JPH0647687A - Position control device in force control robot - Google Patents

Abstract

PURPOSE:To reduce the correcting number when the first part is inserted to a specific position of the second part so as to reduce the time of the assembly work extensively by correcting a fuzzy rule with a data input newly every time of an actual assembly work. CONSTITUTION:When a set condition is satisfied, the first part 5 is inserted to a specific position of the second part 6. In this case, the data to make the detecting values of the reaction and the moment detected when the first position correcting value is inferred, and the sum of the position correcting amounts corrected during the repeating operations, in a set, is added to the collection of the basic data to produce a fuzzy rule beforehand, and the former fuzzy rule is reproduced by a rule producing member 9. As a result, the fuzzy rule is to be corrected by the data input newly every time of an actual assembly work.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a position control device for a force control robot that corrects a positional deviation between parts and assembles them with high accuracy in an automatic assembling work for fitting and assembling parts.

[0002]

2. Description of the Related Art In a production line for precision parts, for example, when a robot automatically performs an operation of inserting a pin into a cylindrical member and assembling the pin, the end of an arm of the robot grips the pin to be fitted. The basic operation is to move the cylindrical member toward the small hole, repeat the operation for alignment, and then insert the pin into a predetermined position of the small hole to fit the pin.

In such fitting work, a high accuracy is required for the relative position of the pin and the small hole, but there is a position error that cannot be ignored due to variations in manufacturing the above-mentioned parts. However, there is a problem that it is difficult to assemble the cylinder member with high reliability by the positioning operation by simply repeating the operation toward the small hole of the cylindrical member.

Therefore, RCC (remote center compliance) is adopted as a device that absorbs this position error by a mechanical function. However, this RCC
When the pin is inserted into the small hole, it does not detect the reaction force or moment when the pin comes into contact with the opening, so that there is a drawback that the contact state or contact force of parts cannot be detected. For this reason, there is a problem that the position error cannot be corrected with flexibility, and the parts contact each other with a strong force to damage the surface.

In view of such problems, the present inventors have previously proposed a position control device for a force control robot.

In this force control robot, as shown in FIG. 6, a force sensor 3 is attached to the tip of an arm 2 of the robot 1. The force sensor 3 detects a reaction force and a moment, and the pin 5 is connected via the force sensor 3.
A gripping part 4 for gripping is provided. This grip 4
Is for holding the upper end of the pin 5 and holding it vertically while moving it. The pin 5 is inserted into the small hole 6a from above the cylindrical member 6.

Further, the robot 1 moves the pin 5 or the cylindrical member 6 to move the force sensor 3 when the final target position where the pin 5 is fitted to the predetermined position of the small hole 6a is set as a reference position. A fuzzy rule is generated from a set of reaction force and moment data based on the detected values of 1 to estimate a position correction amount with respect to a reference position, and a control device 11 that determines the movement amount of the pin 5 based on this position correction amount It is equipped with.

The control device 11 stores a memory 12 that stores a set of data including a force applied to the grip portion 3 of the robot 1 and a position correction amount, and a set of the force and the position correction amount in an actual work. A central operation for adding a data to be stored to a pre-stored set of data, generating a fuzzy rule from this set of data, and estimating a position correction amount by fuzzy inference from the reaction force detected by the force sensor 3. Processing unit 1
3 and 3.

Then, the control device 11 clusters a set of force and position correction amount based on the detection value output from the force sensor 3 by the extracted force element and position correction amount element, After integrating into the defined number of groups, the characteristics of each group are replaced with membership functions. After that, the optimum rule is selected according to the membership function from the reaction force detected during the actual assembly work, the position correction amount is estimated, and the movement amount is determined.

Next, the assembling work by the robot will be described with reference to the flowchart of FIG.

In this assembling work, a cylindrical member having a small hole with a diameter of 10 mm and a depth of 30 mm is used as an object to be fitted.
A case where a pin having a diameter of 10 mm and a length of 40 mm is fitted as a fitting will be described.

