ITPI20010007A1 - METHOD FOR THE CONTROL OF AN ARTICULATED AND / OR DEFORMABLE MECHANICAL SYSTEM AND ITS APPLICATIONS - Google Patents

Description

Description of the Industrial Invention entitled:

METHOD FOR THE CONTROL OF AN ARTICULATED MECHANICAL SYSTEM AND / OR

DEFORMABLE AND ITS APPLICATIONS "

The present invention relates to a method for

control of an articulated and / or deformable mechanical system,

such as for example an apparatus for making complexes

movements to objects, an automaton or an artificial face.

The invention also relates to operating applications

according to the above method.

Mathematical models are known which, based on data

preset or detected by sensors, allow you to manage

simultaneously a limited number of actuators, correlated or

unrelated, in such a way that with the set of them

actions, synergistic and / or competing, it is possible to determine,

in space and time, the desired evolution of the structures is connected to them and / or the variables controlled by them. THE

complex mathematical models that address these problems

they try to describe, by means of systems of equations

integro-differential, the dynamics of the actuators and of the

structures connected to them taking into account the constraints and mutual influence of the various elements that make up the controlled system. Once the desired evolution has been established for the latter, for example a sequence of movements, the model is used to derive the values to be assumed over time by the joint variables in the various actuators. The results of the calculations are then translated into real motion of the controlled system, providing their driving system with the values that the joint variables of the various actuators must assume in space and time.

As an example of what has been stated, the implementation of the movement of a hexapod is mentioned, which represents the best that can be done with traditional control systems. To move this structure it is necessary to create a motion in which the points of contact between the ends of the actuators and the surface vary over time with discontinuity. Classically, in order for the robot to carry out this movement, it is necessary to write the equations that correlate the joint variables of all the actuators and solve these equations over time to obtain the succession of the states of the various motors instant by instant. For a detailed discussion see Haruhiko Asada, Jean-Jacques E. Slotine, Robot Analysis and Control, John Wiley & Sons, April 1986. References to current androids, managed with classic control systems and therefore intrinsically limited in their functions, are: the humanoid "Cog", invented by Rodney Brooks of MIT in Boston, able to emulate the movements of the human upper limbs, or "M2", also of MIT, able to walk. Another example is NASA's "Robonaut", more precisely of the Johnson Space Center in Houston, which will be used to carry out spacewalks working to repair probes and space stations in space in place of astronauts. Again we can mention "Jack", from the Electrotechnical Lab of Tsukuba (Japan), an android capable of helping the elderly to live alone. At the Science University of Tokyo they are also making "Face Robot", a project led by Hideotoshi Akasawa. This android face will be able to recognize and react to different expressions.

Bibliographic references relating to the above mentioned androids are contained in Peter Menzel, Faith D'Aluisio, Robosapiens, evolution of new species, MIT Press, 2000.

The object of the present invention is to make possible the control of articulated and / or deformable mechanical systems of even complex structure for which the description with a system of equations according to the known art would be difficult or impossible.

Another object of the present invention is to provide a control method for articulated and / or deformable mechanical systems using a virtual graphics development environment thanks to which it is possible to interpret the data provided by a set of sensors and simultaneously manage a large number of actuators, with the possibility of constructing a model that closely corresponds to reality both in terms of dynamics of the actuators and of induced deformations, in such a way as to be able to plan and control, within the virtual environment, what will be the succession of movements that will be they want to get.

A further object of the present invention is to provide a control method of the aforementioned type in which it is possible to build an unlimited database of configurations and / or movements.

It is still another object of the present invention to provide a control method of the aforementioned type thanks to which it is possible to obtain new movements on the basis of the database obtained without necessarily recalculating them with the virtual model, that is to study off-line the desired range of movements, to save it in the database and with it control the movements and / or the induced deformations of the real actuators.

The essential characteristic of the control method according to the present invention consists in replacing the classical mathematical models with 3D solid models built in a virtual environment, using, in a conceptually new way, advanced three-dimensional graphics software instead of systems of equations. In other words, the basic idea is to use information and data obtained from objects created in a virtual environment to drive similar real objects.

More precisely, with an advanced graphics system the model of a real structure, even a complex one, is reproduced in detail and a virtual model is created for each of its actuators. The constraints and limits possessed by the real object are attributed to the virtual model thus created, since the greatest degree of freedom exists in the virtual environment. As in the classic case, the key information that the model provides for controlling the actuators are the values to be assigned over time to the joint variables, such as the coordinates of the moving end of a linear actuator, or the angle of rotation of the axis of a motor. This information, after being sent to a piloting system, makes it possible to realize what is simulated in the virtual environment in reality.

