JPH06289918A - Method for controlling learning of robot - Google Patents

Abstract

PURPOSE:To provide the learning control method of a robot, by which the learning efficiency of the robot having a learning function can be improved. CONSTITUTION:The learning control method for correcting an input value supplied to the driving element of the robot based on an error occurring between a target orbit becoming the target of a reproduction operation at the time of causing the robot to repetitively execute the reproduction operation and the real orbit of the robot at the time of the reproduction operation is provided with a step p3 calculating the speed of the target orbit, p4 multiplying the speed error between the target orbit and the real orbit by a coefficient (learning gain) reduced in accordance with the lapse of time and adding a feed forward value supplied to the driving element of the robot at the time of the previous reproduction operation to the result and a step p5 setting the value obtained in the step p4 to the feed forward value for the driving element at the time of the subsequent reproduction operation.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a robot learning control method suitable for use in, for example, an industrial robot that performs repetitive reproduction operations.

[0002]

2. Description of the Related Art As is well known, various techniques for repetitively performing a replay operation of a robot have been developed. In this type of technology, for example, a learning control method described in the following document (a) is known. (A) S. Arimoto: Learning C
ontrol Theory for Robotic
Motion, International Jou
rnal of Adaptive Control
and Signal Processing, 4-
6,546 / 564, (1990)

In the learning control method described in this document (a), a negative feedback loop of position error and velocity error is individually configured for each degree of freedom of the robot, and each degree of freedom prepared in advance is set. The robot is regenerated according to the target trajectory of the corresponding joint. Then, the velocity error between the target trajectory at this time and the actual trajectory generated during the reproducing operation is measured, and this velocity error is multiplied by a constant (hereinafter referred to as a learning gain). Furthermore, this multiplication result is added to the input value (hereinafter referred to as feedforward value) given to the drive element of the robot at the time of the previous reproduction operation,
By providing the addition result as a new feedforward value to the drive element of the robot in the next reproducing operation, the accuracy of the reproducing operation is improved.

According to this learning control method, the actual trajectory of the robot is sequentially corrected while the reproducing operation is repeated several times.
It converges near the target trajectory prepared in advance.
For this reason, the dynamic effects such as centrifugal force and Coriolis force, which are observed in a robot having multiple degrees of freedom, are finally corrected, and high-speed and highly accurate trajectory control becomes possible. Further, theoretically, as the value of the learning gain is larger, the actual orbit converges to the vicinity of the target orbit with less reproduction operation.

[0005]

However, in the above-described conventional learning control method, when the learning gain by which the velocity error is multiplied is made too large, the actual trajectory is not changed due to the influence of the measurement error or the like occurring at the velocity detection at each joint. It vibrates as it approaches the end of the orbit. This is introduced, for example, by the following document (b). (B) Nanjo et al .: Influence of forgetting factor on robot trajectory control using learning, Proceedings of 9th Robotics Society of Japan, No. 1,391 / 392, (1991)

FIG. 3 shows a conventional learning control method.
This is applied to a robot having a degree of freedom, and shows an example of an actual trajectory when the reproducing operation is performed 20 times when the learning gain is too large. As shown in this figure, the starting point S of the trajectory
Therefore, the actual trajectory A2, which substantially coincides with the target trajectory A1, vibrates toward the end of the trajectory. That is,
In the conventional learning control method, the size of the learning gain is practically limited, and therefore many reproducing operations must be performed in order to converge the actual trajectory near the target trajectory, and thus the learning efficiency is poor. There was a drawback.

The present invention has been made under such a background, and an object thereof is to provide a learning control method for a robot which can improve the learning efficiency of a robot having a learning function.

[0008]

SUMMARY OF THE INVENTION In order to solve the above-mentioned problems, the present invention provides a target trajectory which is a target of the regenerating operation when the robot repeatedly performs the regenerating operation, and an actual robot at the time of the regenerating operation. In a learning control method for correcting an input value supplied to a drive unit of the robot based on an error generated between the trajectory and the trajectory, any one of a position error, a velocity error, and an acceleration error between the target trajectory and the actual trajectory. Or a second step of multiplying the error measured in the first step by a coefficient (learning gain) that decreases with the passage of time, a multiplication result of the second step and a previous step. Third addition of the input value (feedforward value) supplied to the drive unit during the reproduction operation
And a fourth step of supplying the addition result of the third step to the drive unit as an input value (feedforward value) for the next reproducing operation.

[0009]

According to the present invention, in the first step, any one of position error, velocity error, and acceleration error between the target trajectory and the actual trajectory is measured, and in the second step, the first step is performed. The measured error is multiplied by the coefficient (learning gain) that decreases with the passage of time, and in the third step, the multiplication result of the second step and the input value (feedforward value) supplied to the drive unit of the robot during the previous playback operation. ) And, in the fourth step, the third
The addition result of the step is supplied to the drive unit as an input value (feedforward value) for the next reproducing operation. As a result, if the feedforward value to the drive unit of the robot is corrected for each repetitive reproduction operation, the vibration of the actual trajectory that occurs near the end of the trajectory is suppressed.

