JP3436713B2 - Robot oscillation detection device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device for detecting oscillation of a robot, for example, an articulated robot.

[0002]

2. Description of the Related Art For industrial robots, such as articulated robots, higher speed, higher accuracy, and lower cost are desired. In order to meet these requirements, when controlling the servo motor that drives the robot arm, the control gain for controlling the servo motor is optimally adjusted according to the dynamic characteristics of the tip of the robot arm that is the final control target. There is a need to. Since the dynamic characteristics of the articulated robot vary depending on the posture of each arm and the load acting on the robot, gain schedule control is performed to switch the control gain according to the posture and load of each arm. In order to perform this gain schedule control, it is necessary to set the optimum gain according to the posture and load of each arm, and therefore the adjustment of the control gain of one articulated robot becomes a huge amount, and a huge amount of labor and labor. It takes time.

Such adjustment of the control gain has been performed manually by the operator. However, the adjustment accuracy varies depending on the skill level of the operator, and after the displacement of the arm, vibration remains in the arm. Problems such as slower arm displacement may occur. Moreover, it takes a lot of trouble.

A device for automatically adjusting the control gain is used in order to eliminate such a problem. During the control gain adjusting operation, it is detected that the robot has oscillated and the robot is damaged. It needs to be prevented. As a conventional technique for detecting oscillation, a method for detecting oscillation of a servo motor is disclosed in Japanese Patent Laid-Open No. 8-116688. In this oscillation detection method, the mean square value of the servo motor current command value is obtained and compared with the set value to detect the oscillation of the servo motor.

[0005]

DISCLOSURE OF THE INVENTION Problems to be Solved by the Invention
In the oscillation detection method disclosed in Japanese Patent Publication No. 8, the current command value contains analog high-frequency noise or contains low-frequency signal components including DC components. In order to obtain, the measured current command value needs to be filtered by a high-pass filter or the like, which requires a filter and requires time and labor for data processing.

An object of the present invention is to provide an oscillation detecting device for a robot which can detect oscillation without requiring time-consuming data processing.

[0007]

According to a first aspect of the present invention, a predetermined oscillation detection evaluation function for determining oscillation based on a position deviation between a position command value and an actual operating position is set. Oscillation detection function setting means, oscillation detection function calculation means for calculating the oscillation detection evaluation function by substituting the position deviation after the command angle position arrival time when the position command value reaches the target value, and calculation by the oscillation detection function calculation means An oscillation detecting device for a robot, comprising: a determining unit that determines that the value is in an oscillating state when the value is equal to or more than a predetermined threshold value.

According to the present invention, the position deviation between the position command value and the actual operating position is substituted into the oscillation detection evaluation function for calculation, and when the calculated value is equal to or greater than a predetermined threshold value, the oscillation is generated. It is determined to be in a state. Since the position deviation is based on the actual position and does not include noise or low frequency signal components, no filtering process is required for the evaluation calculation. Therefore, oscillation can be detected without requiring time-consuming data processing. Also,
By using the position deviation after the command angular position arrival time, the oscillation state of the robot can be determined without being affected by the position deviation before reaching the command angular position arrival time.

According to a second aspect of the present invention, a state variable of a control system in actual operation is substituted, a gain adjustment evaluation function for evaluating a control gain is calculated, and the control value is the minimum. An oscillation detection device for a robot, which is provided in a control gain adjustment device for a robot that selects a gain as an optimum gain, detects the oscillation of the robot when the control gain is set and operates for evaluation. The function calculation means is characterized by calculating the oscillation detection evaluation function by substituting the position deviation during the calculation period in which the gain adjustment evaluation function is calculated. According to the present invention, the oscillation of the robot can be detected during the calculation period in which the gain adjustment evaluation function is calculated. This makes it possible to determine whether the robot is in the oscillation state while automatically adjusting the control gain and prevent damage to the robot.

The present invention according to claim 3 is characterized in that the oscillation detection evaluation function is a mean square value of the position deviation.
According to the present invention, since the oscillation detection evaluation function is the root mean square value of the position deviation from the time when the command value matches the target value, it is possible to detect that oscillation has occurred without being affected by overshoot or the like. It can be reliably detected.

[0011]

1 is a block diagram showing a configuration of a control gain adjusting apparatus 1 including a robot oscillation detecting apparatus 50 according to an embodiment of the present invention, and FIG. 2 is a control gain adjusting apparatus. 1 is a system diagram showing a schematic configuration of 1. The control gain adjusting device (hereinafter, may be simply referred to as “adjusting device”) 1 is a control gain Ksp of each axis of an articulated robot (hereinafter, may be simply referred to as “robot”) Ksp,
A device for adjusting Ksv, including a control structure 3 for controlling the robot 2, a measurement structure 4 for measuring an actual state of the robot 2, and a calculation structure 5 for calculating an optimum gain. Including.

The robot 2 has a base 6 and a base 6 having a base end portion.
A first arm 7 that is angularly displaceable about a first axis L1 and a second arm 8 whose base end is angularly displaceable about a second axis L2 at the tip of the first arm. The base end portion includes a mounting arm 9 connected to the tip end portion of the second arm 8 so as to be angularly displaceable around the third axis L3. The first arm 7 is
The base 6 is angularly driven by a servo motor about a first axis L1, and the second arm 8 is angularly driven by a servo motor about a second axis L2 with respect to the first arm 7 for mounting. The arm 9 is angularly driven by a servo motor about the third axis L3 with respect to the second arm 8. An article 10 such as a tool, a device, and an object to be transported can be removably attached to the tip of the attachment arm 9. The articulated robot 2 having such a plurality of angular displacement axes L1 to L3 is selectively driven by each servo motor.
It is possible to drive the arms 7 to 9 by angular displacement to move the article 10 in a complicated trajectory.

Such an articulated robot 2 has a high speed,
Higher precision and lower cost are desired, and in order to satisfy this requirement, in controlling each servomotor that drives each arm 7 to 9, according to the dynamic characteristics of the tip of each arm 7 to 9, It is necessary to optimally adjust the control gains Ksp and Ksv for controlling the servo motor. The dynamic characteristics of the articulated robot 2 are the same as those of the arms 7-9.
Since the position varies depending on the load acting on the robot 2 depending on the posture and the weight of the article 10, gain schedule control is performed to switch the control gain according to the posture and load of each arm 7-9. In order to perform this gain schedule control, it is necessary to set an optimum gain according to the postures and loads of the arms 7 to 9, and the adjusting device 1 is used to set this optimum gain.