The advantages and application potential of this way of operating are considerable. It is possible to model very complex real systems for which the description with a system of equations would be difficult or impossible, having as the only limit the computing capacity of the computer used to create the virtual environment. Furthermore, it is possible to eliminate all the problems connected with writing and solving very complex systems of equations and therefore the number of actuators that can be controlled at the same time increases considerably. In this way, therefore, it is possible to obtain the results of classical robotics with greater ease, solving the problem of the kinematics of the various actuators, of which it is necessary to find the values assumed in the time of position, speed and acceleration.

Further characteristics and advantages of the method for controlling articulated and / or deformable mechanical systems according to the present invention will become clearer from the following description of a practical application thereof, provided by way of non-limiting example.

In the accompanying drawings:

Figures 1, 2 and 3 schematically show the cycle of operations through which it is possible to obtain the movement of a real system with the control method according to the invention;

Figure 4 schematically shows the components of an apparatus operating according to the control method object of the present invention;

figure 5 schematically shows the motorization system of the embodiment example of the method according to the invention;

Figure 6 schematically shows an actuator used in said example;

Figures 7 and 8 show a side view and a front view of a motor unit;

Figure 9 schematically illustrates the driving electronics of the motors;

Figure 10 schematically shows the modeling step of the application example according to the present invention;

Figure 11 shows an example of a matrix used for driving the application example according to the present invention;

- figure 12 schematically shows the hierarchy of bones and locators.

With reference to figures 1 and 2, in its more general execution mode, the control method according to the present invention provides a first phase in which the components and their mutual relationships are analyzed, according to the application objectives, of an articulated mechanical system and / or deformable having sensors and / or actuators as interface, hereinafter referred to as "real system" for brevity, which must be managed by the control system. A virtual model is then created, using advanced 3D graphics software, building the three-dimensional models of all the structures and all the actuators that make up the real system in the virtual environment. These models are detailed by parameterizing the variables that are used to move the corresponding real components and by modeling the constraints and interdependencies to which said components are subject.

The virtual model thus created is then graphically "animated", ie moved and / or deformed, in the desired way and the information necessary for the control of the real system is extracted from the previously parameterized variables. In particular, model animation can be simplified with the use of motion capture techniques. According to these techniques, the movement to be reproduced is acquired by the real system with appropriate detection systems, processed with special software and the information useful for the construction of the animated sequences is supplied to the virtual model which reconstructs the positions of the actuators and the trends of the variables from them. parameterized.

The information extracted from the model is then input to the control drivers for driving the real system.

By way of example, the software called MAYA, marketed by Alias / Wavefront, or equivalent, can be used as an advanced 3D graphics system. For the motion capture techniques, the software called Gypsy Motion Capture System marketed by Meta Motion, or equivalent, can be used and the detection systems for the acquisition of the movement to be reproduced can be constituted by magnetic sensors, or equivalent. The MAYA software itself or the Filmbox Animation software marketed by Kaydara, or equivalent, can be used as processing software for the data acquired with these detection systems.

The realization of the virtual model is aimed at obtaining the variables necessary for driving the articulated and / or deformable mechanical system which, in its most general form, is composed of n actuators, k constraints, j joints and p passive subsystems moved and / or deformed by the actuators. These variables represent the control parameters with which to drive a large number of actuators in space and time in parallel so as to make the real system and / or the highly complex physical systems dependent on it assume the desired movements and / or deformations. To best construct the virtual model of this system according to the method object of the invention, the passive subsystems are projected into the virtual environment by defining their movement and / or deformation properties while maintaining their proportions and mutual distances unchanged (modeling phase ). The actuators, constraints and joints are graphically created and located in their correct spatial positions. The system control variables are then defined, ie the values to be extracted, in space and time, from the virtual model to allow the control of the real system (parameterization phase). The space-time motor patterns of each individual actuator (animation phase) are created according to the possible methods:

by setting a certain trend of the control variables of the actuators, the model returns the movements and / or deformations induced on the real system (direct kinematics);

- from movements and / or deformations induced graphically on the virtual model it is possible to derive the trend of the control variables of the actuators that allow them to be realized in the real system (inverse kinematics);

- from movements and / or deformations acquired through motion capture techniques, both offline and in real time and projected on the virtual model, it is possible to obtain the trend of the control variables of the actuators that allow them to be created, being the model interfaceable with the most common commercial motion capture systems.

The passive subsystems moved and / or deformed by the actuators can be reproduced in a graphic way having the foresight to create them similar from a geometric point of view (in scale), keeping the distances between them proportional.

