JPS60161539A - Sensation-of-pressure discrimination control apparatus - Google Patents

Abstract

PURPOSE:To control the sensation-of-pressure discrimination of a robot near to human sensation, by detecting two-dimensional distribution of component forces in three directions by a sensation-of-pressure sensor array having predetermined constitution and applying required processing thereto by a microcomputer to perform control. CONSTITUTION:The fine sensation-of-pressure sensor module 21 of a sensation- of pressure sensor array 20 is constituted of a pressure receiving plate 23 and a sensation-of-pressure sensor 25 comprising a monocrystalline silicon ring and said sensor 25 is constituted of a plarality of diffusion type strain gauges 27. By this constitution, pressure component forces in three directions are detected with high accuracy without generating mutual interference and histeresis by the high density arrangement type small sensor 21 and the two-dimensional distribution of pressure component forces in three directions is measured through a scanner amplifier 33. The measured value is operated on the basis of predetermined algorithm through a microcomputer and converted to fundamental quantity data required in the discrimination of sensation of pressure and the sensation-of-pressure discrimination of a robot near to human sensation is controlled by feed-back control.

Description

[Detailed description of the invention] [Technical field to which /i Akira belongs] This invention is a method for attaching a pressure sensor to the /\end of an intelligent robot, the manipulator of an automatic assembly machine, or the sole of the foot of a mobile robot. The present invention relates to a pressure recognition control device that recognizes sensations corresponding to human sensations such as touch, pressure, slippage, and hardness, and controls robot hands, manipulators, or feet based on this recognition.

[Prior art and its problems] The force applied to the hands or feet of a robot, etc. that performs work in place of a human being is three-dimensional and has a certain distribution. The magnitude, direction, and distribution of this applied force on the pressure-receiving surface change as the hand (manipulator) or foot moves. Humans have a sense for this force, that is, not only a sense of pressure, but also a sense of touch, a sense of slippage, a sense of hardness, etc. Therefore, in order to control a machine's hands or legs as highly as humans, it is necessary for the machine to have these senses that humans have. Among these, the most basic sense required is three-dimensional pressure sensation and recognition of its distribution state.

Here, three-dimensional pressure sensation is the feeling of force in three directions in total, including not only the force perpendicular to the pressure-receiving surface but also the force in two directions on the pressure-receiving surface, and the moment in three directions associated with the movement of the hand or foot. It is desirable to have a sense of Sensors shown in FIGS. 1 to 3 have been announced as having a sense of detecting component forces in three directions. These conventional sensors include an elastic support 1 such as an elastic ring body, a cross-shaped leaf spring, or an elastic block, and a number of strain gauges 3 attached in each direction of the support.
It consists of, but is equivalent to the object grasped by node,
or a larger size (e.g. each stroke length is 100
~200 mm), so it is not used as a pressure-receiving surface of the hand or foot, but is mostly attached to the wrist 5 or ankle, and is used to detect the overall force acting on the hand or foot. It's just that. The end of that,
These conventional sensors cannot detect tactile sensations, slip sensations, and hardness sensations; if such sensory information is required, it is necessary to separately provide a special sensor for that purpose on the pressure-receiving surface. However, since those special sensors are also quite large, 1. Due to differences in pressure-receiving space and sensory position, it was possible to obtain only rough sensory information in an area corresponding to the palm of the hand, and it was not possible to obtain subtle sensory information in the area corresponding to the fingers.

Furthermore, as mentioned above, one of the characteristics of human senses is the ability to recognize pressure sensations distributed in a planar manner. The pressure-receiving surface of a robot or the like has a certain size, and when the object to be grasped is not flat or the hand or foot moves in a certain way, the force acting on the pressure-receiving surface is not uniform. Therefore, it is necessary to recognize the pressure sensation in the form of a distributed JAM, and through this recognition, it is possible to have a sense of the pressure-receiving surface, the center of action of the force, and the change in m1 at that time. Sensors shown in FIGS. 4 to 6 have been published as sensors that have a sense that allows the distribution of applied force to be recognized. First, Figure 4 (A)
The sensor shown in FIG. 1 uses a silicone rubber cord 15 made of conductive rubber and a metal electrode 17, and utilizes the fact that the contact area of the rubber cord changes due to pressurizing force, and the resistance value changes.