The control structure 3 includes an operation command means 11 and
Position control means 12, speed control means 13, and detection means 1
4, a differential calculation means 15, an observer 16, a calculation means 17 for feedback of a twist amount, a calculation means 18 for feedback of a twist speed, and first to third subtraction means 1
9 to 21 and addition means 22. For example, the drive control of the first arm 7 will be described as an example. The operation commanding means 11 may be a servomotor (hereinafter, simply referred to as a "motor") that drives the first arm 7 to perform angular displacement with respect to the base 6. ) Position command value θr that commands the angular position of the output shaft of 23
A signal representing ef is given to the first subtraction means 19. The detection means 14 detects the actual angular position of the output shaft of the motor 23, and a signal representing the detected position value θm is used as the first subtraction means 19.
Give to. The first subtraction means 19 subtracts the position detection value θm from the position command value θref to obtain the position deviation Δx, and gives a signal representing the position deviation Δx to the position control means 12.
The position command value θref and the position detection value θm are angles with the position before the start of operation as a reference.

The position control means 12 has a predetermined position control transfer function, calculates the position control transfer function by substituting the position deviation Δx, and obtains a speed command value Vref for correcting the position deviation Δx. , And gives a signal representing the speed command value Vref to the second subtraction means 20. The signal representing the position detection value θm output from the detecting means 14 is the differential calculating means 15
The differential calculation means 15 differentiates the detected position value θm with respect to time to obtain the angular displacement speed of the output shaft of the motor 23, and a signal representing the calculated speed value θms is used as the second subtraction means 2
Give to 0. The second subtraction means 20 determines the speed command value Vref.
Calculate the speed deviation ΔV by subtracting the calculated speed value θms from
A signal representing the speed deviation ΔV is given to the speed control means 13.

The speed control means 13 has a predetermined speed control transfer function, calculates the speed control transfer function by substituting the speed deviation ΔV, and obtains a torque command value Tref for correcting the speed deviation ΔV. , And gives a signal representing the torque command value Tref to the third subtracting means 21. Third subtraction means 21
Is further given a signal indicating the torque correction value Ts from the adding means 22, calculates the torque control value Tcom by subtracting the torque correction value Ts from the torque command value Tref, and the signal indicating the torque control value Tcom is sent to the motor 23. Given. The motor 23 operates based on the torque control value Tcom, that is, angularly displaces the output shaft.

Position detection value θ output from the detection means 14
The signal indicating m is also given to the observer 16, and the signal indicating the torque control value Tcom output from the third subtracting means 21 is also given to the observer 16. The observer 16 estimates the twist amount θs and the twist speed θss. Specifically, the observer 16 has a predetermined twist amount calculation function and a predetermined twist speed calculation function, and calculates the twist amount θs and the twist speed θss by substituting the position detection value θm and the torque control value Tcom. The observer 16 calculates a signal representing the twist amount θs by the calculating means 17
Then, a signal representing the twisting speed θss is given to the calculating means 18.

The calculating means 17 has a twist amount feedback function determined based on the control gain Ksp, calculates the twist amount feedback function by substituting the twist amount θs, and outputs a signal representing the calculated value Kθs. Adder 2
Give to 2. The calculation means 18 has a torsion speed feedback function determined based on the control gain Ksv, calculates the torsion speed feedback function by substituting the torsion speed θss, and adds a signal representing the calculated value Kθss to the addition means 22. Give to. The adding means 22 is used for each computing means 1
The calculated values Kθs and Kθss from 7 and 18 are added to obtain the torque correction value Ts, and a signal representing this torque correction value Ts is given to the third subtracting means 21.

As described above, the control structure 3 is composed of the detecting means 1
The first negative feedback circuit having 4 feeds back the position detection value θm, and the detecting means 14 and the differential calculating means 1
A second negative feedback circuit having a speed calculation value θms
Is fed back to the detecting means 14, the observer 16,
Third having each calculation means 17, 18 and addition means 22
Twisting amount θs and twisting speed θ
The motor 23 is controlled by feeding back ss. The angular displacement of the output shaft of the motor 23 is transmitted to the first arm 7 via the speed reducer, which drives the first arm 7 for angular displacement and controls the motor 23 to make the first displacement.
The arm 7 can be drive-controlled.

When the first arm 7 is driven to be angularly displaced by the motor 23, the angular displacement of the output shaft of the motor 23 is transmitted to the tip of the first arm 7 via the elastic deformation element 25. The elastic deformation element 25 includes the speed reducer and the first arm 7
The angular displacement of the motor output shaft is transmitted to the arm distal end after the elastic deformation of the elastic deformation element 25 is superimposed.

FIG. 3 is a diagram showing the relationship among the detected position value θm, the arm tip position θa, and the twist amount θs. As described above, the position detection value θm is represented by the angle around the axis line of the output shaft from the reference angular position 27 before the operation of the motor 23, specifically, the output shaft. In FIG. 3, the axis line of the output shaft is shown in alignment with the first axis L1 for easy understanding. The position detection value θm is a value converted into an angle after being converted by the speed reducer.

The arm tip position θa is obtained from a reference angular position of the tip portion of the first arm before starting the operation of the motor 23 (matched with the reference position of the output shaft and attached with the same reference numeral to facilitate understanding) 27. , The angle about the first axis L1. Although the arm tip position θa matches the position detection value θm if the elastic deformation element 25 does not exist,
Due to elastic deformation of the elastic deformation element 25, for example, elastic bending of the first arm 7, the position detection value θm deviates. Such elastic strain is the twist amount θs estimated by the observer 16, and the twist amount θs is represented by θs = θa−θm (1) and is a value obtained by subtracting the position detection value θm from the arm tip position θa. Equivalent to. Thus, the arm tip position θa
Is a physical quantity that changes depending on the elastic deformation element 25, and naturally the twist amount θs is a physical quantity that changes depending on the elastic deformation element 25.

Thus, in the control system, the twist amount θs, which is a physical quantity that changes depending on the elastic deformation element 25 present on the trailing side of the motor 23, and the twist speed θss, which is a differential calculation value thereof, are fed back and controlled. Thus, the motor 23 can be controlled based on the twist amount θs and the twist speed θss, and the elastic deformation element 25 can be controlled.
It is possible to control the drive of the first arm 7 based on the elastic deformation of, and thus to control the position of the tip portion of the first arm 7. By optimally determining the respective control gains Ksp and Ksv in the respective computing means 17 and 18 of the third negative feedback circuit which feeds back the twist amount θs and the twist speed θss, the first resulting from the elastic deformation of the elastic deformation element 25. The angular displacement vibration around the first axis L1 at the tip of the arm 7 can be suppressed to the minimum.

Although the control structure 3 has been described in relation to the first arm 7, the control structure 3 can control the second arm 8 and the mounting arm 9 as well. Specifically, the operation command means 11 first subtracts a signal representing a position command value θref that commands the angular position of the output shaft of the servo motor that drives the second arm 8 and the servo motor that drives the mounting arm 9. In addition to being provided to the means 19, the detection means 14 is capable of detecting the angular position of the output shaft of each motor that drives each arm 8 and 9, and each of the other means 12, 13, 15
.About.18 and 20 to 22 operate similarly in relation to the respective arms 8 and 9.