For the parameterization phase, in which the models of actuators, constraints and joints are created, the following steps are carried out:

- a basic skeleton is made;

- the volumetric properties of it are created;

- the base skeleton is placed in space; the handling and / or deformation properties are defined;

- the influence that each individual actuator and / or joint has on the passive subsystems dependent on them is defined.

To create the basic skeleton, what is called "bone" (skeleton) is created in the most common 3D graphics software. It is a vector oriented with respect to a reference frame placed on its application point. To assign volumetric properties to it, a cylinder is drawn which, positioned in correspondence with the bone, is grouped into a single structure with the bone. The generic command is create group, this group can be called an actuator, whose application point represents a joint. This object is positioned and scaled based on the position and size of the actual object. The joint of this object can be bound or not to other structures thus defining its movement. To define the individual deformation properties, each actuator is associated with a flexor, a feature found in graphics software; is one of the characters that can be assigned to elements of a virtual environment. In order for the flexors to have an influence on the passive subsystems dependent on them, they are interconnected both with the latter and with the actuator, selecting them all at the same time and correlating them through the generic rigid bind command. The areas of effect and influence of each flexor on the deformations and / or movements of the dependent subsystems are specially modified using the paint tool panel. Being able to adapt the virtual structure to any subsystem and any real configuration of actuators, it follows unlimited potential.

In the animation phase, through the structures previously developed, the space-time motor patterns of each individual actuator are implemented according to the methods previously indicated. The spatial trend of the movements and / or deformations of the actuators and their interconnected structures can be defined through the spatial keys of the tool set driven key, present in all 3D development environments. Using this tool, it is possible to constrain the geometric parameters corresponding to movements and / or deformations of the virtual structures created to known values (keys). The software will move the model according to the trajectories that satisfy the characteristics of the model created up to this moment and the passage through these known values according to one of the three methods previously described. The associated deformations are therefore fixed by replicating the real physical movements and / or deformations that occur in the actuator itself.

The control variables of the virtualized actuator are associated with another structure, the locator. It is a constraint structure that can be associated with one or more elements and influences their parameters. In this way, attributes of the locator (control variables) are created, the variation of which determines a numerical variation of the corresponding geometric parameters of the flexor passing through the spatial keys.

The trend over time of the movements and / or deformations of the actuators and their interconnected structures can be decided by defining the time keys with a time slider, tool. present in all 3D development environments. The states of the model are fixed in the time slider at the desired instants for each movement. The software will build all the intermediate states of the movement of the actuators over time.

Therefore, by acting on each individual actuator, movements and / or spatial-temporal deformations and their interconnected structures are obtained. If you want to act directly on the structures interconnected to the actuators and obtain the consequent movements and / or deformations of the actuators, it will be necessary to enclose these structures in another structure present in the 3D software, the deformation lattice.

During the execution of the movements and / or deformations in the virtual model, the values over time of the previously defined control variables are extracted and saved. This procedure can be simply carried out with the appropriate commands of the 3D graphics software, such as getAttr and fwrite. These values are saved in an ASCII file and represent the values that the control variables of the real system must assume over time (set points). The file will contain tab delimited values in matrix form. In each column the control variables will be represented, actuator by actuator, while in the lines the time is represented. The file is therefore in a standard format commonly accepted by any system and represents the input to the actuator control drivers. The drivers therefore receive in input the set points over time of the corresponding variables that control and drive their motion.

All these files together make up a database to draw from. Any decision-making algorithm, artificial intelligence software, data analysis, etc., or simply an operator, can update the movement to be carried out in real time by drawing it from the database. Piloting can take place in two ways:

- off-line operation

- online operation

In the first case, the decision-making algorithm will operate by drawing the movements from a mass memory support of a computer. In the second case, the information necessary to drive the actuators will be extracted in real time from the movement of the virtual model and sent, again in real time, to the control drivers. The latter procedure is illustrated in Figure 3.

Figure 4 illustrates the general ways in which the method object of the present invention can be put into practice. In it, 1 indicates a computer, in particular a personal computer, on which the advanced 3D graphics software is implemented, connected via RS232 serial connection to a control electronics indicated as a whole with 2, in turn connected to a driver of a closed-loop controlled motors unit. Control transmission means 3 exit from the motor unit, generally consisting of transmission cables, which act on a system 5 to be controlled comprising actuators 6 and subsystems 7 to be moved and / or deformed.