The sensor shown in No. 411(B) has silicone rubber cords 15 and metal electrodes 17 arranged in a grid pattern, and detects by applying a scanning method used in ITV (industrial television) and the like. Furthermore, the sensor shown in FIG. 5 uses a differential coil (not shown) or a Hall element (not shown) to detect the vertical movement of the tips of many thin rods (pins) 18 that slide vertically in response to the shape of the object being tested. By detecting this, the shape of a three-dimensional part (three-dimensional part) is recognized.

The devices shown in FIGS. 6(A) to 6(D) use conductive rubber, etc., which can be applied as a rotary force or a sensor. The example shown in FIG. 8(A) is a combination of a conductive rubber sheet 7 and a metal electrode 8, in which insulating electrode supports 9a are distributed on the substrate 8 and metal electrodes are attached to the tops of each. ing. Elastic materials such as sponge combs are distributed between the electrode supports 8, and a flexible conductive rubber sheet is adhered or placed on top of the elastic materials. As can be easily seen, the elastic material is compressed downward in the figure at the location where the load F is applied, and the conductive rubber sheet 7
By detecting electrical contact between the electrode 8 and the electrode 8, the position where the load F is applied or its distribution can be known. In FIG. 8(B'), a large number of electrodes 8 are arranged under a conductive or pressure-sensitive rubber sheet 7 formed of an elastic sheet and are distributed dotted on the surface of a substrate 8. , load F
By detecting the electrode 8 that is in electrical contact with the conductive rubber sheet 7 at the location where the load is applied, the location where the load is applied or its distribution is detected.

In the example of FIG. 6(C), insulating rubber sheath)? Conductive rubber 7 is embedded in rows on the lower surface of a, and electrode strips 8 are attached in rows on the upper surface of an insulating substrate 9 opposite to this. Since the conductive rubber 7 and the electrode strips 8 come into electrical contact at the intersection points of the matrix subjected to the load F, the load distribution can be determined from the position distribution of the contact points. Figure 6 (D)
Very thin metal plate or foil made of helium copper etc.8
In the example where a is used as the Tg pole, the elasticity of this jQ8a itself or the elasticity of its insulating cover 8C is utilized. The base 9 to which this composite is attached is made of metal, for example, and a portion 9C of its peripheral surface serves as an electrode surface, and the metal foil electrode 8a and the base 8 are insulated by an insulating plate 8d. When subjected to a load F, the composite of the electrode 88 and the cover 8C bends and the electrode foil 8
Since & is in electrical contact with the electrode surface 9C on the base body side, if a large number of electrode foils 8 are arranged along the base body 8, the load distribution can be known.

However, all of these conventional sensors only have sense in a direction perpendicular to the pressure receiving surface of the sensor.

Furthermore, in order to detect frost distribution 7j, it is necessary to make each module of the sensor small, but due to structural limitations, it is not possible to make it sufficiently small, and the detection mechanism of the mainly used conductive goose Because of this, there is non-linearity of the detection output with respect to the applied force and a narrow dynamic range, so it is only used as a tactile sensor for detecting contact or pressure-receiving area, and is a pressure sensor that can replace humans. It was insufficient. Therefore, with the conventional sensor device, it was not possible to obtain pressure recognition control similar to the human sense.

The performance required of the load distribution detector targeted by the present invention is as follows.

(a) The load detection unit must be small, with dimensions of several millimeters or less, and the units must be integrated in one distributed load detector with as much redundancy as possible.

(b) The component forces of the load can be well separated and detected without mutual interference.

(C) The relationship between load and detection output is linear, there is no hysteresis error in measurement, and the dynamic range, that is, the measurement range is wide.

(d) The load distribution detector itself has high rigidity and will not deform when subjected to a load and will not change the load distribution.

