SE458427B - Robot equipment with position indicator on robot hand - Google Patents

Abstract

The industrial robot (2) is fixed to the floor (1), and has a foot (21). A column (22) is rotatable around a vertical axis in relation to the foot.. A lower arm (23) is rotatably fitted to the column by means of a ontal shaft. An upper arm (24) is fitted to the lower arm, and also has a robot hand (25) attached to it. An intermediary part (27) is fitted to the robot hand, and has a tool fixture (28) carrying a gripper (29). On the intermediary part is an optical indicator (30) which is used to determine the robot hand position and orientation in space. The robot has a control system (5) permitting programming of a movement pattern, with subsequent operation of it in accordance with that pattern. The control system recovers a sequence of position theoretical values from the programme, compares them with known actual values for the robot hand position, and operates motors in the various axes of the robot so that the hand follows the programmed theroretical position values. The indicator determines directions along at least three or at most four lines of sight to three or four points with known positons. (Provisional basic advised week 8741)

Description

458 427 In order to obtain correct position information relative to a fixed coordinate system, it is further required that the robot's attachment (eg the mounting of the robot's foot in a floor) is done with high accuracy.

If manufacturing and assembly tolerances are not kept very small, problems with interchangeability also arise, ie it will not be possible to replace one robot copy with another without reprogramming.

It also becomes difficult to obtain, with maintained accuracy, a larger working area for the robot than that obtained when the robot is permanently mounted.

The invention intends to provide a robot with a position sensor device which makes it possible to obtain high accuracy of the position information independent of manufacturing and fastening tolerances and independent of deformations of the parts of the robot. The invention further intends to provide a robot, where robot copies can be exchanged for each other without reprogramming without any requirements being placed on small manufacturing and attachment tolerances. A further object of the invention is to provide a robot, whose position information is independent of the position of the robot's attachment point is known, whereby the robot can be allowed to move arbitrarily within a room and its working area can be made much larger than has hitherto been possible. A further object of the invention is to provide a robot whose position information is independent of active sources in the environment, whereby the robot can also be allowed to move arbitrarily in an environment not prepared in advance, for the sake of location information, prepared or possibly not even measured.

The invention further aims to provide a simple and robust sensor device for generating the desired position information.

What characterizes a robotic equipment according to the invention is stated in the appended claims.

The invention will be described in the following in connection with accompanying figures 1-7. Fig. 1 shows an industrial robot with an example of a position sensor equipment according to the invention. Fig. 2a shows the principle of the sensor mounted on the robot in Fig. 1. Fig. 2b shows the equipment connected to the detector surface for extracting the positions of specified phenomena and determining 3,458,427. the directions of the lines of sight (from the donor to these phenomena). Fig. 3 shows a block diagram of the position sensor device and Fig. Å shows how it is connected to the robot's control system. Fig. Sa shows a typical position of the different calculation areas (sub-pictures) that each participating vision system has been assigned to evaluate. Fig. Sb shows some of the coordinates used and Fig. 6 the vector designations used for the positions of the phenomena and the sensor. Fig. 7 shows a flow chart for the calculations that from the sensor signals give the position of the robot.

Pig 1 shows an industrial robot 2, which is fixedly mounted on a base 1, for example a floor. The robot is of a per se known type and has a foot 21. A pillar 22 is rotatable about a vertical axis relative to the foot 21.

A forearm 23 is rotatably mounted to the column 22 by means of a horizontal shaft. An upper arm 2H is mounted to the forearm 23 and a robot hand 25 to the upper arm ZH, both by means of horizontal axes of rotation in the figure.

Attached to the robot hand 25 is an intermediate part 27, which has a tool holder 28, which carries a gripper 29. On the intermediate part 27 an optical sensor 30 is arranged, which is used in the manner described below to determine the position and orientation of the robot hand in space. The robot has a control system 5, which enables programming of a movement scheme of the robot with accompanying operation of the robot in accordance with this movement scheme. The control system retrieves a successive sequence of position setpoints from the program, compares these with known actual values for the position of the robot hand and drives the motors in the different axes of the robot so that the robot hand follows the programmed position setpoints.

To determine the position of the robot hand, a sensor 30 is arranged on the robot hand. In the manner specified below, the sensor system determines the direction HOS at least three and preferably four aiming lines from the sensor to three (alternatively four) phenomena with known positions (in the coordinate system the position indication is desired). These phenomena are selected from a larger number of phenomena in such a way that they are partly within the sensor's field of view, and partly are recognizable by the sensor system without confusion.

The phenomena discussed in the description typically refer to corners, holes or boundaries between different surfaces on the object, but also light points, controlled or uncontrolled from the position sensor device, and light patterns projected on the object can be referred to. The combination effect of objects at different distances or in different positions relative to the sensor can also be referred to (eg moirê patterns or crossing bars at different distances from the sensor). Of course 458 427, any number of phenomena can consist of holographic images or patterns.