In the following description, the articulated and deformable mechanical system controlled with the method according to the invention consists, by way of non-limiting example, of an artificial face, called for simplicity "android head", whose aesthetics, expressiveness and mimicry are similar to human ones. According to the general operating principle described above compatibly with the available technologies, with reference also to figure 5, the android head, generally indicated with 10, comprises a skeletal support structure 11 to which artificial muscles (actuators) 12 are attached, connected to a artificial leather 13 that covers the assembly. The actuators 12 are operatively connected, by means of transmission of mechanical power 14, to a group of motors 15 and encoders 16, which by means of an electronic motor control (motor driver) 17 are controlled by signals obtained from a virtual model of the human face reconstructed in an advanced 3D graphics environment loaded on personal computer 1 equipped with Windows NT 128MB RAM operating system and graphics card with OpenGL libraries or equivalent systems. The system components are described in more detail below.

Support skeletal structure

The support structure 11 is a resin reproduction of a skull and part of the human cervical spine intended to act as an anchor for the actuators 12, and to be covered with artificial leather 13. The structure has an internal compartment (the one normally reserved to the brain mass) within which are housed, by means of constraints, sheaths for the transmission cables 14 which transmit the mechanical power from the motor group to the actuators. The sheaths innervate the support structure through through holes allowing the positioning of the cables tangential to the skin.

The artificial leather 13 is made in the present embodiment, by means of a completely manual technique, first creating an alginate cast from life, on which, after solidification, a cast of alabaster gypsum is manually placed which, after solidification, will constitute the negative of the face from which at the end, a mold is obtained into which silicone rubber will be poured which, once solidified, will form the artificial skin.

Artificial Muscles

The actuators 12 which perform the function of the muscles are a silicone structure 20 with a substantially elongated elliptical shape, schematised in Figure 6, the core of which is the transmission cable 14 of the mechanical power drawn from the motor unit. The end of the sheath 21 within which the cable is arranged is fixed to the support structure at the point of origin of the muscle, while the transmission cable 14 is bound to the skin by any suitable means, generally indicated with 22, for example by tightening by means of a grain its end in the hole of a metal cylinder embedded in a silicone protuberance of the internal surface of the skin at the point where the real muscle ends. The actuators are made by molding by pouring the liquid silicone into a hollow mold having the shape of the specific actuator and letting it dry. It is therefore possible to create actuators of any shape and size. The linear movement of the cable 14 inside the silicone mass 20 causes its longitudinal shortening and a consequent transverse swelling, allowing both to pull the skin 13 and to deform it in the desired way. The artificial muscle therefore exchanges with the skin both tangential and normal actions to the internal surface of the skin due to contact with its ellipsoidal surface, and those due to the direct action of the end of the cable at the point where it is anchored to the skin itself. The stiffness of the cable sheath transmission system allows the feedback to be carried out for the control of the actuators by measuring the displacement directly on the motors and not on the actuators.

Engines group

As shown in Figures 7 and 8, the motor group consists of twenty-six linear motors divided on two banks. Each motor is composed of a coil integral with a pivoting arm 4 28 immersed between two fixed magnets 29 (north and south) in the shape of a 45 ° circular sector. In each bank, the rocker arms of each motor are equally spaced from each other and rotatably mounted on an axis 30 independently of the others. Finally, the magnetic circuit closes on a frame 31 which supports the fixed magnets 29 and the rotation axis 30. As a result of the rotation, the linear displacement of the end 28a of each rocking arm 28 is obtained from which the mechanical power is taken. necessary. In this way, with a maximum rotation angle of 45 °, a maximum linear displacement of 20 mm is obtained. These motors are capable of delivering a force of 12.5 N at start-up with a maximum current absorption of 0.5 A. The maximum achievable acceleration is 15.7 m / s <2>, while the maximum speed is 0.78 m / s. Repeatability and precision depend on the characteristics of the transducer 32, which in this case has the Hall effect, and on the driving mode which normally allow to reach repeatability of ± 5 μm and precision of ± 0.08 mm. The cable 14 necessary to transmit the power from the motor to the respective actuator is fixed to the power end 28a of the rocking arms by means of a dowel 33, while the sheath 21 is referred to the bank by means of a threaded bushing 34 which engages in the holes made on a plate 35 placed on the front of the bank. A fan system, not shown, ensures the convection necessary for the dissipation of the heat produced by the motors.