[Object of the Invention] In view of the above-mentioned requirements and the above-mentioned problems of the prior art, the present invention provides an array of minute pressure sensor modules capable of detecting force in three directions with high activation speed. By attaching an array of pressure sensors to a pressure-receiving surface such as a robot's hand or the sole of a foot, it is possible to directly detect three-dimensional pressure sensation and the state of distribution of that pressure. By processing the signal group using a predetermined algorithm, we can obtain not only the force components in the three directions, but also the moment of the force components in the three directions, the center of action of the force, the total force of action, etc., as well as the sense of contact, slippage, and hardness. An algorithm that enables advanced control of the functions of the robot's hands and legs by simultaneously recognizing these sensory information at high speed. is used to grasp the object with the minimum gripping force that does not deform the object, and to lift the object without impact.
It is an object of the present invention to provide a pressure recognition control device that can perform advanced motion control equivalent to the motions of human hands and feet, such as inserting or rotating an object to be grasped into a mounting hole.

[Key Points of the Invention] This invention creates a minute pressure sensor cell based on the principle of a semiconductor strain meter configured to be able to individually detect three-direction components of force applied to a pressure-receiving surface, and a large number of these cells. A pressure sensor array arranged in a single-sided array is attached to a pressure-receiving surface such as the hand of a robot, manipulator, or robot foot, and the output of the pressure sensor array is output on or near the substrate of the pressure sensor array. A scanner is installed to sequentially take out each row, and an analog switch for each cell that opens and closes depending on the signal from the scanner is installed in the pressure sensor, the mounting hole of the pressure sensor cell, or the board directly below the cell, and if necessary, the characteristics of a semiconductor strain meter. A correction circuit and an amplifier are provided to prevent confusion with the output of other pressure sensor cells, and the detection signals collected at appropriate time intervals are corrected for the characteristics of the strain meter and converted into force units, and then Organize pressure information distributed two-dimensionally for each component of force, store it in a signal processing storage device in a bell series, and then combine the three components of force using the pressure data read out from the signal processing storage device, total pressure, pressure The center, pressure distribution map, pressure-receiving area, moment, etc. are calculated and stored in the primary arithmetic storage device, and the contact sense, sliding sense, and shape and hardness sense of the object to be touched are necessary for controlling the robot's hands and feet. etc., the necessary calculation results are stored in the secondary calculation storage device, and the data stored in the signal processing storage device, the primary and secondary calculation storage devices, and the algorithm for the basic movement of the robot are used. A predetermined calculation is performed, the necessary calculation result is stored in the tertiary calculation storage device, and the signal obtained by calculation is sent to the external robot drive control device, and the pressure sense recognition from the external robot drive control device G is performed. It is designed so that it can also receive the commands necessary to carry out the operations.

[Embodiments of the invention] Hereinafter, the present invention will be described in detail with reference to the drawings.

FIG. 7 shows the main part of an embodiment of the invention. Here, 20
2 shows the entire piezoelectric sensor array, and 21 is a pressure sensor module as a unit micro module constituting the pressure sensor array 20. The pressure sensor module 21 consists of an upper pressure receiving plate 23 that receives applied force and one or two (or three) pressure sensor cells 25 fixed under the pressure receiving plate.

The pressure sensor cell 25 has a ring-shaped pressure sensitive structure made of single crystal silicon, and has a plurality of diffusion shapes on the pressure receiving surface of the pressure structure, that is, the surface perpendicular to the surface in contact with the pressure receiving plate 23. A strain gauge 27 is formed, and three components Fz, Fx, and Fy of the force applied to the pressure receiving surface are detected by changes in the resistance values of these strain gauges. A large number of the above-described pressure sensor modules 21 are arranged and fixed at high density in a planar array (matrix) on the lower common substrate 28,
Pressure 1: sensor array 20 is formed. At this time, if the lower end of the sensor cell 25 is vertically fitted and fixed in the parallel groove 31 formed in the substrate 28'' or in the mounting hole (not shown) drilled for each cell, reliable assembly can be achieved. This is preferable because it allows The output signals of the pressure sensor module 21 force 1, etc. are autonomously scanned 111 by a scanner amplifier (integrated circuit) 33 mounted on or in the vicinity of the substrate 28, and are amplified by the CPU (central processing unit) of the microcomputer, which will be described later. Processing section) This facilitates the formation of barrel rings.