First, the pin 5 is attached to the grip portion 4 of the robot 1.
Hold at. Then, the arm 2 is moved to a depth of 30.
A constant pressing force is applied by the force control mode so that the pin 5 is inserted into the small hole 6a of 10 mm by 10 mm (step S
31).

At this time, when the pin 5 contacts the opening of the small hole 6a or the inner peripheral surface, the reaction force is detected by the force sensor 3 (step S32).

Next, it is judged whether or not the reaction force, which is a force component in the direction perpendicular to the inner peripheral surface of the small hole 6a, is smaller than a preset condition ε, and the reaction force is larger than the set value ε. If there is, the position shift amount is estimated by fuzzy inference (step S3).
3).

Subsequently, the position of the grip portion 4 of the robot 1 is corrected (step S34), the process returns to step S32, and the force sensor 3 detects the reaction force.

The position correcting operation for detecting the reaction force and correcting the position again is repeated until the force applied to the grip portion 4 satisfies the setting condition ε.

If this condition that the reaction force is smaller than the preset value ε is satisfied, the small hole 6 is set in the force control mode.
The pin 5 is inserted to the bottom of a and the pin 5 is fitted to the cylindrical member 6 (step S35).

[0018]

By the way, the position control device in the above-mentioned conventional force control robot is generated based on a set of data including a set of reaction force and moment data and a position correction amount before assembly work. Since the fuzzy rule is used, there is an advantage that it is possible to deal with the above-mentioned position error correction with flexibility.

However, since this fuzzy rule is fixed based on a set of data including the reaction force and moment data and the position correction amount before the assembling work, the actual work is performed. In the case where a reaction force and a moment that do not conform to the above rule are detected in the above, the robot 1 repeatedly performs the position correcting operation of the gripping portion 4 until a preset condition is satisfied, and thus the working time becomes long. was there.

FIG. 9 shows the above-mentioned conventional position control device.
It is an experimental result when assembling the pin 5 and the cylindrical member 6, but when a reaction force and a moment which do not conform to the fuzzy rule are given, the number of trials of position correction is increased as shown in the figure. Despite the increase
The problem remains that the number of corrections tends to never decrease.

An object of the present invention is to solve the above problems.

[0022]

In order to achieve the above object, the present invention detects a reaction force and a moment acting on the gripping part by gripping the first part by the gripping part provided at the tip of the robot. In a position control device in a force control robot that performs position correction based on a detection value from a sensor, and inserts the first component into a predetermined position of the second component to assemble the first component, A reaction force and a moment detected by the sensor when the first component or the second component is moved in an arbitrary direction when a final assembly position for fitting the component into a predetermined position is set as a reference position, and A means for inputting a set of data, which is a set of a position correction amount corresponding to the movement amount, to the memory as basic data,
From the set of data, extract the elements including the reaction force and the moment that are correlated to the respective elements of the position correction amount,
Means for generating a fuzzy rule for estimating the position correction amount by integrating the extracted elements into a predetermined number of groups by clustering, and further replacing the characteristics of each group with a membership function; and the fuzzy A means for repeating the operation of estimating the position correction amount based on a rule and correcting the position of the grip portion until the detection value of the sensor satisfies a preset condition, and a means for detecting when first estimating the position correction amount. The fuzzy rule is generated by adding data that forms a set of the reaction force and the moment and the sum of the position correction amounts corrected during the repetitive operation at the time of position correction to the basic data set for generating a fuzzy rule in advance. And a control device including means for regenerating

[0023]

In the present invention, first, the first part is moved in an arbitrary direction, and the data of a set of the detected value of the reaction force and the moment detected by the sensor at this time and each value of the movement amount is set. Store the set as basic data in memory.