The positions of the phenomena in a certain coordinate system are assumed to be known and can in a number of different ways be fed into the sensor system, or alternatively the control system from which the latter can be retrieved, during programming or installation of the robot. From these pre-known positions and the directions determined by the sensor system, the sensor system calculates the sensor's position in the current coordinate system in the manner described below. The phenomena can, for example, be fixed and the robot's position is then calculated in a fixed coordinate system. Alternatively, the phenomena may be phenomena on the holder of the workpiece (or the workpiece itself) which the robot is to machine or handle and the position of the robot is then obtained in relation to the holder or the workpiece.

In a manner known per se, the robot is provided with angle sensors in its axes, which provide the control system with information about the current angles of rotation in the axes. This information is used by the control system to distribute a behavioral error on the various robot axes. These sensors can be made with significantly lower accuracy than was necessary with previously known robots, where the sensors are used to determine the position of the robot hand.

For natural reasons, the sensor 30 can usually not be placed at the point whose position one really wants to determine. This point of interest consists, for example, of the gripping claws of a gripping device, of the tip of a welding electrode or of the edge of a machining tool. With knowledge of the position of the sensor in space and of the orientation of the robot hand, however, the position of the mentioned point can easily be calculated. Information on the orientation of the robot hand can, as will be described in more detail below, be obtained with the aid of the position sensor system according to the invention. Since the sensor 30 is preferably placed as close to the actual point as possible, sometimes no greater accuracy is required in determining the orientation of the hand, and therefore the angle sensors arranged in the robot shafts can alternatively be used to provide information about the orientation of the hand.

In order to facilitate a position and orientation determination of the robot hand_abouting its orientation relative to the phenomena, additional sensors can be arranged on the hand in such a way that during the operation of the robot at least one of these sensors (or all sensors together) always has a sufficient number of phenomena within its field of view. . Such an additional sensor is shown in broken lines in the figure and is denoted by 31. 5, 458 427 For determining the position and / or orientation of the other robot parts, these parts can also be provided with position sensor systems according to the invention, if desired.

As an example, the figure shows how a sensor 32 is arranged on the forearm 23.

A set of phenomena is then specified so that at least three and preferably four during the robot's operation are within the sensor's 32 field of view (here too, of course, additional sensors can be arranged to increase the chances that a sufficient number of phenomena are within the sensor's field of view).

The position sensor equipment thus arranged for the forearm 23 can be used for determining the position and orientation of the forearm. In the case of a sliding robot, for determining the current position of the robot within its sliding area, such a sensor can be placed on, for example, the Foot 21.

As an alternative to mounting several sensors on the robot hand, the sensor equipment can be provided with more specified phenomena than those shown in Fig. 1.

These phenomena are then selected to be those which are placed in such positions and in sufficient numbers, eg six to eight, that, regardless of the position of the robot hand on orientation, always at least three and preferably four of the phenomena are within any donor's. field of view.

As will be described below, the sensor system for determining position and orientation uses three of the phenomena at a time. The robot always knows its position and the sensor's orientation and control system can therefore be programmed to select and use three phenomena suitably placed within the sensor's field of view in each position for each position determination.

For easily understood reasons, the accuracy of the position determination becomes greater the closer the robot hand is to the phenomena. It may therefore be appropriate to have a set of selected phenomena at such a great distance from the robot that they can be used within the robot's entire working area and in addition an additional set that is placed as close as possible to the work point or points where high accuracy is required. Fig. 2a shows the principle of the sensor 30 in Fig. 1. The sensor has a wide: angle lens 300, for example Canon's fisheye optics 7.5 mm f / 5.6, which is symmetrical about the axis 301 of the sensor. semi-spherical and is so designed that it has as large a 458 427 6, at an opening angle as possible. Preferably, the opening angle should be of the order of 2! steradians (the larger the part of the space angle than is covered, the better). perpendicular to the axis 301, an image-receiving surface 302 is arranged (eg a CCD surface, an example is Tektronix TK 2008). The lens images the outside world on the detector (= the image-receiving surface). The focal length of the lens should be made short as this partly leads to the desired large aperture angle, and partly a sufficiently good focus for all current distances to the phenomena to which directions are to be determined. Since the purpose of the sensor is to obtain as much accuracy as possible in determining the directions from the sensor to the phenomena, the lens is preferably designed so that incoming light beams with the directions 0 (in rig za) (between the line of sight of the phenomenon and the optical axis) are imaged. so that the angle står is in an approximately linear relationship to the distance (on the detector surface) from the optical axis to the point 31k on the surface at which the light rays from the phenomenon are imaged.

Fig. 2b shows in block diagram form the sensor system 704, which has an arrangement of vision systems (image processing systems) 706-708, of a type known per se, their superior control system 703 and calculation unit 702, which arrangement is used to evaluate incoming images and generate the direction information which is then the starting point for the calculations of the sensor's position,.

The sensor 30, consisting of, among other things, a wide-angle optics and a two-dimensional CCD surface (for example TK 2038 from Tektroníx), emits a signal in the form of a video signal (with, for example, the number of lines in 2008 and ZOH8 points per line).