The diffusion type strain gauge 27 of the pressure sensor cell 25 is arranged at a location that is highly sensitive and is theoretically unaffected by the other two forces. For example, four strain gauges that detect the force component Fz perpendicular to the pressure-receiving surface are formed near the left and right outer and inner edges on the center line parallel to the pressure-receiving surface, and one component Fx of the force parallel to the pressure-receiving surface is detected. Four strain gauges are formed near the outer edges of two lines tilted at an angle α0 in both left and right directions of the center line perpendicular to the parallel lines mentioned above, and other components FY of the force parallel to the received pressure (flo) are detected. Four strain gauges are formed at positions on the material mechanical neutral axis near the center of the two rings tilted at the angle α0 mentioned above.The angle α0 is such that only the force perpendicular to the pressure-receiving surface is applied. Sometimes, the angle is selected at a position that does not cause distortion, for example, in the case of a circular ring, it is selected around 38.6 degrees.In addition, these strain gauge groups are connected by bridges for each force component, respectively. Electrical signals Ez, E according to force components
Output x and Ey. Here, the strain gauge forming surface for the diffusion type strain gauge 271y 114515 can be formed not only on one side of the pressure sensor cell 25 but also on both sides, and if the strain gauge is formed on both sides, the wiring density will be lower.
This has the advantage that crosses between wiring lines can be reduced. Furthermore 1. Although strain gauges can be formed on the outer and inner circumferences of the ring, it is obvious that manufacturing is easier if they are formed only on the flat surface.

Moreover, the pressure sensor cell 25 is made of single crystal silicon,
Since the diffusion type strainer 27 is formed on the silicon, it can be easily manufactured using general conventional planar technology (integrated circuit manufacturing technology), and is extremely small (e.g., several mtn to 1 mm) with well-matched characteristics. diameter). For example, a predetermined thickness (for example, 0.8+pm)
A pressure sensor cell made of single-crystal silicon Ueno\, which has a predetermined conductivity type (e.g., N-type), a specific resistance (e.g., 1 to 10Ω·cm), and a predetermined crystal orientation (e.g., (ill)). Diffused strain gauge groups and metal wiring were fabricated in corresponding areas using maskless ion beam processing and AfL7. After forming using a planar technique such as % bonding, the pressure sensor cell can be precisely cut out using a wire soaker/atto method, laser processing, or machining such as etch cutting. Therefore, the pressure sensor module 2 as shown in FIG.
For example, 1i1 can be integrated on the substrate 28 at a high density of about 25 to 100 pieces per 1 cm2.

As described above, the pressure sensor array 20 is formed by arranging a large number of pressure sensor cells 25 or pressure sensor modules 21 in a planar array at high density to independently detect the three components of the force applied to the pressure receiving surface. The applied force can be decomposed into force components in three directions, and the two-dimensional distribution of the force components in the three directions can be detected independently and with high density. Furthermore, at this time, the diffusion type strain cage 27 of the pressure sensor cell 25 is placed in a location where it is highly sensitive and is theoretically unaffected by the forces in the other two directions, so that the pressure applied to the pressure-receiving surface is The force can be well separated and detected without mutual interference, and since the excellent single-crystal silicon is used as the pressure-sensitive structure, the linearity between the applied pressure and the detected output is good, making it easy to measure. It has the advantage of not having the above hysteresis, has a wide dynamic range, and can be detected as a pair of tension and compression by combining the horizontal pressure sensor cells 25.

In this way, the pressure sensor array 20, which has minimized cell dimensions and is highly integrated, is placed on a predetermined pressure-receiving surface, such as the surface of the palm, fingers, or foot of a robot's hand, or the tip of a manipulator of an automatic assembly machine. Attach it to the surface. Note that the shape of the pressure sensor cell 25 is not limited to that shown in FIG. 7, and various shapes are possible. For example, in the case of using a group of connected pressure sensor cells in the form of a strip in which a portion of the lower or upper part of each cell is integrally connected for each column, there is an advantage that assembly such as positioning and fixing becomes easy.

