot in practice.

For a given existing robot, one possibility is to include a parameter for the error term "d" in the software that is used to control the robot and then compile a
Robot specific instruction set.

An advantage of this solution is that it can be implemented without dismantling or adjusting the robot. A disadvantage, at least in a factory employing more than one robot, is that it is necessary to compile a different set of instructions to each robot.

Another possibility is to use the above technique during robot manufacturing to adjust the physical positions of the angle encoder components until the zero errors are too small to be detected due to the resolution. coders. However, this can be difficult given the small angles.

Between these two extreme solutions (external software or physical elimination), there are two other solutions which can be advantageous. One is to incorporate in the robot means that can be adjusted, either during manufacturing or during maintenance, to add or subtract a difference on the readings of the gross angle given by each of the coders. The second is to provide means to use the above process automatically, each time the robot is started. This technique is particularly suitable for robots in which each angle encoder generates a reading in the form of pulses applied to an up-down counter.

Robots always perform a fairly complicated initialization sequence, and the following variant may be advantageous. Zero determination means such as relatively imprecise limit switches can be used at the start of the initialization sequence giving the robot sufficient accuracy to be able to reach a fixed reference point in a permanent and known location in advance. Once the zero error has been determined by the above method, the joint in question can be returned to its actual zero position and, while the joint is in this position, the up-down counter responding to the pulses encoder can be reset.

The present invention does not relate directly to the particular techniques for forming fixed or mobile reference points, nor to the techniques available for bringing these points into coincidence nor, alternatively, for measuring the differences in position.

However, as a general rule, the errors in the repeatability of a robot do not exceed l / lOth of mm, while the absolute errors can reach 1 mm. The exact values depend on parameters such as the arm length (e.g. 300 to 500mm), the resolution of the angle encoder (e.g. I for 6000), and the configuration of the robot arm. The means used to establish coincidence or to measure the difference in positions must be at least of the same order of magnitude as repeatability. This implies the use of micrometric screw gauges or any other precision measurement equipment, for example based on light interference.

In its two aspects, the method can be implemented either manually or automatically, using servo systems.

The method can be modified in different ways and, for example, a plurality of reference points can be used, an average being taken on the different measurements. The error difference is then known in a fraction of the angle resolution. A fractional deviation can be represented by a configurable output from a CAD system, or by physically adjusting the positioning of the components of the angle encoder, even if it cannot be corrected by the other techniques mentioned above.

A fractional error deviation of zero can also be obtained directly (without average) by the second aspect, since the readings of the incrementation of the angle encoder are not taken directly into account in the calculation. However, the angle readings influence the position in which the movable reference point comes to rest, and it may be advantageous to use not only a plurality of fixed reference points, but also to compare the results obtained by moving 1 ' forearm in each of its positions clockwise and counterclockwise.

The present description disregards the other sources of errors (absolute and relative) present in a robot arm. If such errors have a greater effect than the zeroing errors, they must be taken into account before the implementation of the method.

Claims (9)

1. Method for setting a robot by determining errors  zero setting in at least one of the angle encoders  a robot arm, with at least two segments articulated in  series around joints having parallel axes and  each comprising an angle encoder, comprising in series  a shoulder joint, an arm segment (10, 20), a  elbow joint, a forearm segment (12, 22)  having a wrist end and a point of  reference close to the forearm wrist end,  said reference point being constituted by a fixed point  on the longitudinal axis of the forearm segment, while  being said "mobile" reference point since the arm is  mobile, the method comprising the following steps  - definition of a fixed reference point (R), this point is  located in a fixed position with respect to the axis (O) of  the shoulder joint  - displacement of the mobile reference point in coincidence  with the fixed reference point (R), the arm taking  successively each of the two configurations  symmetrical possible  - reading of the two angles (bl, b2) of the articulation of  elbow and  - determination of the zero error in the angle encoder  of the elbow joint as being equal to half  of the difference between the sum of the two angle readings  and the theoretical sum. 2. Method according to claim 1, characterized in that  said mobile reference point is materialized by a  ball fixed on the forearm, the reference point  fixed (R) being materialized by a block receiving a seat  for the ball. 3. Method of setting a robot by determining errors  zero setting in at least one of the angle encoders  a robot arm, with at least two segments articulated in  series around joints having parallel axes and  each comprising an angle encoder, comprising in series  a shoulder joint, an arm segment (10, 20), a  elbow joint, a forearm segment (12, 22), and  a reference point near the wrist end of  the forearm rod, said reference point being  consisting of a fixed point on the longitudinal axis of the  forearm segment (12, 22), while being said point of  "mobile" reference since the arm is mobile, the process  including the following steps  - definition of a fixed reference point, this point is  located in a fixed position with respect to the axis (O) of  the shoulder joint  - bringing the arm in the two theoretical configurations  which should bring the mobile reference point in  coincidence with the fixed reference point  - measurement of the difference between the two positions occupied  actually by the moving reference point; and  - deduction of the zero error of the angle encoder  the elbow joint, equal to a function  trigonometric of the measured difference of positions  divided by the length "1" of the forearm. 4. Method according to claim 3, characterized in that the  fixed reference point is chosen so that said  theoretical angles are multiples of Sao. 5. Method of manufacturing an arm robot consisting of  multiple segments, characterized in that the timing  zero, in at least one bend angle encoder of said  robot, is physically obtained by the process according to one  any of the preceding claims. 6. Multi-segment arm robot including means  reading correction at least the angle of the  encoder of the elbow joint, to take into account  the error of zero determined by the method according to one of the  claims I to 4. 7. Robot according to claim 6, characterized in that  said means include means for adding or  subtract an error correction term from said  readings. 8. Robot having a multi-segment arm and having  means for digital reading of at least one angle encoder,  said readings being obtained by an up-down counter  for measuring angle variations, said robot comprising  means for automatically implementing the method  according to one of claims 1 or 3 during the sequence  initialization, means to set the elbow angle  in its exact zero position, and means of reset  at zero of the up-down counter when said code angle  is in its zero position. 9. Method of implementing a robot using an arm  articulated comprising the following stages  - determination of the error of zero, at least in the  robot angle encoder by the method according to one of  claims 1 to 4  - zero error compensation by commanding the  movement of said robot arm by means of software  programmed to take into account said error.