[0010]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing the configuration of a control circuit used in an embodiment of the present invention. In this figure, only the control system for one degree of freedom of the electric motor drive type robot is shown, and the other degrees of freedom have the same configuration and are not shown.

In FIG. 1, reference numeral 4 denotes an arithmetic processing unit for performing various digital arithmetic operations, which is composed of a CPU (central processing unit), memories 1 to 3, etc., which are not shown. The arithmetic processing unit 4 includes an input / output interface circuit such as a D / A (digital / analog) conversion element, but the illustration is omitted here for simplicity.

Each of the memories 1 to 3 has a storage capacity enough to store at least the time T required to move the target trajectory divided by the control cycle T _S.
Then, the trajectory data θ of the target trajectory is stored in the memory 2.
_ref (t) is the target position θ of the motor 9 for each control cycle time T _S
Pre-stored as _ref (T _S i) (where i is an integer). The stored contents of the memories 1 and 3 and their uses will be described later.

The arithmetic processing section 4 is connected to a motor 9 for driving the robot via a comparator 5, a multiplier 6, a comparator 7 and an amplifier 8. Further, the motor 9 is connected to the comparator 7 via the rotation speed detector 11 and the multiplier 10, and is also connected to the comparator 5 via the rotation position detector 12.

The comparator 5 receives the control cycle time T _S from the memory 2.
Calculate the difference between the target trajectory data θ _ref (t) supplied for each and the actual trajectory θ (t) measured by the position detector 12,
This result is output. The multiplier 6 multiplies the output of the comparator 5 by the positive constant k ₁ and outputs the result. This multiplier 6
Is output to the amplifier 8 via the comparator 7.

On the other hand, the speed detector 11 measures the actual speed dθ / dt (hereinafter referred to as θ ^* (t)) of the motor 9 and outputs the measurement result. The multiplier 10 multiplies the actual speed θ ^* (t) measured by the speed detector 11 by the positive constant k ₂ and outputs the result. The output of the multiplier 10 is supplied to the amplifier 8 via the comparator 7 after the sign is inverted. Further, the amplifier 8 amplifies the output of the comparator 7 and outputs it to the motor 9. The above is the feedback control system in this control circuit.

With such a configuration, the motor 9 is driven in a direction in which the actual trajectory θ (t) has an error of “0” with respect to the target trajectory θ _ref (t). At this time, in order to secure the stability of the feedback control system, the values of the positive constants k ₁ and k ₂ cannot be extremely increased practically. Further, when the robot is operated at a high speed, the motor 9 cannot be made to completely follow the target trajectory θ _ref (t) due to the influence of the centrifugal force and the Coriolis force which are seen in the robot having multiple degrees of freedom. Therefore, in order to correct an error based on such a cause, feedforward control described below is performed in addition to the above feedback control.

Next, the feedforward control operation based on the operation of the CPU will be described with reference to the flow chart shown in FIG. In the figure, first, in step p1, the data u _k stored in the memory 1 is cleared to "0". Here, the data u _k is a feedforward value given to the drive element of the robot during the k-th reproduction operation. In this case, the feedforward value u ₁ is “0” because the reproduction operation is performed for the first time.

Next, when proceeding to step p2, in the feedback control system shown in FIG. 1, the feedforward value u ₁ in the memory 1 is supplied to the amplifier 8 every control cycle T _S ,
The first playback operation is performed. At this time, the velocity θ ^* over the entire real trajectory measured by the velocity detector 11 for each control cycle T _S.
Store (T _S i) in the memory 3. Then, when the reproduction operation of the first time is finished, the feedforward value in the memory 1 is corrected by the following steps p3 to p5.

At step p3, the target velocity θ ^* _ref (t) is calculated based on the target trajectory _θref (t) stored in the memory 2. That is, as shown in the following equation (1), the time t
The velocity θ ^* _ref (T _S i) at = T _S i can be calculated by dividing the difference between the target positions in the front and rear control cycles by the control cycle T _S. However, i is an integer representing the order of control cycles. θ ^* _ref (T _S i) = {θ _ref (T _S (i + 1))-θ _ref (T _S i)} / T _S ………… (1)

Next, when proceeding to step p4, the difference between the speed θ ^* _ref (t) calculated in step p3 and the actual speed θ ^* (t) stored in the memory 3 during the previous reproducing operation. Is calculated for each control cycle T _S , and this result is multiplied by the learning gain φ. Here, the learning gain φ is obtained by, for example, a function represented by the following equation (2) so that the learning gain φ has a smaller value as it approaches the end of the trajectory. However, φ ₀ and λ are positive constants, and e is a natural logarithm.