Moreover, the semiconductor strain gauge 2 in the pressure sensor cell 25
7 is output to the outside through the wiring on the substrate 28, but since the number of wirings from the high-density pressure sensor array 20 is enormous, the wiring inside the pressure sensor cell 25 and the joint between the pressure sensor cell 25 and the substrate 29 are Each device has its own way of wiring and connections. For example, the strain gauges 27 may be distributed on both sides of the pressure sensor cell 25 for each detection force, or an epitaxial layer having an opposite conductivity type may be formed on the single crystal silicon substrate, and a strain gauge may be formed in the epitaxial layer. By forming a diffusion layer penetrating the epitaxial layer, the diffusion layer and the substrate section can be configured as part of the power supply wiring of the strain gauge 27, or alternatively, the substrate section 28 can be multilayered to connect the power supply line and the signal line between each layer. The power supply line is connected to the pressure sensor cell 25 from the groove 31.

FIG. 8 shows an entire embodiment of the invention. As shown in the figure, electrical signals El, EY, and Ez corresponding to the magnitude of three component forces output from the bridge circuit 21A at each point of the matrix-wired pressure sensor array 20 are sequentially applied to each row and column by a scanner 33A. Read out at high speed. At that time, an analog switch (not shown) that opens and closes in response to the row and column address signals (control voltages) of the scanner 33A is installed in the pressure sensor cell 25 or in the mounting hole of the pressure sensor cell or on the board 28 in the bridge circuit 21A of the pressure sensor module 21. By selectively operating those analog switches, other pressure sensor modules 21
Not to be confused with the output of If power supply control is performed, power consumption will be reduced, which is even better.

The pressure number 4g read by the scanner 33A is amplified by the amplifier 33B, converted into a digital signal by the analog-digital (AD) converter 41, and then taken into the CPIJ (p-cpu) 43 of the microcomputer and stored in the ROM ( A predetermined calculation formula stored in the read-only memory 45 corrects the temperature characteristics of the strain gauge 27, non-linearity of sensitivity, variations in sensitivity for each cell 25, and interference of component forces in other directions. The converted pressure data is organized in the form of pressure sense data distributed two-dimensionally for each force component and stored in the signal processing storage device 47 in time series. The above-mentioned temperature compensation is determined by calculation using a detection signal of a temperature sensor such as a thermistor provided near the pressure sensor cell 25.

Next, using the pressure data read out from the signal processing storage device 47 and the basic calculation algorithm stored in advance in the ROM 45, the g. , composite total pressure (the sum of the pressures applied to all cells 25, which corresponds to the total acting force), center of pressure (the center of gravity point where the sum of the moments of force at that point is zero, and the center of force action) Hit), 3
Calculate the directional moment, pressure receiving area (calculated from the total number of non-zero pressure sense data and the arrangement pitch of the cells 25), etc.
These calculated basic data are stored in the primary calculation storage device 48 in chronological order.

Furthermore, using the data stored in the signal processing storage device 47 and the primary calculation storage device 48 and the basic calculation algorithm stored in advance in the ROM 45, p.c.
The PU 43 calculates the sense of contact, the sense of sliding, the shape of the object to be touched, the sense of hardness, etc. necessary for controlling the limbs of the robot, and the necessary calculation results are stored in the secondary calculation storage device 51. The sense of touch (tactile sensation) is determined based on a signal that exceeds a certain predetermined threshold. The slider stores pressure data in storage devices 47 and 48.
Since the information is stored in time series, slippage due to insufficient gripping force is calculated based on temporal changes in the pressure distribution on the pressure-receiving surface, and R
The gripping control calculation algorithm (grasping algorithm) stored in the OM 45 allows soft handling control to be performed on the hand drive mechanism 53 to prevent slipping. Further, recognition of the shape of the object to be touched can be obtained by calculation based on temporal changes in pressure distribution and a typical shape recognition algorithm. Furthermore, by appropriately selecting the material distribution of the pressure-receiving plate 23, the elasticity of the object can be determined by calculation, and the material quality and hardness can be determined.

On the other hand, the host computer 5 via the interface 57
The data stored in the signal processing storage device 47, the primary and secondary calculation storage devices 48 and 51, and the RO
Using the robot's basic motion algorithm (work calculation algorithm) stored in advance in M47,
The CPU 43 allows the robot to control basic motions such as smooth running, grasping, lifting, inserting, rotating, etc. at high speed (
or control position), and sends the calculation results to the host computer 59 as an external robot control device via the interface 57, and also sends the necessary calculation results to the
The next calculation is stored in the storage device 55. Upper computer 59
Then, the hand drive mechanism 53 is smoothly driven based on the calculated control amount or control position data.