This signal is routed to a unit 701, whose function is to extract frames from the large image. This is done by control unit 703 (based on previous calculations) ordering unit 701 to take out the sub-pictures which have the lower left corner of the picture element with the coordinate.xi ', yi', for 1 = 1 t111 u.

(The number of phenomena and vision systems can vary according to the invention, but for the sake of clarity it is described here with U pieces of vision systems.) Equipment of type 701 is marketed by several companies, including Colorado Video Incorporated (CVI).

A figure of how these sub-images lie in the big picture, on a special occasion, is seen in Fig. Sa. The sub-pictures are, for example, 256 x 256 pixels large. These sub-pictures are led to the four vision systems 705-708 arranged for the purpose, which, when loaded with their respective sub-picture, are started by control unit 703 to process the current sub-picture with respect to the object (fí; 506-509) the control means. ordered. At a specific time,. 1 458 427 that in Fig. 5a, the phenomena are 506-509. These phenomena have been chosen as they are programmed to be selected in this particular time section. The positions that control unit 703 commands unit 701 to locate the lower left corner of the sub-pictures depend partly on the previous position of the phenomenon (fi) (1 the whole picture), which information is obtained from calculation unit 702, and partly on the positions_§, Ä and speeds X, Å the calculating means 52 generates. X, Z, Å, Å and x and yi (1 = 1 to M) are weighed together in control unit 703 so that the objects 506-509 in the next calculation are predicted to come in the middle of each sub-picture (sub-picture S01 to 50U in Fig. Sa, corresponding to i = 1 to U).

Calculation unit 702 calculates the position of the phenomena in the total picture xí = xi '+ xí ", yi = yi' + yi", i = 1 to H. The positions of the sub-pictures xi ', yi' are obtained by the unit 702 from control unit 703 and the positions of the phenomena in the respective sub-image Xi ", yi" receives it from the respective vision system (705-708).

All units (30, 701-708) are clocked by unit 703 as well. Unit 703 does this at a rate coordinated with unit 50 (see Fig. 3), signal c1 ensures that a common rate is maintained.

Arranges the outputs from unit 701 to the vision systems on standard video footage for 256 x 256 pixel images and the control signals (with information about the current phenomenon to evaluate the image with respect to, as well as start / / stop and other coordination information) from control unit 703, and the information paths from or vision system for calculation unit 702, gem RS 232 communication channels (9600 baud, ADLP10 protocol) an ASEA's vision system IVR 2600 EIII can be used, as each unit 705-708, without significant changes. The function described above is hereinafter referred to as the sensor system 703.

Control unit 703 and calculation unit 702 can be designed with specially built circuits according to known principles. Alternatively, they may be a suitably programmed (micro) computer.

When changing phenomenon fi to another, this is controlled by a signal (F) coming from unit 51, this signal also contains information about where in the big picture this phenomenon can be expected to be. Unit 703 ensures that this change of phenomenon takes place.

Figure 3 shows a block diagram of the position sensor equipment, where the sensor system 70ü is described as a block. A control unit 50 emits clock pulses and 458,427, possibly for coordination and synchronization, additional information necessary for the other units of the equipment. This information is in the figure denoted by cl. The sensor system TOM is designed according to figure 2b and emits signals xí and yi to the units 51 and 52. The unit 52 is a calculation unit which, with the aid of the signals xi and yi and stored data on, among other things, the positions of the objects, calculates information defining the robot hand position Ä and orientation Å. The calculation unit 52 also determines by deriving the vectors Ä and §_the velocity vectors Å and Å. These four vectors or signals are supplied to the robot's control system for controlling the robot's movement. Control unit 51 receives information Ä and šß about the position and orientation of the robot hand as well as information xi and yi about the position of the images of the phenomena. on the detector. Depending on this information, control unit 51 selects the appropriate set of phenomena (from the number of phenomena specified in the programming) for the further calculations and sends this information partly to unit 52, signal i, (so that it knows which phenomena have been used), partly to 7GB device for enforcement (signal F). The signal to unit YOU includes the probable location of the phenomenon in the big picture.

Unit 70U can be made in the manner shown in Fig. 2b. Unit SD can consist of a simple clock pulse oscillator or of a suitably programmed microcomputer. The unit 51 can consist of a specially built logic connection or a microcomputer. Calculation unit 52, which is to perform the calculation operations described below in connection with Fig. 7, may in a manner known per se consist of digital or analog circuits or of a suitably programmed microcomputer.

When programming (learning) the system, the vision systems must be taught in a known way (described in the manual for the vision systems) to recognize existing phenomena. This is suitably done with one of the vision systems, after which the other vision systems receive the information via the mass memories built into the vision systems or via, for example, an RS 232 link. The coordinates corresponding to each phenomenon in surrounding coordinate systems are stored in units 52 and 51 in an order corresponding to the order of the phenomena.