The adjusting device 1 of the present embodiment is a device for adjusting each control gain Ksp, Ksv to an optimum value. The measurement structure 4 includes a twist acquisition unit 43 that acquires a twist amount θs and a twist speed θss related to the control gains Ksp and Ksv, and a deviation acquisition unit 44 that acquires a position deviation Δx. Specifically, the twist acquisition means 43
Is provided with a signal indicating the twist amount θs and a signal indicating the twist speed θss, and the twist amount θs and the twist speed θss are acquired from the observer 16. Further, the deviation acquisition means 44 is given a signal representing the position deviation Δx from the first subtraction means 19, whereby the position deviation Δx is obtained.
To get.

The arithmetic structure 5 comprises a gain adjusting function setting means 30, a gain adjusting function arithmetic means 31, and a selecting means 3.
2, oscillation detection function setting means 33, oscillation detection function calculation means 34, and determination means 35. Arithmetic structure 5
Is realized by, for example, a personal computer.

The gain adjustment function setting means 30 is an optimum value of the control gains Ksp and Ksv that minimize the twist amount θs and the twist speed θss when the robot 2 is controlled by the control structure 3 as described above. Optimal gain K
The gain adjustment evaluation function Jg for obtaining the optimum gain Ksv-opt which is the optimum value of the sp-opt and the control gain Ksv is set. The gain adjustment evaluation function Jg is a function for evaluating the control gains Ksp and Ksv.
The gain adjustment function setting means 30 gives a signal representing the set gain adjustment evaluation function Jg to the gain adjustment function calculation means 31.

The gain adjusting function computing means 31 is comprised of the computing unit 3
7 and a storage unit 38, and the storage unit 38 can store the set gain adjustment evaluation function Jg. A signal indicating the twist amount θs and a signal indicating the twist speed θss are given to the calculation unit 37 from the twist acquisition unit 43.
The calculation unit 37 stores the twist amount θs and the twist speed θss estimated based on the detected position value θm of the output shaft of the motor 12, which is a state variable of the control system during actual operation, in the storage unit 38.
The gain evaluation function Jg is calculated by substituting it into the gain evaluation function Jg read from.

A signal representing the calculated value of the gain adjusting function Jg calculated by the calculating section 37 is given to the selecting means 32 from the calculating section 37. The selection means 32 includes a calculation unit 37.
The control gains Ksp and Ksv, which reduce the calculated value of the gain adjustment function Jg, are set to optimal gains Ksp-opt and Ks.
Select as v-opt.

The oscillation detection function setting means 33, when the bot 2 is controlled by the control structure 3 as described above,
An oscillation detection evaluation function Jv for determining the oscillation is set based on the position deviation Δx between the position command value θref and the position detection value θm which is the actual operating position. The oscillation detection function setting means 33 is realized by input means such as a keyboard and a mouse, and the gain adjustment function setting means 3 is provided.
It may be combined with 0 or may be individual. The oscillation detection function setting means 33 gives a signal representing the set oscillation detection evaluation function Jv to the oscillation detection function calculation means 34.

The oscillation detection function calculation means 34 includes a calculation section 39.
And the storage unit 40, and the set oscillation detection evaluation function Jv can be stored in the storage unit 40. Calculation unit 39
A signal representing the position deviation Δx is given to the signal from the deviation acquisition means 44. The calculation unit 39 substitutes the position deviation Δx into the oscillation detection evaluation function Jv stored in the storage unit 40 to calculate the oscillation detection evaluation function Jv.

A signal representing the calculated value of the oscillation detection function Jv calculated by the calculating section 39 is given to the judging means 35 from the calculating section 39. The selection unit 32 determines that the oscillation state is in the oscillation state when the calculated value of the oscillation detection function Jv by the calculation unit 39 is greater than or equal to a predetermined threshold value. The threshold value for determining the oscillation may be a value that can be arbitrarily set by a keyboard and mouse used as the oscillation detection function setting means 33, or an input means such as a keyboard and mouse provided separately. , Or may be a fixed value.

FIG. 4A is a graph showing an example of the position command value θref, the horizontal axis is the time t (seconds), and the vertical axis is the position command value θref (degrees). FIG. 4 (2) is a graph showing an example of the angular position θa of the arm tip portion corresponding to FIG. 4 (1), the horizontal axis is time t (seconds), and the vertical axis is the angular position θa (degrees). Is. FIG. 4 (3) corresponds to FIG.
It is a graph which shows an example of twist amount (theta) s corresponding to (1) and FIG.4 (2), a horizontal axis is time t (second), and a vertical axis is twist amount (theta) s (V). In FIG. 4C, the twist amount θs is shown as a voltage waveform of the signal output from the observer 16.

Since FIG. 4 is a graph that can be taken by each of the arms 7 to 9, for example, the respective arms 7 to 9 will be generically referred to below and the gain adjustment evaluation function Jg will be specifically described. The adjusting device 1 has a control gain Ks of the articulated robot 2.
In the apparatus for adjusting p and Ksv, the target physical quantity to be minimized is at least the twisting speed θss about each axis, and in the present embodiment, the twisting quantity θs is further included. The gain adjustment evaluation function Jg is the twist speed θ.
Offset value θss_m of ss and the predetermined twisting speed
It includes the time integral of the value obtained by multiplying the square of the difference between and by the time weighting coefficient ng (t). Gain adjustment evaluation function Jg of the present embodiment
Is a function evaluated based on FIG. 4 (3), and is expressed by the following equation (2).

[0035]

[Equation 1]

Here, t0 is the operation start command time, tj is the command angular position arrival time, and tf is t.
It is the evaluation end time several seconds after j, α and b are constant weighting factors, and θs_m is an offset value of the twist amount. The evaluation end time tf is, for example, a time 2 seconds after the operation start command time. Each constant weighting coefficient α, β is determined by the shape of each arm 7-9. The offset value θs_m of the twist amount is a value determined by the bending of each arm due to the weight of the article 10 and the like, and FIG.
It is a value corresponding to each time on the straight line obtained from the twist amount θs. This straight line is (5/8 tf,
θs_m3) and (7 / 8tf, θs_m4) are given by a function representing a straight line. Where θs
_M3 is the average value of θs in the third section (tf / 2 ≦ t <3tf / 4), and θs_m4 is the fourth section (3
It is the average value of θs in tf / 4 ≦ t <tf). Further, the offset value θss_m of the twisting speed is determined by the bending of each arm due to the weight of the article 10 and the like, like the offset value θs_m.