Next, from the data set, elements including a reaction force and a moment which are correlated with the respective elements of the position correction amount which is the direction opposite to the moving direction are extracted, and these extracted elements are extracted. By clustering, the number of groups is integrated into a predetermined number, the characteristics of each group are replaced with a membership function, and a fuzzy rule for estimating the position correction amount of the first component is generated.

After that, when the first part is actually inserted into the second part, the position correction amount is estimated based on the fuzzy rule from the detection value when the first part contacts the second part. Then, the position correction operation of moving the gripping portion that grips the first component to the predetermined position is repeated until a preset condition (for example, a condition in which the detection value of the sensor is 0) is satisfied.

Here, when the set condition is satisfied,
The first part is fitted into the predetermined position of the second part.

At this time, the fuzzy rule is preliminarily set to the fuzzy rule as data which is a combination of the detected value of the reaction force and the moment detected when estimating the initial position correction amount and the sum of the position correction amounts corrected during the repetitive operations. Is added to the set of basic data to regenerate the previous fuzzy rule.

As a result, the fuzzy rule is regenerated at the time when the first part is fitted into the predetermined position of the second part, so that the fuzzy rule is not fixed as in the prior art but is actually a fuzzy rule. It will be corrected by newly input data for each assembly work.

Then, each time the correction is performed, the position correction operation for moving the grip portion to a predetermined position converges and the number of times decreases, so that the number of position corrections decreases in proportion to the increase in the number of assembly operations, and the working time increases. Significantly shortened.

[0030]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is an overall configuration diagram of a position control device in a robot according to a first embodiment of the present invention.

The robot 1 has the same basic configuration as the conventional example shown in FIG.
Is equipped with. That is, the control device 7 stores a memory 8 storing a set of data including a reaction force and a moment applied to the grip portion 3 of the robot 1 and a position correction amount, and a reaction force and a moment and a position correction amount. A rule generation unit 9 for generating a fuzzy rule from data to be added, adding data obtained in actual work to a pre-stored data set, and re-generating a fuzzy rule from this data set, and a force sense Central processing unit 1 for estimating the position correction amount by fuzzy inference generated by the rule generation unit 9 from the reaction force and the moment detected by the sensor 3.
And 0.

As a method of giving data to the memory 8, an ideal assembly position of the pin 5 and the cylindrical member 6, that is, a final assembly position in which the pin 5 is fitted to a predetermined position of the small hole 6a is a reference position. Then, the pin 5 is moved in an arbitrary direction based on this reference position (the amount of movement in the direction opposite to this moving direction is the position correction amount).

At this time, when the pin 5 comes into contact with the opening of the small hole 6a, the force sensor 3 detects the reaction force and the moment, and the relative positional deviation amount from the reference position is obtained. The amount of positional deviation is a set of the detected value of the reaction force and the moment and the value of the movement amount, and a data set consisting of a large number of these sets is input to the memory 8.

Specifically, the data is obtained by detecting the force applied to the grip portion 4 of the robot 1 by the force sensor 3 when the pin 5 contacts the cylindrical member 6, and the 6 components of the force at this time are detected by the robot 1. The position correction amount and the six components of the force are stored in the memory 8 as a force applied to the tip. The six components are Fx: force in the x direction, Mx: moment in the x direction, Fy: force in the y direction, My: moment in the y direction, F.
z: force in the z direction, Mz: moment in the z direction,
In the fitting work like this example, Fx, Fy, Mx,
The four components of My are the main data.

Next, in the rule generator 9,
In order to find an element having the strongest influence on each position correction amount x, y from a set of various data of components, each element of the position correction amount in the direction opposite to the moving direction is set from the data set. Extract correlated reaction force and moment elements.

Table 1 is an example showing the correlation between the position correction amount x and the force Fx.

[0038]

[Table 1]

Subsequently, these sets are integrated into a predetermined number of groups by clustering, the characteristics of each group are replaced with membership functions, and a fuzzy rule for determining the movement amount of the pin 5 is set. To generate.