Motor control electronics

The electronics have been realized with a modular philosophy and allows the housing of the drivers of the motors controlled in closed cycle, of the programmable microprocessors, of an electronic card for interfacing with a personal computer and of the power supplies. Figure 9 schematically shows the control electronics. In the specific case each module, indicated generically with 36, houses a programmable microprocessor 37 and four drivers 38 for the motors. The motherboard 41 carrying the modules 36 includes an RS232 serial connection indicated with 39 and a power supply 40. The protocol has been organized on 3 bytes: the first for the addresses, the second for the position and the third for the speed of movement . The baud rate is 57.6 Kbaud with a command execution time of 520 μm; all the actuators execute the command in 12.5 m / s. The microprocessor can be programmed via software. The control is carried out in a closed cycle by reading the angular velocity of the coils 27 by means of the Hall effect sensors 32 suitably calibrated. With this system the motors can be operated according to very accurate dynamics, until deformations of the actuators and of the artificial skin corresponding to those foreseen by the virtual model are obtained.

Virtual model

The creation of a virtual model of a human face allows to derive the space-time trajectories necessary for piloting an anthropomorphic head or android head (real system) capable of simulating human mimic expressions. The virtual model of the human face created highlights the fact that with this technique it is possible to control a large number of actuators in parallel so as to make even highly complex physical systems dependent on them assume, such as the artificial skin described above , the desired deformations and therefore the required facial expression.

To best reproduce a human face, the physical model of the skin is projected into the virtual environment while maintaining its proportions unchanged and defining this structure as a deformation surface (modeling phase); the virtual model of each individual muscle is then created by parameterizing it and locating it in the correct position (parameterization phase); the dynamics of the mimic expressions are then created by defining the spatial-temporal motor patterns to be associated with each individual actuator to obtain the desired surface deformations (animation phase).

a) modeling

To project a surface and its reference system from physical space into the virtual development environment, it is necessary to identify a certain number of points by tracing meridians and parallels on the surface itself and use a feeler device. The latter is a commercial movable member consisting of a base, three mechanical arms interconnected by relative joints and a punch on the end. By means of a manual positioning system, the feeler allows the punch to identify any position in the space with respect to its base; the location comes RS232 to the pc. The points previously traced on the surface to be projected are thus probed, mapped in a three-dimensional reference system and brought back into the virtual development environment. These devices, thanks to their management programs, are compatible with most commercial software. The development environment recreates from these points the parallel lines that interpolate them and from them the surface model felt. As also schematized in Figure 10, in the specific case a surface model was created consisting of thirteen lines of surface section, counterclockwise, starting from the middle of the face and fifteen lines of cross section starting from the mouth. To create oral cavity spacing, you can manually add points for the inside of the mouth to these curves. Selecting all the curves thus obtained will create one

surface using the loft command common for 3D software. Obviously the whole face both real and in fantasy forms can also be created manually.

b) parameterization

To create the models of the artificial muscles, the following steps were followed:

making a basic skeleton;

creation of its volumetric properties duplication of this structure (actuator and base) and positioning of its clones in the application sites of the physical actuators (artificial muscles)

definition of the deformation properties of each individual actuator;

- indication of the influence that each individual actuator has on the surface. .

To create the basic skeleton, a bone is created as described previously, which is assigned volumetric properties. This structure represents a generic actuator that is positioned and scaled according to the position and size of the artificial muscle. The point of application of the actuator must be positioned at the point of attachment of the artificial muscle on the support structure and its end at the point of attachment of the artificial muscle on the artificial skin The opening / closing of the mouth is implemented by creating a bone control and positioning it on the jawline. It is positioned where the jaw really rotates during the opening of the mouth and chained with the four actuators of the part below the jaw. This allows a rotation on the new bone created with respect to a fixed point, so as to define by rotation and not by translation the opening and closing of the mouth itself.

Once all the actuators have been positioned for half of the face, a basic support bone is constructed. The two half faces therefore have their own base bone. We then move on to the creation of a general control parent bone which is connected to the control bones of the two half-faces and therefore to all the actuators, thus creating a hierarchical structure starting from the parent bone. Once this structure is obtained, the actuators residing on the median axis of the face and a parent bone of them are created. The parent bone of the mid-axis actuators is also enslaved to the general control parent bone.

To define the individual deformation properties, as previously indicated, a flexor is associated with each actuator. In order for the flexors to have influence on the face, they are interconnected both with the cylindrical surface around the bone and with the face surface, selecting them simultaneously with the respective bones, the face surface and the cylinders and correlating them using the rigid bind command. Once all the flexors have been created and sized appropriately, the influences they have on the surface are indicated. The areas of effect and the deformations of each actuator are modified by observing the influence that these will have on the surface of the face. Through the special paint tool panel it is possible to highlight the influence of flexor on interconnected structures and modify it easily. In this way the expressive potential is unlimited, as the virtual structure can adapt to any face and any real configuration of the actuators.