When starting up the system, no unit in the system knows which phenomena are in the field of view or where they may be. At start-up, unit 703 ensures, before giving a clear signal on signal line cl, that the entire large image is scanned and evaluated for all probable phenomena. This sequence is pre-programmed either with logic circuits 9 458 427 or as a suitably executed program in a microcomputer. The large image is scanned in rows with steps of, for example, half the size of the image field (eg 128) and also in columns, in order to have the maximum opportunity to find all phenomena. When the phenomena are found, information is identified which identifies the phenomena found and their locations via signal channel F up to unit 51, which when all phenomena present in the image are found determines which phenomena are initially to be used and sends this to unit 52 and unit 7OU.

In Fig. 2a, the coordinate systems used in the calculations are exposed. Using information from the vision systems described above and from the units 701 and 703, a pair of signals x1, yi (preferably digital) are obtained from the calculation unit 702 for each phenomenon, which is a measure of the position of the phenomenon's image in the overall picture. These signals are typically spaced on the detector surface partly in the x-direction from the axis 301 and partly in the y-direction from the axis 301.

With the aid of the signals x and y, as will be described in more detail below, the image of the phenomenon on the detector surface in polar coordinates a, p can be determined. From these coordinates are then obtained the spherical coordinates q, 6 (with respect to an origin which lies the distance R above the detector) which defines the direction of the aiming line 313 to the current phenomenon.

Nonlinearity 1 of the sensor, etc. can give rise to incorrect directional information, so the two-dimensional coordinate x, y from the evaluation system may have to undergo a correction process. This can consist of storing in advance in another, and more linear, manner varying and measured angles and at the same time with the sensor measured coordinates in such a way that correction of the values given by the sensor (continuously) can take place.

Optionally, unit 703 and unit 50 may be embodied as one unit. Optionally, unit 702 and unit 52 can also be designed as one unit. Units 51 and 703 can also be designed as one unit.

Fig. H shows how the position sensor equipment according to the invention is connected to the robot's control system. The sensor system YOU outputs the signals x, y to the calculation unit 52, which in turn calculates a vector Ä which defines the position of the robot hand. This signal is applied to a comparator 53. To a second input on the comparator, position setpoints X0 are fed from the robot's software 58 at a suitable rate. From the comparator, a signal is obtained at each moment which defines the position deviation of the robot hand. This position deviation is applied to a so-called shaft divider 54, in which the deviation is transformed into the coordinate system of the robot, ie divided on a fault in each of the axes of the robot. The shaft divider 54 thus has an output for each of the robot's shafts. In the figure, only the output signal - Ar1 - for axis 1 is shown.

This signal is applied to the regulator 56 of the shaft 1. From the unit 52 a vector Å is also obtained which defines the instantaneous speed of the robot hand. In unit 55 this speed is compared with a setpoint io for the speed.

The difference between the two values supplied to the unit 55 is then divided on the different axes of the robot, ie transformed over to the coordinate system of the robot. The unit 55 has an output for each of the robot shafts, but only the output signal Ar1 for the shaft 1 is shown in the figure. This signal is applied to the regulator 56 of the shaft 1. The regulator 56 controls the unit 57, which contains a power amplifier and drive motor for the shaft 1, in a manner known per se.

The output signal from the drive unit 57 thus consists of a rotation of the robot's axis 1, and this rotation - in the figure denoted by 71 - in turn affects the position and orientation of the robot hand and the sensor 30. With the aid of the system shown in Fig. H, the robot will controlled so that the sensor 30 follows the path determined by the successively fed position setpoints from the program unit 58.

In Fig. U, for the sake of simplicity, only the position determination and control have been shown. Adjustment of the orientation of the robot hand can be made in a corresponding manner 1 depending on the information about the actual orientation §_ obtained from the calculation unit 52 and on setpoints for the orientation, which are fed by the software 58.

Pig Sa shows the full range of images and the position of the sub-pictures at a given time. The number of pixels m-n can be, for example, 2048x20Ä8 and is determined by the design of the detector 302. The frames 501-504 can have the same size or be different sizes. A sub-picture can, for example, consist of 256x256 picture elements. The number of fields can be both smaller than or larger than U. The contour line 505 of a selected object has four characteristic points, which constitute the above-mentioned phenomena used in the position determination. The corner 509, the tip of the nose notch 508, the center of the hole 507 and the center of the rounded tongue 506 constitute these phenomena. Their positions in the sub-pictures are determined by the vision systems in a known manner, and since the position of each sub-picture in the overall picture is known, the positions of the phenomena in the picture are thus known. Fig. 45b shows some of the designations used in the calculations performed by the calculation unit 52. The figure schematically shows the surface of the detector 302 in Fig. 2, seen from above. The detector's right-angled coordinate system x, y is marked, as are the polar coordinates u, p for each of the three phenomena a, b, c.

Fig. 6 shows the three phenomena a, b and c, which in a certain time section are used for the position / orientation determination. The vectors Ä, É1 and §2 define the position of the sensor 30, the phenomenon b and the phenomenon c, respectively, in relation to phenomenon a. The vector_§3 is the vector product of the vectors B and B. The vectors P, P and P define the positions of the phenomena a, b -4 -2 -a -B -c and e in the coordinate system where the positions of the phenomena are defined.