In this way, supervisory control and autonomous control are performed at high speed, with high response, and with high precision based on the control commands from the host computer 58, and the gripping force is maintained at a minimum level to the extent that the object to be gripped (for example, a fruit) is not deformed. To have a robot hand, etc. perform advanced soft handling operations comparable to those of a human hand, such as gripping with a hand, lifting and releasing a gripped object without impact, inserting the gripped object into a mounting hole, and rotating it. I can do it. In addition, feedback signals containing information such as grasping, lifting, releasing, inserting, turning, etc. are output from the hand drive mechanism 53 to the host computer 58, but pressure recognition cannot be performed based on these feedback signals. The command signals necessary for
43. . Moreover, the load on the external host computer 5θ is significantly reduced by the various calculation operations described above by the p,* CPU 43, speeding up control operations, and facilitating smooth control. Note that each of the storage devices 47.4θ, 51, and 55 used for arithmetic control may be provided independently, but it is also possible to overlap and share them.

Furthermore, although scanner 33A and amplifier 33B are typically provided on or near sensor array 20, A[
l converter 41. The microprocessor 61, such as the CPU 43 and the interface 57, is an LSI of several chips.
(Large-scale integrated circuit), and can be installed near the hand drive mechanism 53 or stored in the armpit of the hand.

Next, the contents of the above-mentioned main calculation algorithm used in the apparatus of the present invention will be explained.

■ Three-direction moment As shown in FIG. 9, the total output X, Y of the three component forces of the pressure sensor arrays 20 and 20 on both sides of the object 7I.

Z, X', Y', Z', the distance z between both pressure sensor arrays, and the Y direction output Y when both pressure sensor arrays 20 and 20 are placed in the same plane position.

(Y) and the distance fIIEi x between both arrays,
Each moment in three directions using the following formula) Mx, My, Mz
can be used.

Mx= (Y, Y' (7) difference)XMzMY= (X,
Difference in X')X Sentence 2 Mz= (Difference in Y, (Y))
This is a mechanical value that is essential for tasks such as grasping unevenly loaded objects and screwing in objects.

■ Sliding Award As shown in Fig. 10, when slipping occurs due to lack of grasping force, etc., the distribution of force on each module 21 of the sensor array 2o changes with time to, to +Δt, etc., so the time series Slippage can be determined by calculating the change. Therefore, regardless of the algorithm for grasping and lifting objects by soft handling, which will be described later, slip prevention 1
L to j can be controlled at high speed.

(2) Shape Recognition As shown in FIG. 11(A), when the object 71 is flat, the pressure distribution 73 on the pressure receiving surface does not change as the gripping force ΣP increases (pg + P2). However, as shown in FIG. 11(B), when the object 71 has a cylindrical surface, the gripping force ΣP increases (
PO+P2), the pressure distribution 73 on the pressure receiving surface changes only in one direction. Moreover, as shown in FIG. 11(C),
When the object 71 is spherical, the gripping force ΣP increases (Pa→P2
), the pressure distribution 73 on the pressure receiving surface changes in all directions. Therefore, the shape can be recognized by the difference in the temporal change of the pressure receiving surface. In addition, to in the figure is the initial time, 1. ．．
12 indicates increased time.

■ Determination of elasticity or material @12 As shown in Figure 12, different elastic pads 75 (each compliance is C, C2) are placed on the upper pressure receiving plate 23 of the pressure sensor array 20, and the load received by each sensor is h. , P2, the compliance C8 of the object 71 to be grasped is P2-p.

It is calculated using the formula.

However, this is a relationship between spring constant and force, and if this compliance c0 is used as data for an algorithm for object sorting or soft hand linking, object sorting and soft handling become possible.

In this way, by determining the elasticity of the object being grasped and indirectly estimating the material, it is possible to control the minimum grasping force to avoid deformation or damage of the object being held, that is, soft handling. Although it may not always be possible to select a suitable selection, it can be used as auxiliary data for the selection.