In addition to the shape, it is possible to equip the virtual face with colors, shades and shadows until it is completely similar to a human face. This procedure can be performed directly from the virtual graphics software or exemplified by applying a texture. The texture is an image that, mapped on the surface, follows its deformations. As a texture it is therefore possible to select the digital photo in bitmap format or equivalent of the desired face. The texture, while being projected from a plane, adapts in 3D and automatically changes dynamically with the surface. c) animation

In this phase, through the structures previously developed, the dynamics of the mimic expressions are implemented by defining the spacetime motor patterns to be associated with each individual actuator to obtain the desired surface deformations.

c1) Spatial deformations

The trend in space of the deformations of the actuators and of the face surface can be decided by defining the spatial keys through the tool set driven key. Using this tool it is possible to constrain to values (keys geometric parameters corresponding to deformations of the virtual structures created. The software moves the model according to trajectories that satisfy the characteristics of the model created up to this moment and the passage through these values. The associated deformations are therefore fixed by replicating the real physical deformations that occur in the actuator itself.In this specific case we have limited only the minimum and maximum parameters of extension / shortening and swelling of each flexor by acting on the scale factors along the longitudinal and transverse axis.

In order to be able to control the scaling factors of several actuators at the same time, you can use the locators (in common 3D software) organized hierarchically. The locator is a structure to which it is possible to associate one or more elements and influence the variables at the same time. For each locator the associated elements are the actuators while the variables are the scale factors. In this way, attributes of the locator are created, the variation of which determines a numerical variation of the corresponding geometric parameters of the flexor passing through the spatial keys. The procedure chosen is to create a hierarchy of locators that controls the influence of the various actuators, grouping adjacent actuators into a single locator. In the specific case, the locators created are:

- mouth: zygomatic muscles, risorium, mouth corner depressor, upper lip levator and lower lip depressor;

- forehead: eyebrow and frontal corrugator muscles;

eye: eye muscles;

- nose: levator muscles of the upper lip adjacent to the nose;

maxilla: muscles that move the jaw.

Locators created separately are in turn interconnected in groups up to the parent locator. Thanks to this hierarchical structure it will be possible to act on the attributes of the parent locator to simultaneously influence the movements of all the actuators in the desired way. In fact, the expressions are defined as attributes of the parent locator. In the specific case, the expressions smile, disgust and indecision have been defined. Bearing in mind that man is able to make a very large number of expressions, it is possible to model approximately forty basic expressions, combining which, we obtain the majority of human facial expressions. The basic expressions can be divided into separate groups: forehead, eyes, nose, mouth and jaw. For a detailed discussion see Peter Ratner, Mastering 3D animation, Allworth Press, New York, 2000. By appropriately varying the percentages of the basic expressions over time, the desired expression is obtained. In the specific case, an expression of disgust can be obtained by deforming the variables of the previously defined locators by 100%. The panorama of possible expressions and combinations is unlimited. Through a variable it will be possible to monitor which expression is made to assume the model.

c2) Temporal deformations

The trend over time of the deformations of the actuators and of the face surface can be decided by defining the time keys, time elider, a tool present in all 3D development environments. In the time slider, the states of the model are fixed at the desired instants for each expression. The software reconstructs all the intermediate states of the deformation trend of the actuators over time and therefore the expression.

Therefore, by acting on the space-time deformation of each actuator, the deformation that the skin undergoes can be varied until the desired expression is obtained. If you want to act directly on the skin surface and observe the consequent deformations of the actuators, it will be necessary to enclose the external surface in another structure present in the 3D software, the deformation latex. This structure encloses the surface of the face in a dense three-dimensional lattice as desired. By acting on the points of this grid it is possible to modify the expression at will.

Control software for interface between the virtual model and the real system

During the execution of the expression, the spatial-temporal deformations of the actuators are saved in an ASCII file and renamed according to a correspondence table between the previously defined metric scale and verbal attributes. The file will be an array consisting of tab-delimited real values. The columns are the actuators, the lines the time expressed in frames (25 frames / second) and the elements the percentage of deformation of the linear actuator, obtained from the previously defined attributes. This matrix is of the format accepted by the driving software. A typical matrix obtained for the aforementioned disgust expression is shown, for a half face, in Figure 11.