Table 2 is an example of group integration by clustering.

[0041]

[Table 2]

After this, the characteristics of each group are replaced with membership functions.

Table 3 is an example of what is replaced by the membership function and adopted in the IF-THEN rule described later.

[0044]

[Table 3]

Where C is the force Fx integrated into the group
, W is the width with respect to the center value.

To explain the above points in more detail,
First, by clustering as shown in Table 2 with the force element and the position correction amount element extracted as shown in Table 1, for example, the set of force and position correction amount is integrated into a predetermined number of groups. In this case, a force element having a correlation with each element of the position correction amount is extracted from the data set by multivariate analysis for finding out which of the four components the positions x and y have a correlation with. .

In this multivariate analysis, the central processing unit 10 obtains the variance for each force element, and if the value is below a certain value, it is assumed that it is irrelevant to all the elements of the position correction amount, and the fuzzy rule Specify not to add to the variable. After that, the partial correlation between each element of the position correction amount and each element of the force is obtained, and if the absolute value is equal to or larger than a predetermined value, it is taken as a candidate for the variable described in the fuzzy rule.

Therefore, the other candidates listed above are not added to the fuzzy rule variables. This is done in the same way for the other elements of the position correction amount, and only the elements of the forces having a correlation with each other are extracted.

As a result, the factors that have the strongest influence on each position correction amount, such as the fact that the position correction amount in the x direction is correlated with the force Fx acting in the x direction as shown in Table 1, are extracted.

Incidentally, although this Table 1 is considerably simplified, if this is also obtained in the y direction, the elements having a correlation with the total position correction amount can be grasped.

After that, in order to integrate these elements into groups, a distance function D of the following formula by the least square method is defined.

[0052]

[Equation 1]

By this operation, a pair (set) of i and j that minimizes D is found, and one group is formed. Therefore, if this calculation process is repeated until the number of groups reaches a predetermined number, for example, the position correction amount x
And the forces Fx are integrated into the group of the closest values.

Next, in order to generate a membership function from the characteristics of each group, the average value for each group is obtained. Then, when this average value is used as the representative value of each group, the center value C shown in Table 3 above is obtained.

Then, by determining the respective widths W from the respective central values C, the membership function as shown in FIG. 2 can be generated.

In this example, this membership function is an isosceles triangle represented by a center value C and a width W (length of the bottom surface).

Specifically, the average value of each element of the integrated group is set as the central value of the membership function, and the central value other than the maximum or the minimum for each element, for example, larger than A for A and A The difference between the center value closest to A and the center value smaller than A and closest to A is the width of the membership function whose center value is A. Further, when the maximum or minimum central value is, for example, B, the absolute value of the difference between B and the central value closest to B is twice the absolute value of the membership function of which the central value is B. It is the width.

For example, when the data of the central value C is the following six types, 2.0, 3.0, 5.0, 1.0, -2.0, 8.0 First, this is rearranged in order. And −2.0,1.0,2.0,3.0,5.0,8.0 Then, the width W is obtained from this value by the above method.

Here, for each of the above elements, a central value other than the maximum value (8.0) or the minimum value (-2.0), for example, assuming that A is 3.0, a value larger than 3.0. (5.0, 8.0) and the center value 5.0 which is the closest to 3.0 and a value smaller than 3.0 (-2.0, 1.
The central value (2. 0, 2.0) and closest to 3.0 (2.
The difference (5.0-2.0) with respect to 0), that is, 3.0, is defined as the width of the membership function whose center value A is 3.0.

When the maximum or minimum center value B is 8.0, for example, the difference between 8.0 and the center value (5.0) closest to 8.0 is 8.0. (8.0-
5.0), that is, twice the absolute value of 3.0 (3.0 × 2), that is, 6.0, is defined as the width W of the membership function whose center value is B. As a result, the following results are obtained.

