FR2737024A1 - Process for training industrial robots and object digitisation by following geometric form and contours - involves utilising sensor sending position and orientation data as sensor unit is guided over chosen parts of object, and using computer to translate data to three-dimensional data - Google Patents

Abstract

The learning process has an operator (12) who displaces a sensor (10) over the geometric form itself. The sensor has moving internal components to measure position and orientation. The sensor cooperates with fixed detectors (4,5) connected to the work space where reference axes are formed. The sensor is calibrated prior to use then used to collect position and orientation data relative to the reference axes. This data is processed by a computer (24) to generate three-dimensional representation of the object. The operator chooses particular paths (S1,S2,...,SN) that define aspects of the form that will be important in defining the form and important in processing the object. The data is recorded at predetermined instances.

Description

"Geometric shape learning process,
including trajectory or contour, and systems
implementing this process "
DESCRIPTION
The present invention relates to a learning method of geometric shape, in particular of trajectory or contour. It also relates to systems implementing this method.

The term machine is understood to mean a mechanics composed of controlled digital axes. They may in particular be robotic gantries or industrial robots identified in the following under the term of robot. A robot must be programmed to perform complex movements. A programming phase requires either immobilizing the robot, or very expensive workstation development means. The movements of a robot can be programmed in several ways
- by successive manual movements of the robot
(point by point learning);
- through a "dummy" reproducing
the physical characteristics of the robot;
- by description of the movements by means of a
appropriate language;
- by simulation on a computer system.

 Point-by-point learning requires the immobilization of the production chain, or the use of a test cell with characteristics identical to those of the production chain. The use of a mannequin for learning trajectories has many disadvantages. In particular, the mechanical structure of a mannequin often makes it difficult to handle and when handling it, the operator often has difficulty compensating for its inertia. The description of the movements is only possible for simple trajectories or decomposable into a main movement and controlled movements. Simulation on a computer system involves perfectly representing the part concerned and its environment in three dimensions on a workstation.

 Current methods for learning the trajectory and more generally the geometric contour have the disadvantage of requiring heavy or expensive equipment, qualified personnel and sometimes immobilization of a production chain.

 The object of the invention is to remedy the aforementioned drawbacks by proposing a method of learning a geometric shape, which allows recording of a trajectory or of a contour on a real part, without using the machine itself. or any other mechanical device.

 According to the invention, the method for learning a geometric shape, in particular a trajectory of a tool or an outline of a part, within a workspace, is characterized in that an operator moves over said geometric shape an object provided with mobile position and orientation sensing means cooperating with fixed position and orientation sensing means linked to the workspace with which a work mark is associated, this object being previously calibrated , in that a recording of capture data relating to the position and orientation of said object is carried out and in that this recorded data is processed to provide three-dimensional data relating to said geometric shape.

 Thus, with the method according to the invention, it becomes possible to carry out learning of a geometric shape, in particular a tool trajectory or a workpiece contour, with equipment of small bulk, inexpensive and which can be implemented. work by a relatively unskilled operator.

 It is then possible to make recordings of data for capturing the position and orientation of said object at particular points of the geometric shape chosen by the operator.

 However, it is also possible to record data for capturing the position and orientation of said object at predetermined times.

 In a first advantageous application of the learning method according to the invention, there is proposed a method for learning a trajectory of a tool carried by a robotic machine within a workspace including a part to be treated by said tool, characterized in that a prior offline learning of said trajectory is carried out followed by a tool object representing said tool and moved by an operator, and in that this learning comprises a remote tracking method according to the invention for the acquisition of data relating to the position and the orientation of said tool object with respect to a fixed point.

This trajectory learning method comprises the following steps
- learning a benchmark associated with the part at
within the workspace,
- recording of the trajectory of the tool object
moved by the operator,
- referencing the trajectory with the
machine carrying the tool, and
- taking into account the mechanical constraints of the
robot to provide a set of instructions of
movement to be performed by said machine.

 It is particularly advantageous, but not essential, to also provide a step for calibrating the tool object, in order to define the position of a fixed point relative to sensors placed on said tool object. In addition, one can include in the method according to the invention a step of calibrating the working space.

 It is possible preferably to provide, following the step of recording the trajectory, a step of analysis and visualization of said trajectory which generally comprises a step for inserting additional instructions.