Selection of the expression for driving the system

The set of expressions obtained will constitute a database from which to draw. In this specific case, a function has been written which provides as input which expression or combination to adopt. To each expression of the database, a natural number corresponds (0 is the rest value), while to each combination the real number to a decimal digit (10 values) between one expression and another. By increasing the decimal places you can increase the possible combinations. An ASCII file will contain the correspondence table between the verbal definitions of the expressions and the requested metric scale. This function is always active. Any artificial intelligence software, data analysis, decision algorithm, operator, etc. will be able to update the expression to be carried out in real time simply by updating the input values to the function. In the specific case, we have carried out the simulation of a chewing with subsequent hedonistic response to the food tasted. The above function can be exemplified with:

global proc function ($ expression)

{

select locator_father, -setAttr locator_father.expression $ expression; playButtonStart;

playButtonForward;

}

Piloting

Piloting can take place in two ways:

- off-line operation

- online operation

In the first case, the driving software, based on the choice made by the decision algorithm, will operate by drawing the previously calculated expression matrices (trajectories) from a mass memory support in the computer. In the second case, having a latest generation work station available, the virtual environment and the driving software can interact with the decision algorithm by calculating the expression to be performed in real time and driving the linear actuators. Advantageously, common algorithms can be used to correct any deviations between the set trajectory and that actually performed due to errors and / or defects that cannot otherwise be repaired in the articulated mechanical system. A correction algorithm known in the common literature that can be used is Adaptive Learning or equivalent. For more details see: International Workshop On Nonlinear and Adaptive Control: issues in Robotics, Grenoble, France Nov 21-2, Carlos Caundes De Wit (Publisher), 1991; John H. Andreae, Associative Learning: for a Robot Intelligence, Imperial College Press, October 1998.

From the foregoing it is clear that with the method according to the present invention, thanks to the use of a virtual graphics development environment, it was possible to build a model that closely corresponds to a real human face, both in terms of dynamics of the actuators and of skin deformities. This has made it possible to plan and control, within the virtual environment, what will be the succession of expressions to be obtained from the real system, even before it even moves one of its actuators.

With this technique it is also possible to build an unlimited expression database, to obtain new expressions on the basis of the database obtained without necessarily recalculating them with the virtual model, to obtain expressions corresponding to the pronunciation of human phonemes.

Furthermore, the model can be developed with the features of any desired human face and can include the most varied configurations of subcutaneous actuators corresponding to the real ones. Thanks to it it is possible to study off-line the desired range of expression, save it in the database and with it control the real movements of the actuators. The above example highlights the fact that it is not only possible to control a large number of actuators with the method according to the invention, but even highly complex physical systems dependent on them, such as, for example, the artificial skin that reproduces the human face. stretched on a support structure. In addition, once the model has been created, it can be used to obtain information in both directions:

by setting a certain motion of the actuators, the model returns the changes in the expression that this motion induces on the skin,

- starting from a sequence of expressions it is possible to derive the motion of the actuators that allows to realize it.

With the use of motion capture techniques, it is also possible to develop a control system that captures a subject's facial expressions in real time and reproduces them using the real model.

In addition to the application illustrated in the example, the method according to the present invention can be used for the realization, handling and control of anyhow complex articulated and / or deformable mechanical systems, such as for example: anthropomorphic articulated mechanical arms; anthropomorphic mechanical hands; articulated hexapods; androids and / or animatrons of any form; androids, puppets and / or animated dolls; androids, animatrons, mechanical animals, fictional beings and / or puppets capable of mimicking expressions and / or speaking. The fields of application of the aforementioned systems are: robotics, mechanics, cinematography, entertainment, world of toys.

Possible further applications of the system object of the invention can therefore be the following:

- control of an electronic actor, that is an animatrone or his model capable of expressing human movements, expressions, mimicry and dialogues;

control of an electronic teacher, that is an animatrone or his model able to teach by carrying out lessons and talking with the students;

control of replacement prostheses of parts of the human body, such as: arms, legs, fingers, face and organs; and / or orthotic prostheses,

- study, and / or control of demonstrator prototypes, of organic lesions and functional impairments caused by diseases, such as for example: malfunction of muscles and / or organs in general;

- control of an electronic physiotherapist in which models of organs to be re-educated indicate the optimal exercises and evaluate their effectiveness.

Variations and / or modifications may be made to the control method of an articulated and / or deformable mechanical system according to the present invention, without thereby departing from the protective scope of the invention itself.