FIG. 5 is a graph showing an example of the twist speed θss and the time weighting coefficient ng (t). The twisting speed θss is shown as a voltage waveform of the signal from the observer 16 with the horizontal axis as time t (second) and the vertical axis as the twisting speed θss (V). The time weighting factor ng (t) is shown with the horizontal axis as time t (seconds) and the vertical axis as time weighting factor ng (t). The time weighting factor ng (t) is
That is, it is a value that changes with time t. The time weighting factor ng (t) is the second to the fourth section (the second section is tf /
In order to suppress the residual vibration of 4 ≦ t <tf / 2), it is set to increase with the passage of time to prevent overshoot and undershoot in the first section (t0 ≦ t <tf / 4). In order to reduce the power consumption, it is set so that it becomes smaller with the passage of time. In the present embodiment, as shown in FIG. 5, it is set to increase with the passage of time.

That is, if the robot 2 positions each joint axis L1 to L3 at high speed and once settles and then the residual amount vibration occurs, the setting of the hand, that is, the position of the article 10 is delayed, and precise positioning to the target position is difficult. And is not preferable as a response. In view of this point, the above-mentioned gain adjustment evaluation function J
g is defined to be low when residual vibration appears.

FIG. 6 shows the twist amount θs and the twist speed θ.
It is a graph for showing another example of ss. 6 (1)-
In FIG. 6 (4), each horizontal axis represents time t. Figure 6
(1) shows the twist amount θs on the vertical axis, and the observer 16
The signal waveform from is shown. In FIG. 6 (2), the vertical axis represents the twisting speed θss, and the signal waveform from the observer 16 is shown. In FIG. 6C, the vertical axis represents the position deviation Δx, and the signal waveform from the first subtraction unit 19 is shown. In FIG. 6 (4), the vertical axis represents the operating state.

The signal shown in FIG. 6 (4) shows -5V until the operation command is started, 0V is displayed until the target value is reached after the operation command is started, and 5V is displayed when the target value is reached.
This signal is given from the operation command means 11 to the time acquisition means 45 of the measurement component 4, and from this means 45 to the calculation unit 37 of the gain adjustment function calculation means 31 and the oscillation detection function calculation means 39. The operation start command time t0 and the command angular position arrival time tj can be acquired based on this signal.

FIG. 7 is a graph showing yet another example of the twist amount θs. 7 (1) to 7 (4), each horizontal axis represents time t (seconds). 7 (1) shows the position command value θref (degree) on the vertical axis, FIG. 7 (2) shows the position detection value θm (degree) on the vertical axis, and FIG. 7 (3) shows the vertical axis. The twist amount θs (degree) is shown, and FIG. 7 (4) shows the position θa of the tip end portion of each arm 7 to 9 on the vertical axis. FIG. 8 is a graph showing yet another example of the twist amount θs. In FIGS. 8 (1) to 8 (4), each horizontal axis represents time t (seconds).
Indicates. In FIG. 8A, the vertical axis indicates the position command value θref.
8 (2) shows the position detection value θm on the vertical axis.
8 (3), the vertical axis indicates the twist amount θs.
8 (4) shows the position θa of the tip of each arm 7-9 on the vertical axis.

When the robot 2 is controlled by the control structure 3, when the respective control gains Ksp and Ksv are optimally set, as shown in FIG. 7, the vibration of the output shaft of the motor is prevented and the twisting of each arm is prevented. Vibration is prevented and each arm 7-9
If the control gains Ksp and Ksv are not optimally set, it is possible to control the tip position θa of the motor so that it does not vibrate as shown in FIG. However, torsional vibration of each arm occurs, and therefore the tip position θa of each arm 7-9 vibrates. In order to determine the optimum gains Ksp-opt and Ksv-opt that enable control as shown in FIG. 7,
The above-mentioned evaluation function Jg is used for evaluation.

In the present embodiment, when the optimum gain in one posture of one axis is selected, at least one of the posture and the posture is changed to operate the robot 2. Specifically, the arithmetic structure 5 further includes an operation management unit 47, and when the selection unit 32 determines the optimum gains Ksp-opt and Ksv-opt, a signal indicating that the optimum gain has been selected. Is given to the operation management means 47, and a signal for instructing to operate the robot 2 by changing at least one of the axis and the attitude is given to the operation command means 11, and the operation of the robot 2 is controlled based on this. . In this way, the axis and posture are changed sequentially,
Optimal gain Ksp-opt, Ks in each posture of each axis
The v-opt is, so to speak, automatically selected.

FIG. 9 is a flowchart showing the operation of the adjusting device 1. FIG. 9 (1) shows the entire operation.
9 (3) shows the axis adjusting operation, FIG. 9 (4) shows the posture adjusting operation, and FIG.
(5) shows a gain selection operation. As shown in FIG. 9 (1), at step a0, when the operator gives an instruction to the operation management means 47 by a switch operation or the like to start the adjustment work of the control gains Ksp and Ksv of the robot 2, at step a1. The gain adjustment is performed for all the loads acting on the robot 2, and the gain adjustment ends in step a2.

The detailed flow of gain adjustment at full load in step a1 is shown in FIG. 9 (2). Step a
When the process moves from 0 to step a1, the gain adjustment at full load is started in step b0, and in step b1,
The load on the robot 2 is set by the worker removing or replacing the article 10. This load is set to a load for which gain adjustment is not completed while checking the display screen of the operation management unit 47, for example. When the load is set, the operator gives an instruction to the operation management means 47 by operating a switch or the like, the process proceeds to step b2, and based on the management by the operation management means 47, all the joint axes L1 to L1. A gain adjustment for driving the arm around L3 is performed, and when this adjustment is completed, a signal indicating that the optimum gain has been determined is given from the selecting means 32 to the operation managing means 47, and the process proceeds to step b3. .
In step b3, the operation management means 47 determines whether or not the gain adjustment has been completed for all loads, and if not completed, the process returns to step b1 to set a load for which gain adjustment has not been made. When it is determined in step b3 that the gain adjustment for all loads is completed, the process proceeds to step b4, the gain adjustment for all loads is completed, and the process proceeds to step a2.

The detailed flow of gain adjustment in all joint axes L1 to L3 in step b2 is shown in FIG. 9 (3). When the process moves from step b1 to step b2, the gain adjustment for all joint axes L1 to L3 is started in step c0, and in step c1, the operation management means 47 sets which axis operation the gain adjustment is to be performed. The target axis is set to the axis for which gain adjustment has not been completed. When the target axis for gain adjustment is set, the process proceeds to step c2, and the gain adjustment is performed in all postures, specifically, in the operation from all the operation start positions of the arm around the target axis, and this adjustment is performed. When finished,
A signal indicating that the optimum gain has been determined is given to the operation management means 47 from the selection means 32, and the process proceeds to step c3. In step c3, the operation management means 47
It is determined whether or not the gain adjustment for all joint axes L1 to L3 is completed, and if not completed, the process returns to step c1.
The target axis is set to the joint axis for which gain adjustment has not been performed. When it is determined in step c3 that the gain adjustment for all joint axes L1 to L3 has been completed, the process proceeds to step c4, the gain adjustment for all joint axes is completed, and the process proceeds to step b3.