 In a second application of the learning method according to the invention, there is proposed a learning method applied to the digitization of a geometric contour of a part within a workspace to which a work mark is associated. , characterized in that an operator moves on the surface of said outline a feeler tool provided with mobile position and orientation sensing means cooperating with fixed position and orientation sensing means, in that one records position and orientation data of said feeler tool in the work coordinate system, and in that this data is processed to deliver a three-dimensional representation of the part to be digitized.

This scanning process preferably includes the following steps
- probe tool calibration, to define the
position of a fixed point with respect to
sensors placed on said feeler tool,
- learning of a coordinate system associated with the contour
geometric to digitize within the space of
job,
- recording of the tool path-
feeler moved by operator, and
- processing of position data and
the orientation of the feeler tool to provide
a three-dimensional representation of the outline
geometric.

According to another aspect of the invention, there is proposed a system for learning a trajectory of a tool carried by a robotic machine within a workspace including a part to be treated, implementing the method of learning according to the invention, comprising
- position sensing means and
orientation,
- means for calculating position and
recording of described trajectories,
- means to define a fixed reference linked to
the part to be treated, and
- means to calibrate the workspace,
characterized in that it further comprises an object representing the tool the trajectory of which must be learned, this tool object being moved by an operator and comprising a part of said position and orientation sensing means, the other part being linked to the treated part.

 In a preferred embodiment of a learning system according to the invention, the tool object is non-magnetic and the position and orientation sensing means comprise electromagnetic emission means, electromagnetic detection means and means for processing signals from the electromagnetic detection means.

According to yet another aspect of the invention, there is provided a system for digitizing a geometric contour within a workspace, implementing the method according to the invention, comprising
- position sensing means and
orientation,
- means for calculating position and
recording,
- means to define a fixed reference linked to
geometric outline, and
characterized in that it further comprises a feeler tool moved by an operator on the surface of the geometric outline and comprising a part of said position and orientation sensing means, the other part being linked to the workspace.

 Learning a trajectory of a tool carried by a robotic machine within a work space is achieved by moving an object in space and recording its position and orientation.

The object is for example moved manually and identified by a system of analysis of signals emitted by one or more transmitters arranged appropriately around the space of evolution of the object. No rigid mechanical connection is planned with the ground. Furthermore, the actual workspace is preserved and the operator works on a standard part as if he were performing the task manually. The method according to the invention makes it possible not to require heavy infrastructure for programming the machines. It can also be quickly adapted to any type of existing robot.

 Other particularities and advantages of the invention will emerge from the description below, relating to several nonlimiting exemplary embodiments.

To the accompanying drawings
- Figure 1 is a schematic view of principle
of a learning system according to the invention
- Figure 2 is a schematic view of a system
learning implemented in an application
spot welding; and
- Figure 3 is a flowchart of the main
stages of a particular embodiment of the
learning method according to the invention,
accompanied by schematic illustrations of these
steps.

 We will now describe a learning system according to the invention at the same time as the method implemented in this system, with reference to the aforementioned figures.

In a first embodiment illustrated diagrammatically in FIG. 1, a learning system 1 according to the invention comprises
- a system 13, 4, 5, 11, 16 for capturing,
position and orientation,
- a non-magnetic object 10, representing the tool
working 21 carried by a robotic machine 20,
which is handled by an operator 12,
- a system for calculating the tool positions and
recording of the described trajectories, which
can be a personal computer or a station
working 14,
- a fixed repository 6, 7, 8, 9 linked to a part to
treat 15 disposed on a support 2, and
- a workspace calibration system
3, which can be object 10 representing the tool
or a target with sensors.

The system for sensing the position and the orientation of the tool preferably comprises
- at least one transmitter or antenna 4, 5, consisting
for example three coils for coding
electromagnetic emission in a space
three-dimensional,
one or more electromagnetic sensors 11,
a system 13 for processing these signals, and
- an input console 16 connected to the
treatment 13 and used by the operator during
of learning.

 One can for example use the position and orientation measurement system 3SPACE FASTRAK marketed by the company POLHEMUS.

 But one could also consider the use of a system for measuring the position and orientation of a tool in space using a CCD technology mobile camera linked to the tool object and targets linked to the part to be treated or its support, as disclosed in the document FR 2706345 published on December 23, 1994. Other techniques for measuring position and orientation can be used, in particular techniques by laser or ultrasound.

 In a first preferred embodiment, the electromagnetic sensors 11 are fixed to the tool object 10 and are connected by flexible cable to the processing system 13. In another embodiment of a system according to the invention, it is the transmitter which is positioned on the object representing the tool while the sensors remain fixed.