Claims (21)

CLAIMS 1. Method for the control of an articulated and / or deformable mechanical system, comprising passive subsystems to be moved and / or deformed by means of actuators connected to each other and to said subsystems by means of joints and subject to constraints, characterized by the fact of including the following steps: create a virtual model of said system using advanced 3D graphics software; - defining in said virtual model the control variables necessary for driving said system; - animating said virtual model in the desired way to extract from it the time values of said control variables; - save the extracted values; sending the saved values to the control drivers of the actuators of said system for piloting the latter. Method according to claim 1, in which the animation of said virtual model is obtained by setting the desired trend of the control variables of the actuators. Method according to claim 1, in which the animation of said virtual model is graphically induced to derive the trend of the control variables of the actuators which allow to carry out corresponding movements and / or deformations in the real system. Method according to claim 1, wherein the animation of said virtual model is obtained by a motion capture technique. 5. Method according to any one of the preceding claims, wherein the creation of said virtual model comprises the following steps: projection of said passive subsystems in the virtual environment while maintaining unchanged proportions and mutual distances; - creation and localization of actuators, constraints and joints in their correct spatial positions. Method according to claim 5, wherein the step of creating and locating actuators, restraints and joints comprises the steps of: - creation of a base skeleton for each actuator, - creation of the volumetric properties of said base skeleton; - positioning in space of said base skeleton; definition of its handling and / or deformation properties; - definition of the influence that each actuator, constraint or joint has on the subsystems dependent on them. Method according to any one of the preceding claims, in which the extraction of the values and of the control variables is obtained with the getAttr command or equivalent. Method according to any one of the preceding claims, in which the values extracted from the control variables are saved in an ASCII file by means of the fwrite command or equivalent. Method according to claim 8, wherein the set of saved files constitutes a database, from which to draw files of instructions for specific movements and / or deformations with which to drive the motor drivers. Method according to any one of the preceding claims, wherein said articulated mechanical system and / or deformable to check is an artificial face. Method according to claim 10, wherein said passive subsystem consists of artificial leather. Method according to claim 11, wherein the movements and / or the deformations of said artificial skin are imparted through artificial muscles that connect it to a support structure substantially in the shape of a skull and in which the following operational phases are foreseen: - creation of a virtual model of said artificial face using advanced 3D graphics software; definition in said virtual model of the control variables necessary for the movement of said face artificial; - animation of said virtual model in the desired way for extract from it the time values of said control variables: - saving the extracted values; sending of the saved values to control drivers of said artificial muscles for their piloting. Method according to claim 12, wherein the creation of the virtual model of said artificial face comprises the following steps: - projection in the virtual environment of the physical model of the artificial skin of said face while maintaining the proportions unchanged; creation of the virtual model of each single muscle by locating it in its correct spatial positions together with the relative constraints and joints; 14. Method according to claim 13, in which the projection in the virtual environment of the physical model of the artificial skin is achieved by identifying a certain number of points on the surface of the skin obtained by tracing meridians and parallels on it, detecting the spatial position of said points by means of a feeler device, bringing the detected values back into the virtual environment, recreating parallel lines that interpolate said points and interpolating said lines to create a surface. Method according to any one of claims 13 and 14, wherein the creation of a model of the artificial muscles comprises the creation of a bone to which volumetric properties are assigned, the duplication of it in a plurality of clones which are arranged, suitably scaled , in the positions corresponding to each artificial muscle, the creation of a control bone for each semifacial and of a general control parent bone for the two control bones, the creation of resident bones on the median axis of the face and one of their parent bones control also enslaved to said general control parent bone. 16. Method according to claim 15, in which to define the individual deformation properties, a flexor is associated with each actuator recreated in the virtual environment, the influence of the latter on the interconnected structures being highlighted and modified by means of a paint tool panel. 17. Equipment for the control of an articulated and / or deformable mechanical system comprising at least one passive subsystem to be moved and / or deformed by means of actuators connected to each other by joints and constraints, characterized in that it comprises computer means on which a software is implemented advanced 3D graphics for the creation of a virtual model of said mechanical system, suitable for generating instruction files for driving said actuators, motor means being provided for controlling said actuators and a control interface suitable for receiving said files of instructions and to process them to drive said motor means. 18. Apparatus according to claim 17, wherein said control interface comprises programmable microprocessor means connected to said computer means and controlling drivers for said motor means. 19. Apparatus according to any one of claims 17 and 18, wherein said articulated and / or deformable mechanical system is an artificial face comprising a support structure and said passive subsystem to be moved and / or deformed is an artificial skin of said face, said actuators being artificial muscles formed by an elongated ellipsoidal body made of flexible material with a rigid longitudinal core fixed to the end of said body which is connected to said skin and sliding with respect to the other end of said body which is fixed with respect to said structure of support, said core being connected to said motor means. 20. Apparatus according to any one of claims 17 to 19, wherein said motor means consist of linear motors each capable of transmitting a sliding of predetermined amplitude to the respective rigid cores of said artificial muscles. 21. Method for controlling an articulated and / or deformable mechanical system and related equipment substantially as described above and illustrated with reference to the attached drawings