Machine learning device, servo motor control device, servo motor control system, and machine learning method

Patent metadata

Field	Value
Publication number	US-10418921-B2
Application number	US-201816021447-A
Country	US
Kind code	B2
Filing date	Jun 28, 2018
Priority date	Jul 18, 2017
Publication date	Sep 17, 2019
Grant date	Sep 17, 2019

Abstract

A machine learning device is configured to perform machine learning with respect to a servo motor control device including a non-linear friction compensator that creates a compensation value with respect to non-linear friction on the basis of a position command, the machine learning device including: a state information acquisition unit configured to acquire state information including a servo state including position error, and combination of compensation coefficients of the non-linear friction compensation unit, by causing the servo motor control device to execute a predetermined program; an action information output unit configured to output action information including adjustment information of the combination of compensation coefficients; a reward output unit configured to output a reward value in reinforcement learning, based on the position error; and a value function updating unit configured to update an action value function on the basis of the reward value, the state information, and the action information.

First claim

What is claimed is:

1. A machine learning device configured to perform machine learning with respect to a servo motor control device comprising a non-linear friction compensation unit configured to create a compensation value with respect to non-linear friction on the basis of a position command, the machine learning device comprising:
a state information acquisition unit configured to acquire state information from the servo motor control device by causing the servo motor control device to execute a predetermined program, the state information including a servo state including at least position error, and a combination of compensation coefficients of the non-linear friction compensation unit;
an action information output unit configured to output action information including adjustment information of the combination of compensation coefficients included in the state information, to the servo motor control device;
a reward output unit configured to output a reward value in reinforcement learning, based on the position error included in the state information; and
a value function updating unit configured to update an action value function on the basis of the reward value output by the reward output unit, the state information, and the action information.

2. The machine learning device according to claim 1 , wherein the reward output unit outputs the reward value on the basis of an absolute value of the position error.

3. The machine learning device according to claim 1 , wherein the servo motor control device further comprises a velocity feedforward calculation unit configured to create a velocity feedforward value on the basis of the position command, and the non-linear friction compensation unit is connected in parallel to the velocity feedforward calculation unit.

4. The machine learning device according to claim 1 , wherein the machine learning device further includes an optimizing action information output unit configured to generate and output combination of compensation coefficients of the non-linear friction compensation unit, on the basis of a value function updated by the value function updating unit.

5. A servo motor control system comprising: the machine learning device according to claim 1 ; and a servo motor control device that comprises a non-linear friction compensation unit configured to create a compensation value with respect to non-linear friction.

6. The servo motor control system according to claim 5 , wherein the servo motor control device further comprises a velocity feedforward calculation unit configured to create a velocity feedforward value on the basis of a position command, and the non-linear friction compensation unit is connected in parallel to the velocity feedforward calculation unit.

7. A servo motor control device comprising: the machine learning device according to claim 1 ; and a non-linear friction compensation unit configured to create a compensation value with respect to non-linear friction.

8. The servo motor control device according to claim 7 , wherein the servo motor control device further comprises a velocity feedforward calculation unit configured to create a velocity feedforward value on the basis of a position command, and the non-linear friction compensation unit is connected in parallel to a velocity feedforward calculation unit.

9. A machine learning method of a machine learning device configured to perform machine learning with respect to a servo motor control device, comprising a non-linear friction compensation unit configured to create a compensation value with respect to non-linear friction on the basis of a position command, the machine learning method comprising:
acquiring state information from the servo motor control device by causing the servo motor control device to execute a predetermined program, the state information including a servo state including at least position error, and a combination of compensation coefficients of the non-linear friction compensation unit;
outputting action information including adjustment information of the combination of compensation coefficients included in the state information, to the servo motor control device; and
updating action value function on the basis of a reward value in reinforcement learning, the state information, and the action information based on the position error included in the state information.

Assignees

• Fanuc Corp

Inventors

• Oho Yusaku
• Sonoda Naoto

Classifications

• G06N3/045

  Combinations of networks · CPC title

• G05B13/0265

  the criterion being a learning criterion · CPC title

• G06N3/08

  Learning methods · CPC title

• G05B1/00

  Comparing elements, i.e. elements for effecting comparison directly or indirectly between a desired value and existing or anticipated values · CPC title

• H02P23/0031

  implementing a off line learning phase to determine and store useful data for on-line control · CPC title