 The calculation system 14 makes it possible to display the positions of the tool object 10 in the workspace 3 and to check their consistency. It can also be used to modify these positions.

 In an example of application of the present invention to learning a trajectory for an automotive welding robot, illustrated in FIG. 2, the learning system 30 comprises a measurement support 2 on which a workpiece is placed , in this case an automobile body 25, a control and processing system 26, a tool object 10 provided with three-dimensional electromagnetic sensors 11, two antennas or three-dimensional electromagnetic transmitters 4, 5. FIG. 2 a robotic machine 20 carrying a welding tool 21. The sensors 11 of the tool object 10 and the electromagnetic antennas 4, 5 are connected to a signal processing unit 23 which supplies a data processing unit 24 position and orientation of the tool object 10 in the working reference 6 when this tool object is moved along the path 29 by the operator 12. The control and processing system comprises also a control unit 27 which generates control instructions for the machine 20 via control lines 27 ′, as well as an input desk 36 allowing the operator 12 to enter during learning trajectory of additional instructions or commands. This input console 36 is for example connected to the signal processing unit 23 and cooperates with the latter and the calculation unit 24 to enter, store and process additional instructions. For example, in the case of trajectory learning for a painting robot, the operator can enter coded instructions on the input console 36 representing the type of spray gun (large spray gun, round spray gun, small spray) to be implemented according to the portion of trajectory learned.

The learning method according to the invention comprises, with reference to FIG. 3, the following steps
- CT calibration of the tool object 10, moved by
operator 12,
- CS calibration of workspace 3,
- AR learning of work mark 6 or mark
of the room,
- AND recording of the trajectory by
displacement of the tool object 10 in the reference frame
work 6,
- visualization and analysis VT of the trajectory 29,
- RT reference of path 29 with the
robotic machine 20 to be programmed,
- taking into account the mechanical constraints of the
machine 20 for the GT generation of the trajectory
in the language specific to this machine.

 The function of the tool calibration step CT 10 is to know the position of the tool center CO with respect to the sensors 11a, 11b, 11c. In the application of the present invention to an automotive painting robot, the CO tool center is in principle the point of impact of the paint on the bodywork.

 The object representing the tool, designated here by the term of tool object 10, which is used for the development of the trajectory 29, should preferably have a space as close as possible to that of the tool 21 which will be used on the machine 20. The operator 12 can thus better visualize its evolution in space 3. In addition, the tool object 10 must be light enough to be easily handled by the operator 12 and be non-metallic so as not to not disturb the functioning of the learning system 1.

 The step CT for calibrating the tool object 10 thus makes it possible to define the position of the fixed point called the tool center CO with respect to the sensors îî-c positioned on the tool object 10. The position of the tool center must be as precise as possible, because it is on the latter that the precision of the learning process according to the invention depends.

This CT calibration step can be performed in several ways
- by description of the position of the sensors 11 to
from a three-dimensional measurement of the object
tool 10, or from design files
computer aided (CAD);
- by a software function integrated into the
calculation 14. The operator 12 must then position
several times the CO tool reference
with a fixed tip, changing every time
the orientation of the tool object 10. As
example, a matrix calculation of the different
positions of each sensor 11 allows to redefine
the position of the CO tool center.

 The step CS of calibration of the workspace is essential insofar as it conditions the obtaining of high precision for the learning method according to the invention. It can be carried out in several ways depending on the characteristics of the learning system used.

It depends on the interference generated by the working environment on the position sensing system. This interference can also be eliminated by coatings ensuring electromagnetic filtering and deposited on metal parts. It is a question of characterizing the working environment in order to know the disturbances and drifts of signals.

 The calibration CS of the working space must be maximum near the workpiece 15, but may be lower in the areas where the machine 20 is not asked for a high position accuracy.

Several calibration methods can be implemented
A first calibration method consists in positioning a target 30 in the workspace 3.

This target 30 is provided with sensors 31, 32, 33 referenced with respect to each other. the number of sensors depends on the precision required. It is the same for the size of the target 30. Once the target 30 is positioned, the calculation system 24 reads the position for each sensor 31-33 and deduces the position errors read therefrom. By moving the target 30, it is possible to characterize the workspace 3. If the absolute position of the targets 30 is well known, the calculation method is simpler and faster.

 A second calibration method consists in pointing with a perfectly calibrated tool well-known positions of the working space 3. these positions can for example be positions referenced of the workpiece 15. They can be measured beforehand using a measuring machine. The calculation system compares with the positions read and deduces a cartography of the workspace 3.