The vector Ä denotes the position of the sensor 30 in the same coordinate system.

The function of the calculation unit S2 will now be described in more detail in connection with Fig. 7. This unit can be designed with specially built circuits for solution according to known principles of the relationships given below. Alternatively, it may be a suitably programmable (micro) computer. For the sake of clarity, the latter is assumed to be the case, and Fig. 7 shows a flow chart for the calculation process of such a (micro) computer.

The three phenomena used in each calculation give rise to three pairs of related x-y values on the surface of the detector 302. Which directions in space correspond to these x-y values depends in part on the design of the sensor optics. In the following, for simplicity, it is assumed that R in Fig. 2 is constant (independent of 6). However, this need not be the case, but the calculations will be somewhat more complex if this were not the case.

Namely, R (9) is not an unknown quantity but only an additional parameter to take into account in the calculations.

At the beginning of the calculation, the calculation unit receives information from the control unit 51 about which three phenomena are used in the calculation. These phenomena are denoted by a, b, c. An "i" image of a phenomenon on the detector surface has in its coordinate system the position xi, yi. The position of the image in polar coordinates is given by the following equations: d) | l 458 427 12 Q u V1 arctan (T) + (1 - P1 = “X1: * yiz From these polar coordinates are then obtained the spherical coordinates of the line of sight to the phenomenon from the equations : Qi: Gi N01) The function H01) depends on the optics and can be measured and stored in the calculation unit once and for all. However, the optics should be designed so that the function is monotonous in the current range. Hereinafter, the angles T1, '22 and T3 between the three aim lines from the sensor to the phenomena are calculated by means of the following equations: c1gwa-wb Czzq) -qaa-o-Z! C3zcpb-qzc sin 9 a T1 _ arcsin (sinC1) C 6 _9 c e _e (_1). ab _1_ a cot sin (cot 2) cos 2 arctan eggb + arctan ea + 9b sin 2 cos 2 sin 6 "- c TE _ arcsín (sinCz) C e __e C a _e 2. ca 2 c (cot -) sin ( cot -2-) cos »2 arctan ec + ea + arctan ec + ea sin 2 cos 2 458 427 13 sin Bb -I T3 = arcsin (sinC3) G3 e _96 63 e _60 (cot -ZA) sin (cot -2 -) cos 2 arctan + arctan _ bcbc sin 2 cos 2 The first four squares of the flow chart have now been processed, in the fifth step the functions Gi (A) are formed according to: zone y + 2 2. 2 + J] G (A) _ (cosT ) (AcosT2 - bz-A sin T2) - b3 3 f - [AcosT t Jgz-A2sin2T J 1 1 1 (Acosrzï bâ-A2s1n2T2) 2 (s1n2tr3) - where bl is the distance between phenomena a and b be is the distance between phenomena a and c b3 is the distance between the phenomena c and b This function is actually eight different functions, one for each of the eight combinations of characters at the three places in the expression where either a plus or minus sign can occur (the character in front of the expression I bg - A2sin2T2 always has the same value in the two places where the expression fö re- comes). These functions are denoted in the fifth box of the flow chart by Gi (A), where i = 1 ... 8. For each of these functions, zeros 1 are determined in those parts of the function, where each of the two terms of the function as well as AcosT2 É Jb: - A2sin2T2 is both real (denoted by J “2 0 in the sixth box of the flow chart) and positive (denoted by [1 Z 0 in the sixth box of the flow chart). The next box examines how many zeros have been obtained in total. If only one zero has been obtained, the calculation program proceeds to the box marked B, C in the flow chart. If more than one zero point has been obtained, in the next point of the calculation scheme the zero points are selected where A compared with previous calculations would result in an unrealistically high speed of the robot. If after this check only one zero point remains, the calculation program proceeds to box B, C. If there are still several zeros left and if the current passage through the flow chart is the first passage, a new group of three phenomena is selected, in the flow chart denoted by a, b, d If so, the above-described calculation is performed from the beginning for the new group of phenomena.

Thereafter, the position determination is described below for both sets of phenomena, after which the solution which gives a common position is selected. If the current passage through the calculation scheme is not the first passage, according to the above description, calculation of zeros has been made for two sets of phenomena, the calculation proceeds according to the description below and finally the solution is selected that gives a common position for the two sets of phenomena. .

According to the next step in the calculation, denoted by B, the expressions are first: C in the flow chart, sign- A + bï - 52 n I BPCCOS znobï A + bg - ca Y = aPCCÖS znobz bf + so _ on E: QFCCOS 2b1u2 where Ao is the value of A where the function G (A) = 0 and B and C respectively are given by the expressions: + 12 2. 2 B = A0cosT1 - b1 - Aosin T1 -r / Z .2 C = A0cosT2 - bz - Aosin was where the ambiguous characters in the expressions for B and C are selected as the characters in the corresponding expression in the equation Gi (A) which gave the current zero. After this, the quantities s and k are formed according to: _n + '1 -. + = S -' ~. 2 k _ / sin (sY) sin (s-nlsin (SE) 'sin s and the quantities u and 6 according to the_expressions: 458 427 15 u = 2 arctan šïaïšïçï 6 = arcsin (his senses) whereby if x> than the å solution 6 > å is selected and if x 5 å the solution 6 i š is selected.