The detailed flow of gain adjustment in all postures in step c2 is shown in FIG. 9 (4). Step c
When the process shifts from 1 to step c2, gain adjustment in all postures is started in step d0, and in step d1,
The operation management unit 47 sets from which position around the target axis for gain adjustment the arm operation is to be gain adjusted. This posture is set to a posture in which gain adjustment has not been completed. When the posture of the gain adjustment target is set, the process shifts to step d2 to search for the optimum gain, and when this search is completed, a signal indicating that the optimum gain has been determined is output from the selection means 32. It is given to the operation management means 47 and the process proceeds to step d3.
In step d3, the operation management unit 47 determines whether or not the gain adjustment in all the postures is completed. If not completed, the process returns to step d1 and the target posture is set in the posture in which the gain adjustment is not performed. It Step d
If it is determined in 3 that the gain adjustment in all postures is completed, the process proceeds to step d4, the gain adjustment in all postures is completed, and the process proceeds to step c3.

The detailed flow of the search for the optimum gain in step d2 is shown in FIG. 9 (5). When the process shifts from step d1 to step d2, the search for the optimum gains Ksp-opt and Ksv-opt is started at step e0, and at step e1, the control gains Ksp and Ksv for controlling the robot 2 are set by the selection means 32. Is set.
When the control gains Ksp and Ksv are set, a signal representing them is given from the selection means 32 to the operation management means 47,
Further, based on this, a signal is given from the operation management means 47 to the operation command means 11.

Next, at step e3, the robot 2 is controlled and operated by the set control gain. In this operation, for example, each arm is moved three times around each axis. Along with this operation, in step e2, the twist amount θs and the twist speed θss during the movement of the robot 2 are acquired by the twist acquisition unit 43. The twist amount θs and the twist speed θss obtained in step e4 are given to the gain adjustment function calculating means 37, the gain adjustment evaluation function Jg is calculated in step e5, and the calculated values are given to the gain selecting means 32. Next, in step e6, the optimum gain Ksp-op is calculated based on the calculated value of the gain adjustment function Jg.
It is determined whether or not t, Ksv-opt is obtained, and if the optimum gain is not obtained, the process returns to step e1, a new control gain is set, and the operation of steps e2 and e3 and subsequent steps is repeated to obtain the optimum gain. If obtained, step e7
, The search for the optimum gain is completed, and step d3
Move to.

According to such a flow, the adjusting device 1
It is possible to adjust the control gains Ksp and Ksv in all joint axes L1 to L3 and all postures for one load. FIG. 9 shows a flow for performing gain adjustment of all postures in each joint axis, but the present invention is not limited to this.
In each posture, the gain adjustment may be performed according to the flow for adjusting the gains of all joint axes, and in this case also, the control gains Ksp and Ksv in all joint axes and all postures for one load can be adjusted. it can.

FIG. 10 is a flowchart showing a more specific example of the operation shown in FIG. 9 (5), and FIG. 11 is a control gain Ksp, which changes according to the flow shown in FIG.
It is a graph which shows Ksv. 11 (1) shows the control gain at the start of the search, FIG. 11 (2) shows the search state in the control gain adjustment direction, and FIG. 11 (3) shows the gain adjustment in the control gain decreasing direction. FIG. 11 shows the search state.
(4) shows a search state in another adjustment direction, and FIG.
(5) shows a search state in which the gain is adjusted in another descent direction, and FIG. 11 (6) shows a state where the optimum gain is reached. In FIG. 11 (1) to FIG. 11 (6), the X axis in the lower right direction shows the control gain Ksp, and the Y axis in the upper right direction is
The control gain Ksv is shown, and the vertical axis (Z axis) shows the calculated value of the gain adjustment evaluation function Jg.

In the present embodiment, the selection means 32 determines the calculated value of the gain adjustment evaluation function Jg when operating at the current optimum candidate control gains Ksp0 and Ksv0, and other values near the current optimum candidate control gain. Control gain Ksp1, Ks
The calculated value of the gain adjustment evaluation function when operating at v1 is compared, and when the calculated value based on the other control gain is smaller than the calculated value based on the control gain of the optimum candidate, another control gain is set to a new control gain. Then, when the calculated value based on the control gain of the optimum candidate is equal to or smaller than the calculated value based on the other control gains, the control gain of the optimum candidate is selected as the optimum gain. A flow including the operation of the selecting means 32 is shown in FIG.
0.

At step f0 corresponding to step e0,
When the search for the optimum gains Ksp-opt and Ksv-opt is started, in step f1 corresponding to step e1,
The selection means 32 sets the control gains Ksp and Ksv for controlling the robot 2 to optimum candidate values, that is, the optimum candidate control gains Ksp0 and Ksv0. The initial values of the control gains Ksp0 and Ksv0 of this optimal candidate are
The robot 2 is configured so that the calculated value of the gain adjustment function Jg reaches a global minimum value without falling into a local minimum value.
It is a value that is theoretically derived by the configuration of. When the optimum candidate control gains Ksp0 and Ksv0 are set, a signal representing the control gains is given from the selecting means 32 to the operation managing means 47, and based on this, a signal is given from the operation managing means 47 to the operation commanding means 11. It

Next, step f2 corresponding to step e3
Then, the robot 2 is controlled and operated by the set control gain of the optimum candidate. In this operation, for example, each arm is moved three times around each axis. Further step e
In step f2 corresponding to 2 as well, the twist amount θs and the twist speed θss during the operation of the robot 2 are acquired by the twist acquisition unit 43. Further, in step f2 corresponding to step e4 and step e5, the obtained twist amount θs and twist speed θss are given to the gain adjustment function computing means 37, and as shown in FIG. 11 (1),
The gain adjustment evaluation function Jg is calculated, and this calculated value Jg0
Are provided to the gain selection means 32.

Next, at step f3, as shown in FIG. 11 (2), other control gains Ksp1, Ksv1 having different numerical values are assigned to the control gains Ksp, Ksv at predetermined intervals near the optimum candidate control gain. , The twist amount θs and the twist velocity θss are obtained by operating the robot 2 in the same manner as the operation of step f2, and the gain adjustment evaluation function Jg is obtained.
The calculated value Jg1 of is given to the gain selecting means 32. When there are two control gains as in the present embodiment, there are two control gains in the vicinity of each control gain, so that for the current optimum candidate control gains Ksp0 and Ksv0,
The four control gains, that is, one of the control gains has the same value, and the control gain obtained by changing the other control gain is set as the other control gains Ksp1 and Ksv1.
Calculated value Jg1 is obtained. This step f3 corresponds to step e5.