 A third method consists in comparing the differences in positions read by several sensors positioned on the same perfectly balanced tool object 10. The operator 12 moves the tool object 10 in the workspace 3 and the calculation system 14 permanently reads the positions of the sensors 11. An analysis of the differences in positions read makes it possible to know the shape of the lines of electromagnetic field. .

 A fourth method can be designed from the first three methods in order to optimize either the accuracy of the calibration or the time required for this calibration.

 the preferred method is a fifth calibration method using digitization of the part to be treated. One can for example use the 3DRAW three-dimensional digitization system marketed by the company POLHEMUS. It is possible to calibrate the tool object without calibrating the workspace. A distorted three-dimensional representation of the part to be treated is then obtained with a contour recognition system. One then takes real particular points and one thus determines corrections to be made to the distorted representation. This method allows you to perfectly characterize the workspace with the outline of the room. It also has the advantage of allowing visualization of the part and the trajectory, during the analysis and visualization stage VT.

 Once the workspace is characterized and stored in a workstation file, this characterization can advantageously be used to determine possible differences between a standard part and similar parts produced in series. In fact, if representation deformations are detected for a serial part, they will then be due solely to imperfections in this part.

The step AR for learning a reference mark of the part 15, hereinafter called the reference mark-piece or work mark 6, makes it possible to easily readjust the trajectory with the machine 20 which will execute the trajectory 29. The reference mark-piece is learned using the perfectly calibrated tool object 10. The operator 12 points with this tool three or four points 7, 8, 9 perfectly defined in the workspace 3, on the workpiece 15 or on its support 2. Three points 7, 8, 9 are enough to learn the orthonormal coordinate system 6. As an example, the first point gives the origin of the coordinate system; the second point defines, in combination with the first, the axis X of the coordinate system; the third point defines, in combination with the X axis, defines the plane
XOY. A fourth point may be necessary if, for example, the first three points are taken in an imprecise manner. The orientation of the mark 6 is however correct, a fourth point making it possible to position the mark perfectly.

 The ET step of recording the trajectory is carried out with the tool object 10. The operator 12 moves the tool object 10 in the workspace 3.

 In a first embodiment, the operator 12 records the trajectory 29 continuously. For this, the operator 12 reproduces the movement which must be executed by the machine 20. The system 14 continuously records the movement of the tool object 10.

The time interval between two successive positions can be set. This first mode is, for example, implemented for painting or bead welding robots.

 In a second embodiment, the operator 12 learns the path 29 in point to point S1, S2, SN.

For this, the operator 12 positions the tool object 10 successively and requests the calculation system 14 to record its position on each point S1, S2, SN, by means of a trigger or a button. pusher. This second mode is for example implemented for spot welding robots.

 The step VT of analysis and display of the recorded trajectory 29 ′ allows the operator 12 to check the validity of the learned path and to insert additional instructions into it if necessary, or to modify certain points. The operator 12 accomplishes this task by viewing on the calculation station 24 the trajectory 29 ′ of the tool object 10. The instructions are events or orders which must be taken into account by the machine 20. By way of example , these instructions may relate to paint flow or air flow.

 The step RT of referencing the path 29 with the machine 20 makes it possible to define the position of the latter relative to this path 29.

For this, the operator 12 must relearn the three or four points 7, 8, 9 learned for the workpiece marker 6 with the machine 20. This operation can be carried out on a work station 24 if the machine and the working environment are modeled there. If the operation is carried out directly with the machine 20, the values of the points learned by the machine 20 must be supplied to the calculation system 24. It should be noted that certain robots require working relative to their own axes of rotation and not with respect to external points.

 The step GT of taking into account the mechanical constraints and of trajectory generation makes it possible to convert the trajectory 29 into the language of the robotic machine 20. The computer 27 must convert the learned trajectory as a function of the mechanical entity to be programmed, ie the robotic machine, the trajectory position 29 relative to the latter and its own language. The program which results from this step is either a program directly admissible by the machine 20, or a descriptive type file which will be transformed into movement instructions by the machine or another system.

An application of the learning process which has just been described can be provided for the digitization of a model produced by molding, for example the model of a new car. Thus, a system for digitizing a geometric outline within a workspace includes
- position sensing means and
orientation,
- means for calculating position and
recording,
- means to define a fixed reference linked to
geometric outline.

 It further comprises a feeler tool moved by an operator over the surface of the geometric contour and comprising part of the position and orientation sensing means, the other part being linked to the workspace. It can also be provided with means to calibrate the workspace. When no calibration of the workspace is planned, the digitization system then further comprises means for correcting data supplied by the processing means relating to the position and the orientation of the tool. - feeler at any point on the geometric outline.