As can be seen from Fig. 6, the direction from the phenomenon a to b is given by the vector É4 and from the phenomenon a to c by the vector ga. The direction from the phenomenon a to the sensor 30 is denoted by the unit vector A. The vector B3 constitutes the vector product of the vectors Éh and §2. The vector §_can now be solved from the following equation system: B1x B1y B12 AX cosn B B A co Y 32x 2y Zz y = S B -ö 3x B3y B32 AZ cos (l) The position of the sensor 30, which is denoted by the vector Ä is obtained here; ur: æ = ß -w o få The sensor's 30 position has now been determined.

Then the orientation of the sensor is determined by forming the following vectors: Éa'š Z = "° lïa-šl Z = gb-š '° lïršl z ---_- B ° -' fire-Ål Hereinafter, the components na, nb and nc of the normal vector to the sensor detector surface 302 from the system of equations: / ZZZ n cos9 ax ay az aa Zbx Zby Zbz nb = cosßb Zcx Zcy Zcz nc cosßc 458 427 .

If necessary, the translation and rotation speed of the robot hand is then obtained by means of differentiation of the signals Ä and Zr. It has been described above how the sensor device according to the invention is used for both position and orientation determination, but of course the equipment can only be used for position determination or orientation determination.

In Fig. 1, for the sake of simplicity, an industrial robot with only four axes has been shown, but of course the invention is equally applicable to robots with a different number of degrees of freedom. The invention furthermore provides the same advantages also with other working means than the gripper shown in Fig. 1, for example welding equipment, machining tools, assembly tools, etc.

Of course, the positions of the phenomena only need to be known relative to the working point where one wishes the robot to move. If the operating point of the robot is on a moving object, phenomena fixed on this are advantageously specified, whereby automatic tracking of the object is obtained. In this case, however, it should often be appropriate to also have a number of phenomena specified in a fixed coordinate system so that control is always obtained at the position of the robot relative to a fixed coordinate system (so that the system does not drift away). The switching between these two systems can take place continuously, and the control system can, for example, simultaneously keep track of the robot's position in both systems but only control the robot's movement by means of one system, preferably the one that provides the greatest accuracy (if possible).

In order to evaluate a sufficient number of sub-pictures, it is of course possible to equip the sensor system 704 with a sufficient number of additional vision systems to cope with the evaluation of more than one set of phenomena.

Another possibility is to appropriately, and controlled from the control means 703 and / or = 51, coordinate the evaluation to a smaller number of vision systems .; Calculation and / or control means 703, 51 can of course either be divided between the different sets of vision and calculation systems or be more or less separated. Unit 701 can of course also be either divided or multiplied. 11 458 .427 When selecting phenomena, according to known geometric principles, in order to achieve the highest possible accuracy, these should be chosen so that, when the robot is in its desired working range, the aiming lines between the sensor and the phenomena used form angles with each other that are not too great. small. Optimal from an accuracy point of view is when they form angles between each other that are about 900.

The invention has been described above in connection with an equipment which works with optical radiation. It is then obvious that the radiation used can be both inside and outside the visible wavelength band. However, equipment according to the invention can be arranged to work with signals other than light. Ultrasounds and microwaves are examples of such signals.

As can be seen from the above description, a robot according to the invention offers great advantages compared to hitherto known and used robots.

Because the position of the robot hand is directly determined by the sensor, which is arranged at or on the hand, a high accuracy of the position determination is obtained regardless of manufacturing and mounting tolerances of the robot parts and independent of errors in the signals from the angle sensors arranged in the robot's different axes. Likewise, the high accuracy of the position determination is maintained regardless of the deformations that can occur in the robot at high load or high accelerations. These facts make it possible to achieve a robot which, despite significantly lower tolerance requirements and stiffness requirements, exhibits a very high degree of accuracy in the position determination. In addition, a robot equipped with a sensor system described in the invention can find the correct position independently of the temperature of the material (in the robot).

A further essential advantage of the invention is that one robot can be directly exchanged for another, for example in the event of an interruption of operation, without any reprogramming being required.

A further advantage of the invention is that no preparation, special for the locator system, needs to be made by the work object or the environment and a sufficient number of easily identifiable phenomena are naturally present there.

Because the phenomena used in the position determination can suitably be selected fixedly connected to a workpiece and close to the actual working point A 458 427 W point for the robot, the important advantage is obtained that the robot will automatically follow movements or position deviations of the workpiece. Furthermore, by selecting the phenomena near the working point, a very high is obtained. accuracy in the vicinity of the working point, where this is highly desirable, while further from the working point a lower accuracy is obtained, which can be accepted without inconvenience. By obtaining the position determination according to the invention with the desired high accuracy without the attachment point of the robot having to be known, the robot can in principle be allowed to move freely in the workroom, whereby its working range can be extended significantly compared to previously known fixed or limited movable robots.