6.0, 4.0, 2.0, 3.0, 5.0, 6.0 When the center value C and the width W are calculated as described above,
A fuzzy rule consisting of multiple membership functions such as is generated. This is similarly formed for the other elements of the position correction amount.

As described above, the fuzzy rule of this embodiment is generated by a plurality of membership functions,
The interval between the minimum value and the maximum value becomes large, and a fuzzy rule having a large overall width W can be obtained. Then, since the arrangement of the membership function is small, when the reaction force and the moment which do not conform to the fuzzy rule are given, the problem that the processing becomes impossible almost disappears.

By the way, the fuzzy rule is an IF-THEN rule in which the element of force having a correlation with the element of the position correction amount is the antecedent part, and the element of the position correction amount is the consequent part. Use multiple rules as follows:

If "Fx is large" and "Mx is large", move a little in the x direction. (1) If "Fx is small" and "Mx is medium", move largely in the x direction. (2) By adopting this IF-THEN as a rule, the position control device has the force sensor 3 attached to the tip of the robot 1.
The position of the hand of the robot 1 is corrected by estimating the amount of position correction from the detected value, and the position correction operation is repeated until the force applied to the tip of the robot 1 satisfies a preset condition, thereby achieving highly accurate positioning. It is a thing. As the setting condition, a value close to 0 which is a state in which the force sensor 3 does not detect the reaction force is adopted.

Further, in this example, the reaction force and the moment detected when the initial position correction amount is estimated and the sum of the position correction amounts corrected during the repetitive operation during the position correction are combined. By adding data to the set of data for which fuzzy rules are generated in advance and regenerating the fuzzy rules, it is possible to reduce the number of position corrections repeated until the preset condition is satisfied.

Next, the assembling work by the robot will be described with reference to the flowchart of FIG.

In this assembling work, a cylindrical member 6 having a small hole 6a with a diameter of 10 mm and a depth of 30 mm is fitted as an object to be fitted, and a pin 5 having a diameter of 10 mm and a length of 40 mm is fitted as an object to be fitted. The case of fitting will be described.

First, the pin 5 is attached to the grip portion 4 of the robot 1.
Hold at. Then, the arm 2 is moved to a depth of 30.
The pin 5 is inserted into the small hole 6a of 10 mm by 10 mm in the force control mode (step S1).

At this time, when the pin 5 comes into contact with the opening of the small hole 6a or the inner peripheral surface, the reaction force is detected by the force sensor 3 (step S2).

Next, it is judged whether the reaction force, which is a force component in the direction perpendicular to the inner peripheral surface of the small hole 6a, is smaller than a preset condition ε, and if the reaction force is larger than the set value ε. If so, the amount of positional deviation is estimated by fuzzy inference (step S
3).

Subsequently, the position of the grip portion 4 of the robot 1 is corrected (step S4), the process returns to step S2, and the reaction force is detected by the force sensor 3.

The position correcting operation for detecting the reaction force and correcting the position again is repeated until the force applied to the grip portion 4 satisfies the setting condition ε.

If this condition that the reaction force is smaller than the set value ε is satisfied, the pin 5 is inserted to the bottom of the small hole 6a in the force control mode, assuming that the reaction force is almost non-contact,
The pin 5 is fitted into the cylindrical member 6 (step S5).

After that, the data, which is a set of the sum of the force detected when the first position correction amount is estimated and the position correction amount corrected during the series of operations, is set as a set of basic data for generating a fuzzy rule in advance. And regenerate the fuzzy rule (step S6).

As a result, the fuzzy rule is generated in a state where the setting condition ε is satisfied, so in the next assembling work, the reaction force is detected, the setting condition ε is determined, and then the position correction amount is calculated. As is clear from FIG. 8, the estimation operation requires almost only one position correction, and the number of position corrections significantly decreases even if the number of positioning trials increases.

Next, a second embodiment will be described.