 It should also be noted that, in the context of the learning method according to the invention, it is perfectly possible to combine several working marks. Thus, it is possible to envisage, for the same machine, successive teachings in several benchmarks. In addition, provision can also be made for learning several different trajectories in the same work reference.

 Of course, the invention is not limited to the examples which have just been described and numerous modifications can be made to these examples without departing from the scope of the invention. Thus, other position and orientation sensing devices can be used than the electromagnetic devices mentioned in the description above. Furthermore, provision can be made to multiply the number of electromagnetic transmitters and receivers in order to increase the measurement accuracy. The signal processing system can then be configured to discriminate between the position and orientation data delivered by the sensors in response to the emissions from the different transmitters so as to retain only the data ensuring the best measurement accuracy.

Claims (28)

 1. Method for learning a geometric shape, in particular a trajectory (29) of a tool or an outline (18) of a part (15), within a workspace (3), characterized in that an operator (12) moves over said geometric shape (29, 18) an object (10) provided with mobile position and orientation sensing means (11) cooperating with fixed means for sensing position and orientation (4, 5) linked to the workspace (3) which is associated with a work mark (6)  this object (10) being previously calibrated (CT), in that one performs (AND) a recording of capture data relating to the position and orientation of said object (10) and in that one processes (VT , RT, GT) these data recorded to provide three-dimensional data relating to said geometric shape (18, 29).  2. Method according to claim 1, characterized in that one performs (AND) data recording of position and orientation capture of said object (10) at particular points (S1, S2, SN) of the geometric shape chosen by the operator (12).  3. Method according to claim 1, characterized in that one performs (AND) recordings of position and orientation sensing data of said object (10) at predetermined times (T1, T2, TN).  4. Method according to any one of claims 1 to 3, applied to learning a trajectory (29) of a tool (21) carried by a robotic machine (20) within a work space ( 3) including a part (25) to be treated by said tool (21), characterized in that prior offline learning of said trajectory (29) is carried out during which an operator (12) moves on said trajectory (29) ) a tool object (10) provided with mobile position and orientation sensing means (11) cooperating with fixed position and orientation sensing means (4, 5) linked to the working space (3 ) which is associated with a work mark (6), this tool object (10) being previously calibrated (CT), in that one performs (AND) a recording of capture data relating to the position and the orientation of said tool object (10) and in that it processes (RT, GT) these recorded data to provide instructions for controlling the machine (2 0).  5. Method according to claim 4, characterized in that it comprises the following steps  - learning (AR) of a reference (6) associated with the  part (15, 25) within the workspace (3),  - recording (ET) of the trajectory (29) of  the tool object (10) moved by the operator (12),  - referencing (RT) of the trajectory (29) with  the machine (20) carrying the tool (21), and  - taking into account the mechanical constraints of the  machine (20) for providing a set  movement instructions to be performed by said  machine (20).  6. Method according to claim 5, characterized in that it further comprises a calibration step (CT) of the tool object (10), to define the position of a fixed point (CO) relative to sensors (11a, 11b, 11c) placed on said tool object (10).  7. Method according to one of claims 5 or 6, characterized in that it further comprises a calibration step (CS) of the working space (3).  8. Method according to any one of claims 5 to 7, characterized in that it further comprises, following the step of recording (ET) of the trajectory (29), a step (VT) d analysis and display of said trajectory (29).  9. Method according to claim 8, characterized in that the step (VT) of analysis and display comprises a step for inserting additional instructions.  10. Method according to any one of claims 5 to 9, characterized in that the step of recording (ET) of the trajectory comprises a step for inserting additional instructions.  11. Method according to any one of claims 5 to 10 and claim 7, characterized in that the step (CS) of calibration of the working space (3) comprises several positioning sequences within said working space (3) of a target (30) provided with position and orientation sensor means (31, 32, 33), each positioning sequence being followed by a position reading for each sensor (31-33) and d 'a calculation of position errors read.  12. Method according to any one of claims 5 to 11 and claim 7, characterized in that the step (CS) of calibration of the working space (3) further comprises obtaining a mapping of said workspace (3).  13. Method according to one of claims 5 to 12 and claim 7, characterized in that the step (CS) of calibration of the working space further comprises a comparison of the position differences read by several sensors (11 ) arranged on a calibrated tool (10), this calibrated tool (10) being moved by the operator (12), and an analysis of these position differences to obtain a calibration of the working space (3).  