According to a further variant of the invention, the phenomena described above do not have to be naturally occurring phenomena (on the workpiece e d) but can of course be applied there in order to be easy to find and follow with the vision systems.

Another monitoring function present in the workshop can be solved with a device according to the invention, namely the monitoring that no person enters the work area. This is done so that an arbitrarily selected, otherwise invisible symbol in the workshop from the donor is applied in sufficient numbers to staff present in the workshop, in such a way that the donor, when a person is within the donor's field of view, always independent of the person's orientation relative to the donor sees at least one of these symbols. Either one or more of the existing vision systems connected to the sensor 30 can monitor that this symbol does not appear in the sensor's field of view, or another one or more vision systems can be installed to handle this monitoring. As people prefer to approach the system slowly, these systems can scan the large image at the appropriate speed (in the search for the symbol). For example, the symbol can be a pattern in the fabric of protective clothing used in the workshop.

Of course, a separate sensor can also be mounted on or adjacent to the robot. Evaluate the image information from this sensor using other vision systems or vision systems already installed for the sensor 30.

- The person recognition symbol described above can of course be part of or the whole person himself, eg the hand or the face. 458 427 19 There are several other ways of using a robot according to the invention.

One such way is to use the sensor according to the invention only as a position sensor in the control loop and to use conventional speed sensors, eg tachometers in each robot shaft, as speed feedback for the control system. Another alternative is to use the sensor as a complement to conventional position sensors for calibration or control, to increase the accuracy of the position determination, or for measurement to enable the replacement of robots. An additional alternative is to only use the sensor as a speed sensor.

When exchanging a robot for another specimen, which is to perform the same function in the same place, it is not necessary that the positions of the phenomena are known. In this case, it is only necessary for the new robot to be able to retrain the positions at which the positions of the phenomena's images on the detector correspond to those obtained with the original robot in corresponding positions.

The appearance and positions of the phenomena can be entered in a number of ways, eg it can be done in one of the following ways: 1) Manual entry of the appearance via a learning procedure according to the current ASEA vision system (see the manual). The location of the phenomenon can be entered, for example, by manual measurement or with some type of aid, for example in the form of a theodolite. 2) In case the design of the workpiece is entered on a CAD / CAM system (Computer Aided Design / Computer Aided Manufacturing), the appearance of the phenomena selected on the workpiece from different angles can be fed directly from the CAD / CAM system as well as their positions relative to each other and / or the working point of the robot. 3) Another way of entering the phenomena of phenomena is to measure the direction of the phenomenon from two known (relatively different sets of phenomena) and to calculate from it the new phenomenon; its location.

The appearance of the phenomenon is stored in the case described above with ASEA's vision system in the respective vision system, while the position is stored in the calculation unit (52). 458 427 20 Another possibility when using a CAD / CAM system is that the robot's work points and / or entire work program can also be manufactured in the CAD / CAN system, including the parent computer, and fed down to the robot.

Since the positions of the phenomena without intermediaries directly control the robot's actual value and setpoint, the robot can in this way be positioned with a precision approaching the sensor's inaccuracy.

A problem that may arise in certain work situations is that during work operations on the workpiece it is not possible to maintain a satisfactory viewing angle to a sufficient number of phenomena. This can be solved in a number of different ways. Two of them are described below: 1) A sensor of the same type as the sensor 30 attached to another part movable relative to the sensor 30, in a position measurable with the appropriate type of sensor (x) relative to the tool 29 ooh / or the original sensor 30, can temporarily take over or duplicate the calculation of the position of the tool, whereby in the calculation of the position of the working point the relationship between this additional sensor and sensor 30 measured with the sensor described above is used.

A special case of this method is that the original sensor 30 and the duplicating sensor which at this stage have the possibility to more accurately determine their position relative to the phenomena on the workpiece, both use other for both visible phenomena to calculate the position difference between the two sensors and thus between the temporarily use the sensor and tool point. In the case of certain complicated objects and / or operational cases, this procedure may need to be used more than temporarily (then becomes a normal operational case). 2) In applications where the position relation between the tool and a sensor retracted in accordance with 1) is not used in any other way, the position of the tool can also be measured from this sensor in a manner similar to the method described in Patent No. EP 0 ïïü 505. using and relatively the retracted sensor.

The image-receiving surface (detector) can of course be of different types. 'r ex' a con-surface (eg Tekn-enn TK 20148 with zousz biiaeiement), ew- Vidicon or Plumbicon tubes (preferably high-resolution ones) or a photodiode array can be used.

To reduce the number of vision systems needed, of course, a smaller number of vision systems can be multiplexed into the various tasks / sub-pictures, 21 458 427 extremely enough, a single vision system is enough, which in turn can count on the different sub-pictures.