Next, step f4 corresponding to step e6
Then, the calculated value Jg0 is compared with the calculated value Jg1. Since there are four calculated values Jg1 as described above, each calculated value Jg1
And the calculated value Jg0 are compared. If there is a calculated value Jg1 smaller than the calculated value Jg0, the process proceeds to step f7 corresponding to step e1 and the control gains Ksp1 and Ks at which the smallest calculated value Jg1 is obtained among the small calculated values Jg1.
v1 is a new optimum candidate control gain Ksp0, Ksv0
To step f2. Step f2
The operations of steps f4 and f7 are repeated until the calculated value Jg0 becomes equal to or smaller than the calculated value Jg1.

Specifically, as shown in FIG. 2B, the steepest descent direction is obtained by comparing the optimum candidate control gains Ksp0 and Ksv0 with four sets of neighboring control gains Ksp1 and Ksv1, and the following figure (3) ), The control gains Ksp, Ksv are changed only in that direction, and the calculated values Jg0, J
The calculated value Jg0 is the calculated value J while comparing g1 on a one-to-one basis.
Control gains Ksp, K in that direction until g1 or less
Change sv. When the control gain is changed in that direction and the calculated value Jg0 becomes equal to or smaller than the calculated value Jg1, the current optimum control gain is again compared with the neighboring four control gains as shown in FIG. 11 (4). Then, find the steepest descent direction. After that, the control gains Ksp and Ksv are changed only in that direction as shown in FIG.
While comparing 0 and Jg1 on a one-to-one basis, the control gain Ks is increased in that direction until the calculated value Jg0 becomes equal to or smaller than the calculated value Jg1.
Change p and Ksv.

The optimum candidate control gain Ksp0,
If the control gains Ksp0 and Ksv0 of the optimum candidates become equal to or lower than the control gains Ksp1 and Ksv1 in the vicinity as shown in FIG. 11 (6), the change of Ksv0 is repeated.
In step 4, it is determined that the calculated value Jg0 is less than or equal to each calculated value Jg1, the process proceeds to step f5, and the optimum candidate control gains Ksp0 and Ksv0 that have obtained the calculated value Jg0 are the selecting means 32.
The optimum gains Ksp-opt and Ksv-opt are selected by and the gain adjustment is completed in step f6. Steps f5 and f6 correspond to step e7.

In the adjusting device 1 of the present embodiment as described above, the twist amount θs and the twist speed θs during the actual operation are set.
It is possible to evaluate the control gains Ksp and Ksv by substituting s into the gain adjustment evaluation function Jg. The control gain is selected according to this evaluation, the calculated value of the gain adjustment evaluation function Jg is the minimum, and therefore the twist amount θs. It is possible to obtain the optimum gain that minimizes the twisting speed θss.
Control gains Ksp and Ks based on the gain adjustment evaluation function Jg
The evaluation of v is performed by twisting amount θs depending on the elastic deformation element 25.
In addition, the control gains Ksp and Ksv can be adjusted in consideration of the elastic deformation of the elastic deformation element 25, for example, the torsion of the arm.
Specifically, the robot 2 drives the arms 7 to 9 by the motor 23, and the motor 23 is controlled based on the control gains Ksp and Ksv. Therefore, only the position of the motor can be adjusted as in the conventional technique. Based on the control gain Ks
Even if p and Ksv are adjusted, elastic deformation of the arms 7 to 9 is not taken into consideration, and such control gains Ksp and Ks
With sv, the positions of the tips of the arms 7 to 9 cannot be accurately controlled. On the other hand, in the present embodiment, the control gains Ksp and Ksv are evaluated based on the twist amount θs and the twist speed θss depending on the elastic deformation element 25, and thus the adjusted optimum gains Ksp-opt and Ks are obtained.
By using v-opt, the robot 2 can be controlled with high accuracy in consideration of the elastic deformation 25.

In addition, the twist amount θs and the twist speed θss are the position detection values which are state variables of the control system in the control structure 3 in the operation system including the robot 2 and the control structure 3. θm and torque control value T
The twist amount θ
In order to obtain s and the twisting speed θss, it is not necessary to separately provide a measuring device such as a camera outside the operation system, and it is possible to prevent the device from becoming complicated and large.

Further, since the evaluation is performed by the evaluation function Jg so that the twisting speed θss becomes small, the vibration of the arm can be prevented even when the twisting amount is small. That is, the residual vibration generated with a small amplitude can be suppressed more effectively. More specifically, the gain adjustment evaluation function Jg includes a time integral of a value obtained by multiplying a difference between the twisting speed θss and the offset value θss_m by a time weighting coefficient ng (t). The offset value θss_m is
It is set in consideration of the amount of deformation of the arm due to gravity, etc., which enables highly accurate positioning of the arm tip. Further, since the twist amount θs is also an object of evaluation, it is possible to control the vibration in consideration of a large amplitude vibration, that is, overshoot and undershoot, so that the vibration can also be suppressed.

Further, by setting the time weighting coefficient ng (t), the manner of positioning the arm tip can be set. In this embodiment, the time weighting factor ng
Since (t) is set to increase with the passage of time, it is possible to control the robot 2 so that residual vibration that remains even after the passage of time can be preferentially suppressed.

The time weighting coefficient ng (t) can be set small as time passes, and in this case, the initial displacement of the arm tip portion such as overshoot and undershoot can be preferentially suppressed. So that the robot can be controlled.

Further, the control gains Ksp0 and Ks of the optimum candidate
When the calculated values Jg1 based on other control gains Ksp1 and Ksv1 in the vicinity are smaller than the calculated value Jg0 based on v0, the other control gains are set as new optimal candidate control gains, and based on the optimal candidate control gains Ksp0 and Ksv0. When the calculated value Jg0 is equal to or smaller than the calculated value Jg1 based on the other control gains Ksp1 and Ksv1, the optimum candidate control gain is selected as the optimum gains Ksp-opt and Ksv-opt, so that the gain adjustment evaluation function Jg is used. The evaluation work can be reduced as much as possible to select the optimum gain, and the work time of the gain adjustment work can be shortened.

In particular, when there are a plurality of control gains, the predetermined optimum control gains Ksp0 and Ksv0 are compared with other control gains Ksp1 and Ksv1 in the vicinity,
The direction in which the calculated value Jg1 drops sharply is specified, and the calculated value Jg
The work time can be further shortened by changing the control gain in that direction and comparing the calculated values until g1 does not drop.

When the optimum gain for one posture of one axis is selected, the operation management means 47 changes at least one of the posture and the posture. In this way, the axes and postures are sequentially changed, and the optimum gain in each posture of each axis is selected. Therefore, it is possible to automatically select the optimum gain in all postures of all axes under a predetermined load. In this way, the control gain can be adjusted automatically, so compared to the case where the operator manually adjusts the gain, highly precise adjustment is possible, and the work time is short and the robot (2) The production efficiency can be improved and the delivery deadline can be shortened.