14. Method according to one of claims 5 to 13 and claim 7, characterized in that the step (CS) of calibration of the working space (3) comprises a digitization of the part (15, 25) at process within said workspace (3).  15. Method according to any one of claims 5 to 14, characterized in that during the step (AR) of learning the work mark (6), the operator (12) points with the object -tool (10) at least three points (7, 8, 9) predefined in the workspace (3).  16. Method according to any one of claims 5 to 15, characterized in that during the step (ET) of recording the trajectory (29), the operator (12) reproduced with the object- tool (10) the movement which must be executed by the tool (21) carried by the machine (20), and the movement of said tool object (10) is continuously recorded.  17. Method according to one of claims 5 to 15, characterized in that during the step (ET) of recording the trajectory (29), the operator (12) positions the tool object ( 10) at successive points (S1, S2, SN), and the position of said tool object (10) and the positioning instant (T1, T2, TN) are recorded at each of said successive points (S1, S2, SN).  18. Method according to one of claims 1 to 3, applied to the digitization of a geometric contour (18) of a part (15) within a workspace (3) previously characterized which is associated with a reference working (6), characterized in that an operator (12) moves on said contour (18) a feeler tool (10) provided with mobile position and orientation sensing means (11) cooperating with fixed means sensing position and orientation (4, 5), in that the position and orientation data of said feeler tool (10) are recorded in the working reference (6), and in that processes this data to deliver a three-dimensional representation of the part (15) to be digitized.  19. The method of claim 18, characterized in that it comprises the following steps  - calibration of the probe tool (10), to define  the position of a fixed point with respect to  sensors (11) placed on said feeler tool (10),  - learning of a reference (6) associated with the contour  geometric (18) to be digitized within the space  working (3),  - recording of the tool trajectory  probe (10) moved by the operator (12), and  - processing of position data and  the orientation of the feeler tool (10) to  provide a three dimensional representation (19)  of the geometric outline (18).  20. The method of claim 18, characterized in that it further comprises a step for correcting the recorded data, this correction step comprising a recording of particular predefined points of the geometric contour (18).  21. System (1, 30) for learning a trajectory (29) of a tool (21) carried by a robotic machine (20) within a work space (3) including a part to be treated ( 25), implementing the method according to one of claims 4 to 17, comprising  - means (11, 4, 5, 13) for sensing position  and orientation,  - means (14) for calculating position and  recording of described trajectories,  - means (10, 14) for defining a frame of reference  fixed linked to the part to be treated (15), and  - means (10, 14) for calibrating the space of  work (3),  characterized in that it further comprises an object (10) representing the tool (21) whose trajectory must be learned, this tool object (10) being moved by an operator (12) and comprising a part (11) said position and orientation sensing means, the other part (4, 5, 13) being linked to the part to be treated (25).  22. Learning system (1, 30) according to claim 21, characterized in that the tool object (10) is non-magnetic and in that the position and orientation sensing means comprise emission means electromagnetic (4, 5), electromagnetic detection means (11) and means (13) for processing signals from the electromagnetic detection means (11).  23. Learning system (1, 30) according to claim 22, characterized in that the electromagnetic emission means (4, 5) are linked to the part to be treated (15, 25) or to its support (2) , and the electromagnetic detection means (11a-c) are arranged on the tool object (10).  24. Learning system according to claim 22, characterized in that the electromagnetic emission means are arranged on the tool object, and the electromagnetic detection means are linked to the part to be treated or to its support.  25. Learning system (1) according to any one of claims 21 to 24, characterized in that it further comprises means (16, 36) for entering data, and in that these entry means ( 16) operated by the operator cooperate with the position and orientation sensing means (13, 23) and with the calculation means (14, 24) for entering and storing additional instructions during the learning of trajectory .  26. System for digitizing a geometric contour within a workspace, implementing the method according to any one of claims 18 to 20, comprising  - position sensing means and  orientation,  - means for calculating position and  recording,  - means to define a fixed reference linked to  geometric outline, and  characterized in that it further comprises a feeler tool moved by an operator on the surface of the geometric contour and comprising a part of said position and orientation sensing means, the other part being linked to the workspace.  27. Digitization system according to claim 26, characterized in that it further comprises means for calibrating the workspace.  28. Digitization system according to claim 26, characterized in that it further comprises means for correcting data supplied by the processing means relating to the position and the orientation of the feeler tool at any points. of the geometric outline.