[Equation 1]

Next, in step p5, a value obtained by multiplying the learning gain φ by the feedforward value u in the memory 1 is used.
_The result of adding ₁ (t) and the new feedforward value u ₂
Store in memory 1 as (t). And step p6
Then, it is determined whether or not the learning operation is continued. That is, if the number of playback operations has not reached the preset number of learning times, or if the actual trajectory has not reached the preset desired accuracy, the determination result here is "Yes", It returns to the above-mentioned step p2.
As a result, the learning operation associated with the second reproduction operation is performed.

Next, in the second learning operation, the feedforward value u ₂ (t) newly stored in the memory 1 in the above step p5 is supplied to the amplifier 8. After that, while the determination result of the above step p6 is "Yes", the above steps p3 to p5 are repeated for each reproduction operation, and the feedforward value in the memory 1 is corrected each time. The calculation formula of the feedforward value at this time is expressed by the following formula (3).

[Equation 2]

In this way, when the number of reproduction operations reaches the preset number of learning times or when the actual trajectory reaches the desired accuracy, the determination result of step p6 is "N".
Then, the learning operation ends.

As described above, according to the present embodiment, the learning gain by multiplying the difference between the target trajectory and the actual trajectory is reduced toward the end of the trajectory, so that the vibration generated near the end of the trajectory is suppressed. can do.
As a result, the learning gain can be set larger than in the conventional method, and the actual trajectory can be converged to the vicinity of the target trajectory with fewer repetitive reproduction operations.

In this embodiment, the learning gain φ (t) in the above equation (3) is set by the function represented by the above equation (2), but a function value that decreases with the passage of time is obtained. Other functions can be used as long as they are available.

In the present embodiment, the target speed θ ^* _ref (t) is calculated based on _θref (t) stored in the memory 2, but instead, the target speed θ ^* _ref is calculated in advance. (t) may be stored in another memory. Further, the actual speed θ ^* (t) during the reproducing operation is not stored in the memory 3, but the feedforward value for each control cycle is directly calculated by the above equation (3), and the calculated value u _{k + 1} (t ) May be stored directly in the memory 1.

Further, in the present embodiment, the case where the computer (the arithmetic processing unit 4 composed of a CPU or the like) is used as the control circuit is taken as an example, but it may be replaced with other well-known various control elements. It is possible.

Further, in the present embodiment, a playback type robot in which each joint of the robot is driven by an electric motor and a negative feedback control system of the position error and the speed error is individually configured for the electric motor is targeted. Although described, the present invention is also applicable to a robot using a hydraulic cylinder as a drive source and a machine having various servo mechanisms.

Further, in the present embodiment, only the velocity error generated between the target trajectory and the actual trajectory of the robot is used in the calculation of the feedforward value, but the position error and the acceleration error are calculated with a coefficient that decreases with time. It is also possible to perform learning by adding the multiplied value to the feedforward value at the time of the previous reproduction operation.

Further, the present invention can be used in combination with the learning control method described in the above-mentioned document (a). That is, in this document (a), in order to mitigate the influence of measurement error, etc., as shown in the following equation (4), an appropriate positive constant (1-a) is added to the previous feedforward value u _k (t). A learning control method that multiplies by has been proposed and can be used in combination with the present invention. _{u k + 1 (t) =} φ (θ * ref (t) -θ * (t)) + (1-a) u k (t) .................. (4)

[0031]

As described above, according to the present invention, since the vibration of the actual orbit that occurs as the end of the orbit is approached is suppressed, the learning gain can be set larger than in the conventional case. It becomes possible to converge the actual orbit to the vicinity of the target orbit with fewer repeated reproduction operations. As a result, the learning efficiency of the robot can be improved.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a control circuit used in an embodiment of the present invention.

FIG. 2 is a flowchart showing a feedforward control operation based on an operation of a CPU in the embodiment.

FIG. 3 is a diagram showing an example of actual trajectories when a reproduction operation is performed 20 times when a conventional learning control method is applied to a robot having three degrees of freedom and a learning gain is excessively increased.

[Explanation of symbols]

 1 to 3 memory 4 arithmetic processing unit 5,7 comparator 6,10 multiplier 8 amplifier 9 motor 11 rotational speed detector 12 rotational position detector A1 target trajectory A2 actual trajectory

Claims (1)

[Claims] 1. Based on an error generated between a target trajectory which is a target of the reproducing operation and an actual trajectory of the robot during the reproducing operation when the robot repeatedly performs the reproducing operation,
A learning control method for correcting an input value supplied to a drive unit of the robot, which measures any one of a position error, a velocity error, and an acceleration error between the target trajectory and the actual trajectory.
A step, a second step of multiplying the error measured in the first step by a coefficient that decreases with the passage of time, a multiplication result of the second step, and an input value supplied to the drive unit at the time of the previous reproduction operation. And a fourth step of supplying the addition result of the third step to the drive unit as an input value for the next reproduction operation.