(2) Grasping and Lifting an Object As shown in FIG. 13(A), the force Q to lift the object 71 is determined in the form of a composite force of the component forces Fx and Fy at each sensor array 2o. After the object is lifted up, Q=W
becomes. Furthermore, as shown in FIGS. 13(B) and (C), the grasping force R is expressed in the form of Fz. That is, object 7
In order to prevent slippage between 1 and hand 77, Q≦pR=
gFz (g friction coefficient) must be satisfied. Now, suppose that after grasping the object 71 with the hand 77, the hand is lifted by Δδ. When there is no slippage in the hand, Q increases at a rate of Δδ ΔQ=−. (Here C is the total compliance of the object and floor surface). In this case, in order to perform soft handling, it is necessary to increase the grasping force of the hand so that ΔQ ΔR≧□ increases with respect to the increase in Q. On the other hand, if slippage is detected using the method described in the previous section (2), R is increased (regardless of this equation) and a new value p that does not cause slippage is defined. Change tiles to increase R (Q is known). When Q=W, the object 71 moves away from the floor 78 and is grasped by the hand. Thereafter, there is no increase in Q and R for an increase in δ.

FIG. 14 shows these relationships in a flowchart.

In addition, soft handling when placing or releasing an object is possible by performing the reverse operation of the above case, and the object can be released without impact. At this time, soft handling can be made easier by detecting the proximity of the object 71 using another proximity sensor to reduce Δδ or by increasing the compliance of the hand.

■ Insert the object 71 into the insertion hole 81. In addition, fitting together is the basis of assembly work. FIG. 15(A) shows the basic form of this operation. FIGS. 15(B) and 15(C) each show cases where the insertion is incomplete, and a certain moment LM is generated against the insertion force P. By determining the magnitude and direction of this moment H, insert the insert 7 so that the ideal normal insertion shown in FIG. 15(A), that is, the moment M becomes O.
Perform an arithmetic operation to move 1.

■ Application as a walking control sensor Up to Section 18, we have described the case where the sensor device according to the present invention is applied to a robot hand, but as shown in Fig. 16, this sensor can also be used for walking control as a sensory sensor for the robot's legs. Applicable.

(a) Detects the movement of the body's center of gravity while walking and raises the leg at the optimal time.

Movement of the center of gravity □ Time-series calculation of the center of gravity position from the surface pressure distribution of the feet Raising and lowering the feet - Calculation basically similar to lifting the object shown in Figures 13 and 14 (b) Feet for forward walking It detects the blood pressure and the feeling of slipping when kicking, and applies a kicking force (forward force) commensurate with that.

However, in the figure, Fz indicates surface pressure and Fx indicates kicking force.

[Effects of the Invention] As explained above, according to the present invention, a pressure sensor array capable of detecting a two-dimensional distribution of component forces in three directions is provided, and after sequentially extracting the sensor output and performing necessary signal processing, The stored data is converted into various basic quantities related to pressure using a predetermined algorithm and stored.
These stored data are used to convert into various basic control signals necessary for controlling a controlled object such as a robot, and these basic signals are sent to an external control device such as a robot. It is possible to recognize sensations corresponding to pressure sensation, sliding sensation, hardness sensation, etc., and to control the hands of intelligent robots, manipulators of automatic assembly machines, legs of mobile robots, etc. smoothly and safely based on this recognition. In particular, it is possible to easily perform sophisticated and high-viscosity operation control such as soft handling.

[Brief explanation of the drawing]