In the adjusting device 1 as described above, the gain is automatically adjusted, so to speak, without human intervention except for changing the load. Therefore, the robot oscillates, that is, the time is increased due to the set values of the control gains Ksp and Ksv. When a vibration that does not converge even after the passage of time occurs, it is necessary to prevent the damage of the robot by changing the control gain on behalf of the operator. The adjustment device 1 includes an oscillation detection device 50 for detecting oscillation of the robot 2. The oscillation detection device 50 basically includes the oscillation detection function setting means 33, the oscillation detection function calculation means 34, and The determination means 35 is included.

FIG. 12 is a graph showing an example of the position deviation Δx, where the horizontal axis represents time t (seconds) and the vertical axis represents the position deviation Δx (degrees). FIG. 12 shows, for example, each arm 7-9.
Is a graph that can be taken,
9 to 9 collectively, the oscillation detection evaluation function Jv will be specifically described. The oscillation detection device 50 adjusts the control gains Ksp and Ksv of the articulated robot 2 when the robot 2
12 is a device for detecting that the
Based on the position deviation Δx calculated by the first subtracting means 19 as shown in (1), the evaluation is performed and the oscillation is detected. The oscillation detection evaluation function Jv for detecting this oscillation is the root mean square value of the position deviation Δx, and is specifically expressed by the following equation (3).

[0069]

[Equation 2]

Here, t0 is the operation start command time, tj is the command angle position arrival time, tf is the evaluation end time several seconds after the command angular position arrival time tj, and these are the above-mentioned values. This is the same as the variable relating to the time of the gain adjustment evaluation function Jg.

FIG. 13 is a flow chart showing the oscillation detecting operation. At the same time that the search for the control gain shown in FIG. 9 (5) and FIG. 10 is started, the oscillation detection operation is started at step g0. Next, at step g1, when a pair of control gains Ksp and Ksv are set and the robot 2 is operated for evaluation, the positional deviation Δ during the operation is set.
x, that is, the position deviation Δx over the same time as the evaluation of the control gain as shown in FIG. 12 is acquired by the deviation acquisition means 44, and at the time t by the time acquisition means 45.
Is obtained. The position deviation Δx and the time t are given to the calculation unit 39 of the oscillation detection function calculation means 34, and the acquired position deviation Δx is substituted into the oscillation detection evaluation function Jv.
Is calculated, and the calculated value Jv1 is obtained.

Next, at step g2, the judging means 35 judges whether or not the calculated value Jv1 is equal to or larger than a predetermined threshold value Jv0. When the calculated value Jv1 is equal to or larger than the threshold value Jv0, oscillation is performed at step g3. Is determined to be
A signal representing oscillation is given to the gain setting means 32 and also given to the operation management means 47. On the basis of this, in step g4, the selecting means 32 performs one step before the state in which it is determined to be oscillation in the evaluation work of the control gain confirmed not to cause oscillation, for example, the control gain repeatedly performed. Set the control gain that was set,
The operation management means 47 limits the direction of changing the control gain so as to start the search for the control gain in a direction different from the control gain in which the oscillation has occurred, and ends the oscillation detection operation in step g5. Thus, even if the oscillation is detected once and the operation is terminated, when the adjustment work of the control gains Ksp and Ksv is continued, the oscillation detection is started again from step g0.

When it is determined in step g2 that the calculated value Jv1 is smaller than the threshold value Jv0, step g
6, the signal indicating that the control gains Ksp and Ksv do not oscillate is given to the selection means 32 and the operation management means 47. Based on this, the gain adjustment work is continued as it is, and the process proceeds to step g1 to evaluate the operation by the next control gain.

By providing the oscillation detecting device 50 as described above, the positional deviation Δx is substituted into the oscillation detection evaluation function Jv for calculation, and the calculated value Jv1 is a predetermined threshold value Jv.
When v0 or more, it can be determined that the robot 2 is in the oscillating state, and it is possible to prevent the robot 2 from being damaged by the oscillation when the control gain is automatically adjusted. Further, since the position deviation Δx used for oscillation detection is based on the actual position and does not include noise, no filtering process is required for the evaluation calculation.
Therefore, oscillation can be detected without requiring time-consuming data processing.

FIG. 14 is a graph showing the torque control value Tcom and the position deviation Δx. FIG. 14 (1) shows the torque control value at the time of excellent response, and FIG. 14 (2) shows the time at the time of poor response. 14 (3) shows the torque control value at the time of oscillation, FIG. 14 (4) shows the position deviation at the time of excellent response, and FIG. 14 (5) shows the torque control value at the time of poor response. Shows the position deviation,
FIG. 14 (6) shows the position deviation during oscillation. 14
In (1) to FIG. 14 (3), the horizontal axis represents time t (seconds), the vertical axis represents the torque control value Tcom (V), and FIG.
4 (4) to FIG. 14 (6), the horizontal axis represents time t (second).
And the vertical axis represents the position deviation Δx (V). As is apparent from FIG. 14, the torque control value Tcom (V) includes a noise component 53 in each state of excellent response, inferior response, and oscillation, and as described above for removing this noise component. It requires various filtering processes. On the other hand, the positional deviation Δx indicates a waveform that does not include the above noise component, and its removal operation is not required. Therefore, the above-mentioned excellent effects can be achieved.

Further, the oscillation detection evaluation function Jv has a positional deviation Δ
Since it is the root mean square value of x, it is possible to reliably detect the occurrence of oscillation without being affected by overshoot or the like.

FIG. 15 is a table showing calculated values of the evaluation function Jg for explaining the search operation of the control gains Ksp and Ksv according to another embodiment of the present invention. In this embodiment, the calculated value of the gain adjustment evaluation function when operating with a plurality of control gains selected in a predetermined step size within a predetermined range is obtained. Specifically, the control gain Ksp is set in increments of 8 (8Q6Ib / Eb) in the range of 0 to 64 (8Q6Ib / Eb), and the control gain Ksv is -48 to 0 (Ib × Vt / In the range of Eb), it is set in increments of 4 (Ib × Vt / Eb). This setting operation is performed by the selection means 32.

In this way, a plurality of values are set for the respective control gains Ksp and Ksv. This setting of the control gain corresponds to step E1 in FIG. 9 (5). Next, operate the robot in all combinations (step e
3), the twist amount θs and the twist speed θss in the operation are measured (corresponding to step e2) and given to the calculation unit 37 (corresponding to step e4). Next, as the operation corresponding to step e5, the calculation unit 37 obtains the calculated value of the gain adjustment function Jg when operating at the control gain of each combination. Next, as an operation corresponding to step e6, in the selection means 32, the control gains Ksp and Ksv that minimize the calculated value of the gain adjustment evaluation function Jg among the control gains of each combination are set to the optimum gain Ksp-op.
t, Ksv-opt. With this, among the possible values of the control gain, the control gain that can minimize the physical quantity, that is, the twist amount θ and the twist speed θss can be selected as the optimum gain.
It is possible to reliably achieve highly accurate control of a robot or the like. By setting and evaluating the control gain in such a brute force manner by the inventor of the present invention, a control gain showing a response equal to or better than that when the operator manually adjusts the control gain is obtained. It has been confirmed that the probability that can be set is 93.5%.