The position control device in the robot 1 of the present embodiment has the same basic configuration as that of the first embodiment, but has a small hole 6a.
The fuzzy rule is regenerated with the pin 5 inserted to the bottom of the.

That is, similar to step S3 in the flow chart of FIG. 3, whether the reaction force which is a force component in the direction perpendicular to the inner peripheral surface of the small hole 6a is smaller than the preset condition ε1 in the flow chart of FIG. 4 as well. If it is determined and smaller than ε1, the pin 5 is inserted to the bottom of the small hole 6a by the force control mode, and the pin 5 is fitted to the cylindrical member 6 (step S
16).

Thereafter, it is judged whether or not the reaction force is smaller than the preset condition ε2 (step S17), and if the reaction force is larger than the set value ε2, the work is finished. Note that ε
2 is a value smaller than ε1 to make the condition strict.

On the other hand, when the reaction force is smaller than the set value ε2, the sum of the force detected at the time of estimating the first position correction amount and the position correction amount corrected during the series of operations is set. The data is added to the previous set of data and the fuzzy rules are regenerated (step S18).

As a result, since the fuzzy rule is regenerated at the time of inserting the pin 5 in the first embodiment, if the data differs due to noise in the electric signal system or a mechanical error peculiar to the robot 1, A new fuzzy rule is generated and the amount is added. In this example, after inserting pin 5, the judgment is made with stricter ε2 than the setting condition ε1 and the fuzzy rule is played when this condition is satisfied. I am trying to achieve it. Therefore, it is possible to suppress the possibility that the number of repeated position corrections increases until the preset condition is satisfied while the assembly work is repeatedly executed.

Next, a third embodiment will be described.

The position control device in the robot 1 of the present embodiment also has the same basic configuration as that of the first and second embodiments, and regenerates the fuzzy rule with the pin 5 inserted to the bottom of the small hole 6a. However, the processing of data according to the storage capacity of the memory 8 is performed.

That is, similarly to step S3 in the flowchart of FIG. 3, in the flowchart of FIG. 5, whether the reaction force, which is a force component in the direction perpendicular to the inner peripheral surface of the small hole 6a, is smaller than the preset condition ε1. If it is smaller than ε1, the pin 5 is inserted to the bottom of the small hole 6a by the force control mode and the pin 5 is fitted to the cylindrical member 6 (step S26).

Thereafter, it is judged whether or not the reaction force is smaller than the preset condition ε2 (step S27), and if the reaction force is larger than the set value ε2, the work is finished.

On the other hand, when the reaction force is smaller than the set value ε2, it is judged whether or not the storage capacity of the memory 8 has a margin (step S28). If there is no problem in the storage capacity, the first position is determined. Data that is a set of the force detected when estimating the correction amount and the sum of the position correction amount corrected during the above series of operations,
After adding to the previous data set and regenerating the fuzzy rule (step S30), the operation is completed.

On the other hand, when the memory 8 does not have a sufficient storage capacity, a data set having a close numerical value is searched for and integrated into one data (step S29), and then the operation is terminated through step S30.

For example, suppose that the force component for estimating the position correction amount in the x direction has a correlation with the force Fx acting in the x direction and the moment My about the y axis (this relationship is defined by the fuzzy rule). The distance function D for group-integrating the data shown in the previous equation)
Is defined, the pair that minimizes the distance function D can be found and made into one group.

Specifically, in the plurality of groups in Table 2, a set of data closest to the latest input data to be added is searched. For example, when the latest input data of group 3 is closest to the data of group 1, the average value in the data set of groups 1 and 3 is calculated to be 1
Performs arithmetic processing to be integrated into a set of data.

As a result, the data of the two groups are set to 1
One data set can be collected, the capacity of the memory 8 can be secured by that amount, and the number of held data will decrease, so that new data can be added. Therefore, even the general-purpose memory 8 having a limited storage capacity can be used to update the fuzzy rule semipermanently.