Figure 1, Figure 2, Figure 3 (A), (B), Figure 4 (A
). (B), Figure 5, and Figure 8 (A), CB),
(, C) and (D) are respectively perspective views showing conventional sensors, and FIG. 7 is a perspective view showing main parts of an embodiment of the present invention.
FIG. 8 is a block diagram showing the entire embodiment of the present invention, and FIGS. 8 to 16 are explanatory diagrams showing the contents of various algorithms used in the present invention, respectively. 1... Elastic support body, 3... Strain gauge, 5... Wrist, 7... Conductive rubber sheet, 7a... Insulating rubber sheet, 8... Metal electrode, 8a... Metal foil electrode, 8c... Insulating cover, 8d... Insulating plate, 8... Substrate. 8a... Electrode support, 8c... Electrode surface, 15... Silicon rubber cord, 27... Metal electrode, 18... Thin rod, 20... Pressure sensor array, 21... Pressure sensor Module, 21A... Pressure sensor module bridge circuit, 23... Pressure receiving plate, 25... Pressure sensor cell, 27... Diffused strain gauge, 29... Substrate, 33... Scanner amplifier, 33A ...Scanner, 33B...Amplifier, 41...Analog-digital converter, 43...
・Microcomputer central processing unit (CPU), 45...ROM (read only memory), 47...
Signal processing storage device, 49... Primary calculation storage device, 51... Secondary calculation storage device, 53... Hand drive mechanism, 55... Tertiary calculation storage device, 57... Interface, 58 ... Upper computer, 71 ... Object to be grasped, 73 ... Pressure distribution, 75 ... Pad, 77 ... Hand, 78 ... Floor surface, 81 ... Hole. Patent applicant Fuji Electric Research Institute Co., Ltd. Figure 2 1 3 Figure 3 Figure 4 (A) (B) Figure 5 Figure 10 Senna position Figure 11 (A) 1 ΣP=Mo ΣP=Fi ΣP = P2t=to t=
tpit, j”t□+j2ΣP=Po ΣP=P, Σ
P: P2 ward・゛ ward 3 綜綜

Claims (1)

[Claims] 1) 3 of the force applied to the pressure-receiving surface attached to the drive control target
Pressure detection means for detecting a two-dimensional distribution of directional force; a detection signal is sequentially extracted from the pressure detection means and subjected to predetermined signal processing to be converted into pressure detection data in units of pressure; A signal processing storage means for storing each directional force corresponding to the pressure receiving surface; a basic amount of pressure necessary for predetermined pressure recognition based on the pressure detection data read from the signal processing storage means and a predetermined basic calculation algorithm; primary calculation storage means for calculating data and storing the calculated basic quantity data; the basic quantity data read from the primary calculation storage means; the pressure sense detection data read from the signal processing storage means; A secondary calculation storage means for calculating predetermined pressure 9: recognition data using a pressure recognition calculation algorithm and storing the calculated pressure recognition data; and a secondary calculation storage means for storing the calculated pressure recognition data; Accordingly, the pressure sensation recognition data read from the iffff-like calculation storage means, the basic quantity data read from the primary calculation storage means, the pressure sense detection data read from the signal processing storage means, and the predetermined control (
tertiary calculation means for calculating control amount data necessary for drive control of the drive control target using a JIl amount calculation calculation algorithm, and outputting the calculated control amount data to the external control means; the tertiary calculation means and the A pressure recognition control wave Δ characterized by comprising a data communication means for exchanging data between external control means, and an internal control means for controlling the entire processing operation to be performed smoothly. 2. In the device according to claim 1, the drive control target is a robot's limbs or a manipulator of an automatic assembly machine, and the basic quantity data is a composite total pressure of three-directional force components, a pressure center, and a pressure distribution. The pressure recognition data is data regarding contact sense, sliding sense, hardness sense, shape recognition, and material recognition, and the control measure data is data regarding gripping, lifting, insertion, and rotation. Pressure recognition control device, characterized in that the instruction data is data related to grasping, lifting, inserting, turning, releasing, and walking. 3) In the device according to claim 1 or 2, the pressure sensing means converts the force applied to the pressure-receiving surface into a component force Fz in a direction perpendicular to the pressure-receiving surface and a component force Fz in a direction perpendicular to the pressure-receiving surface. A large number of pressure sensor modules are arranged in rows and columns at high density on a substrate to form an array. A pressure-sensing recognition control device characterized by being a pressure-sensing sensor array. 4) In the device according to any one of claims 1 to 3, the signal processing storage means generates and outputs an electric signal corresponding to the magnitude of three component forces from the pressure sensing means. Each column is sequentially extracted and amplified, analog-to-digital conversion is performed, temperature compensation, sensor sensitivity non-linearity compensation, and interference from component forces in other directions are compensated for, and the corrected signals are converted into the pressure sensing data. A pressure recognition control device characterized by storing information in chronological order. 5) In the device according to any one of claims 1 to 4, the primary calculation storage means calculates a composite total pressure, a pressure A pressure sense recognition control device characterized in that a center, a pressure distribution diagram, a pressure receiving area, and a three-direction moment are calculated by σ and stored in time series. (Margin below)