FIG. 16 is a graph for explaining the gain adjustment evaluation function Jga according to another embodiment of the present invention.
16 (1), the horizontal axis represents time t (seconds), the vertical axis represents the twisting speed θss (V) and the time weighting coefficient ng (t), and FIG. 16 (2) represents the horizontal axis at time t. (Seconds), and the vertical axis represents the position command θref (degrees). In the above-described embodiment, the evaluation function Jg that evaluates so that the twist amount θs and the twist speed θss are minimized is used.
As the gain adjustment evaluation function for adjusting the control gain, the evaluation function Jga represented by Expression (4) that evaluates only the twisting speed θss may be used.

[0080]

[Equation 3]

As shown in FIG. 16, the same variables as the gain adjustment evaluation function Jg are used as the variables defining the gain adjustment evaluation function Jga. Even if such a gain adjustment evaluation function is used, it is possible to achieve the same effect as the evaluation of the twisting speed θss as described above.

As another embodiment of the present invention, it may be used for adjusting the control gain and detecting oscillation of a multi-axis robot having two or four joint axes, or a machine tool other than the robot or a crane. It may also be used to adjust the control gain of an elevator or the like and detect oscillation. Further, the twist amount and the speed may be detected by providing a strain gauge on each of the arms 7 to 9 and an acceleration sensor at the tip of each of the arms 7 to 9 instead of the estimation by the observer. The physical quantity related to the gain adjustment may be a physical quantity other than the twist amount and the twist speed.

[0083]

As described above, according to the present invention, the position deviation between the position command value and the position during actual operation is substituted into the oscillation detection evaluation function for calculation, and the calculated value is a predetermined threshold value. When it is above, it is determined to be in an oscillating state. Since the position deviation is based on the actual position and does not include noise or low frequency signal components, no filtering process is required for the evaluation calculation. Therefore, oscillation can be detected without requiring time-consuming data processing. Further, by using the position deviation after the command angular position arrival time, the oscillation state of the robot can be determined without being affected by the position deviation before reaching the command angular position arrival time.

Further, according to the present invention, the oscillation of the robot can be detected during the calculation period in which the gain adjustment evaluation function is calculated. This makes it possible to determine whether the robot is in the oscillation state while automatically adjusting the control gain and prevent damage to the robot. Further, according to the present invention, since the oscillation detection evaluation function is the root mean square value of the position deviation from the time when the command value coincides with the target value, oscillation is generated without being affected by overshoot or the like. Can be reliably detected.

[Brief description of drawings]

FIG. 1 is a block diagram showing a control gain adjustment device 1 including a robot oscillation detection device 50 according to an embodiment of the present invention.

FIG. 2 is a system diagram showing a schematic configuration of an adjusting device 1.

FIG. 3 is a diagram showing a detected position value θm, an arm tip position θa, and a twist amount θs.

FIG. 4 is a graph showing a position command value θref, an arm tip position θa, and a twist amount θs.

FIG. 5: Torsional velocity θss and time weighting factor ng
It is a graph which shows (t).

FIG. 6 is a twist amount θs, a twist speed θss, and a position deviation Δ.
5 is a graph showing x and a position command value arrival time tj.

FIG. 7 is a graph showing a position command value θref, a position detection value θm, a twist amount θs, and an arm tip position θa.

FIG. 8 is a graph showing a position command value θref, a position detection value θm, a twist amount θs, and an arm tip position θa.

FIG. 9 is a flowchart showing a control gain adjusting operation.

FIG. 10 is a flowchart showing a specific example of the flow of FIG. 9 (5).

FIG. 11 is a graph showing a transition of control gain.

FIG. 12 is a graph showing a positional deviation Δx.

FIG. 13 is a flowchart showing an operation of oscillation detection.

FIG. 14 is a graph showing a torque control value Tcom and a position deviation Δx.

FIG. 15 is a table showing calculated values Jg for explaining a control gain searching operation according to another embodiment of the present invention.

FIG. 16 is a graph showing a twist speed θss, a time weighting coefficient ng (t), and a position command value θref for explaining a gain adjustment evaluation function Jga according to another embodiment of the present invention.

[Explanation of symbols]

1 Control gain adjustment device 2 robots 3 Control structure 4 Measuring structure 5 Arithmetic structure 6 bases 7-9 arms 10 articles 11 Action command means 12 Position control means 13 Speed control means 14 Position detection means 15 Differential calculation means 16 Observer 17, 18 Feedback calculation means 19-21 Subtraction means 22 Addition means 23 motor 25 Elastic deformation element 30 Gain adjustment function setting means 31 gain adjustment function calculation means 32 selection means 33 Oscillation detection function setting means 34 Oscillation detection function calculation means 35 Judgment means 37, 39 Operation part 39,40 storage 43 Twist acquisition means 44 Deviation acquisition means 45 time acquisition means 50 Oscillation detector L1 to L3 joint axes

Continuation of the front page (56) Reference JP-A-11-102211 (JP, A) JP-A-6-22591 (JP, A) JP-A-8-116688 (JP, A) JP-A-11-133038 (JP , A) (58) Fields investigated (Int.Cl. ⁷ , DB name) B25J 3/00-3/04 B25J 9/10-9/22 B25J 13/00-13/08 B25J 19/02-19 / 06 G05B 19/18-19/46 B23Q 15/00-15/28 B23Q 17/00-23/00 G05D 19/00-19/02

Claims (3)

(57) [Claims] 1. An oscillation detection function setting means for setting a predetermined oscillation detection evaluation function for judging oscillation based on a position deviation between a position command value and an actual operating position, and the position command value is a target value. When the command angle position arrival time is reached, the oscillation detection function calculation means for calculating the oscillation detection evaluation function is calculated. An oscillation detecting device for a robot, comprising: a determining unit that determines that the robot is in a state. 2. A state variable of a control system during actual operation is substituted, a gain adjustment evaluation function for evaluating a control gain is calculated, and a control gain that minimizes the calculated value is selected as an optimum gain. An oscillation detection device for a robot, which is provided in a control gain adjustment device for a robot, detects oscillation of the robot when a control gain is set and operates for evaluation, wherein the oscillation detection function calculation means is a gain adjustment device. 2. The oscillation detection device for a robot according to claim 1, wherein the oscillation detection evaluation function is calculated by substituting a position deviation in a period corresponding to a calculation period in which the evaluation function is calculated. 3. The oscillation detection device for a robot according to claim 1, wherein the oscillation detection evaluation function is a mean square value of the position deviation.