[0091]

As described above, according to the present invention, the data, which is a set of the sum of the detected value obtained when estimating the initial position correction amount and the position correction amount corrected during the repeated operation, is generated in advance. Since the fuzzy rule is regenerated when the first part is fitted into the predetermined position of the second part, the fuzzy rule is regenerated. The rules are not fixed as in the past, but are modified by newly input data for each actual assembly work. Therefore, since the position correction operation for moving the gripper to a predetermined position converges and decreases with each correction, the number of position corrections decreases in proportion to the increase in the number of assembly works, and the work time is significantly shortened. effective.

[Brief description of drawings]

FIG. 1 is an overall configuration diagram of a position control device in a robot according to a first embodiment of the present invention.

FIG. 2 is an explanatory diagram of a membership function that generates a fuzzy rule.

FIG. 3 is a flowchart of an assembly work operation according to the first embodiment of the present invention.

FIG. 4 is a flowchart of an assembly work operation according to the second embodiment of the present invention.

FIG. 5 is a flowchart of an assembly work operation according to the third embodiment of the present invention.

FIG. 6 is an overall configuration diagram of a position control device in a conventional robot.

FIG. 7 is a flowchart of a conventional assembly work operation.

FIG. 8 is a graph showing the relationship between the number of trials and the number of corrections according to the example of the present invention.

FIG. 9 is a graph showing the relationship between the number of trials and the number of corrections according to the conventional example.

[Explanation of symbols]

 1 Robot 2 Arm 3 Sensor 4 Gripping Part 5 Pin 6 Cylindrical Member 7 Control Device 8 Memory 9 Rule Generation Unit 10 Central Processing Unit

─────────────────────────────────────────────────── ─── Continuation of front page (51) Int.Cl. ⁵ Identification code Office reference number FI Technical display location G05D 3/12 U 9179-3H 305 V 9179-3H

Claims (3)

[Claims] 1. A gripping part provided at the tip of a robot grips a first part, and position correction is performed based on a detection value from a sensor that detects a reaction force and a moment acting on the gripping part. In a position control device for a force control robot that fits and assembles the first component at a predetermined position of the second component, a final assembly position at which the first component is fitted at a predetermined position of the second component is a reference position. At this time, data of a pair of a reaction force and a moment detected by the sensor when the first component or the second component is moved in an arbitrary direction and a position correction amount corresponding to the movement amount are stored. Means for inputting the set to the memory as basic data, and extracting from the set of data the elements including the reaction force and the moment that are correlated with each element of the position correction amount,
Means for integrating the extracted elements into a predetermined number of groups by clustering, and further replacing the characteristics of each group with a membership function to generate a fuzzy rule for estimating the position correction amount; A means for repeating the operation of estimating the position correction amount based on a rule and correcting the position of the grip portion until the detection value of the sensor satisfies a preset condition, and a means for detecting the first position correction amount. The fuzzy rule is generated by adding data that forms a set of the reaction force and the moment and the sum of the position correction amounts corrected during the repetitive operation at the time of position correction to the basic data set for generating a fuzzy rule in advance. A position control device in a force control robot, comprising a control device including means for regenerating 2. The control device, only when the detection value of the sensor reaches a value satisfying a preset condition, the detection value of the reaction force and the moment obtained when estimating the first position correction amount, and the grip portion. 2. The data, which is a set of the sum of the position correction amounts corrected during the repetitive operation of 1), is added to the set of basic data for generating the fuzzy rule in advance.
A position control device for the force control robot described. 3. The control device sets a pair of data including a detected value of the reaction force and the moment obtained when estimating the initial position correction amount and a sum of the position correction amounts corrected during the repeated movements of the grip portion. When adding to the set of basic data for generating fuzzy rules in advance, if there is no storage capacity in the memory, at least one set of data closest to the latest input data is stored in the memory. Search the data,
The position control device for a force control robot according to claim 1, wherein an average value in a set of this data and the latest data is obtained and integrated into one set of data.