When starting up the system, it is not always known which phenomena are visible and / or where they are in the big picture. There are a number of ways to solve this, some of which are described below: 1) The vision systems available in the system may analyze the entire large image including all (or a subset of) specified phenomena by taking sub-images, which may overlap, in any order. and analyze them. 2) Another procedure, which may be appropriate in certain situations, is not to take each pixel (= pixel) but, for example, to select a selection of large image points, such as every eighth point on every eighth line. Then the image thus obtained is roughly analyzed for all (or a subset of) specified specified phenomena. Later, the parts that are found to contain the correct phenomena are finely analyzed.

A number of possibilities for locating objects and for positioning oneself relatively relative to these are opened if a number (typically two or three) of the sensors described above are combined. If two of the sensor systems described above are combined, in addition to positioning oneself relative to a known coordinate system determined by a number (greater than or equal to three) phenomena within the two sensors' combined field of view (greater than or equal to one sensor's field of view), measure distance, position and orientation relative to phenomena, the positions of which are not known relative to any sensor and / or any coordinate system, but which are within the field of view of both sensors. (This is done by knowing the relative position of the sensors and thus two angles and one side 1 a triangle are known. The additional angle needed is obtained from either sensor system.) In particular, three (or more) phenomena on an object can be determined and these points thereafter be used as the phenomena to which the donor's position is relatively related. The coordination of used phenomena (in the different sensors) so that the measurement of the directions of the aiming lines (in the different sensors) refers to the directions of the same phenomena can take place, for example, by so-called image matching. 458 427 22 If three (or more) sensor systems are used, the possibility is obtained to solve a phenomenon directly by solving two triangles in the equation systems instead of going over the distance and the angle of inclination in the other direction as with two sensor systems.

Unlike conventional previously described vision systems, which by means of two cameras equipped with linearly imaging optics (straight line is imaged as a straight line on the camera's image plane, unlike the embodiment of the invention where a change in the angle of view is imaged linearly regardless of where in the field of view it is ) obtains the possibility to determine by means of stereometry the distance and position relative to the cameras, the systems described above can, through their wide-angle optics and imaging of angle approximately linearly to radius on the detector surface, obtain their position and attitude with great accuracy.

If more than one detector surface is used, the vision system, control and calculation units used for one surface can of course handle parts of these tasks for the other detectors as well.

Claims (9)

B 458 427 PATENT REQUIREMENTS Robotic equipment with a plurality of relatively movable parts (22-25) and comprising a sensor device (704, 50, 51, 52) for determining the position of one of said parts. which sensor device comprises a) an optical sensor (30) mechanically connected to said part arranged to image the surroundings of the robot, the sensor device being arranged to provide information (x, yi) defining the direction relative to the sensor of the aiming lines from the sensor to objects in the environment, b) calculation means (52) arranged to be supplied with said direction-defining information. to determine the angles (T1, T2, T3) between the aiming lines and to determine the position of the sensor relative to the objects on the basis of said angles, characterized in that the sensor (30) is arranged to produce a two-dimensional image of the environment and to a the image processing system (705-708) emits an output signal corresponding to the information content of the image that the image processing system is a) arranged for storing information defining the image of each of a plurality of predetermined details (506-509) in the environment, which details constitute said objects, b) arranged to scan the image of the environment generated by the sensor with respect to said predetermined details using the stored information for determining the positions of the details in the image, and to form said direction-defining information on the basis of the positions thus determined. K 2. Bobot equipment according to claim 1, characterized in that the calculating means (52) are arranged to generate information (na, nb, ne, u) from said angles, which defines the orientation of the sensor relative to said details. 458 427 Å Robotic equipment according to one of the preceding claims. t e c k n a d thereof, wide-angle type. k n n e - that the sensor (30) comprises a lens system (300) of Ä. Robotic equipment according to any one of the preceding claims, known therefrom. that it comprises means (101. 70 3) arranged that from the image a; u: en; plural aenumer (501, 502, 5o3, 504) ' Robotic equipment according to one of the preceding claims. is characterized in that it comprises a plurality of vision systems (705-708) arranged to evaluate the image with respect to each set of said details. Robotic equipment according to claim 4, characterized in that it comprises a plurality of vision systems (705-708) arranged to be supplied with a separate image and to evaluate the images with respect to each set of details. Robotic equipment according to any one of the preceding claims, characterized in that it is arranged to alternately evaluate the image with respect to at least two different sets of details and to generate information defining the position of the sensor device relative to each of the sets. Robotic equipment according to any one of the preceding claims, characterized in that it comprises at least two separate optical means (30) arranged to each produce an image of one and the same detail. that the vision systems are arranged to emit signals on the basis of the images which for each of the optical means define the direction from the means to the part, and means arranged to determine the position of the detail relative to the sensor device on the basis of said signals. Robot equipment according to any one of the preceding claims, characterized in that the vision systems (705-708) are arranged to evaluate the images with respect to a predetermined detail and to emit an indication signal in the presence of the image of this detail. .