Article stacking apparatus and machine learning apparatus

Patent metadata

Field	Value
Publication number	US-11084168-B2
Application number	US-201816104682-A
Country	US
Kind code	B2
Filing date	Aug 17, 2018
Priority date	Aug 23, 2017
Publication date	Aug 10, 2021
Grant date	Aug 10, 2021

Abstract

A machine learning apparatus of an article stacking apparatus observes, as a state variable representing the current environmental state, stacking status data indicating the stacking status of an article stacking area and article information data indicating information on an article to be stacked, and acquires, as label data, article placement data indicating a placement of the article in the stacking area. The machine learning apparatus learns article placement data in association with the stacking status data and the article information data using the state variable and the label data.

First claim

The invention claimed is:

1. An article stacking apparatus for controlling a robot to stack a plurality of articles from a setting area into a stacking area, wherein the plurality of articles were previously set in the setting area, the article stacking apparatus comprising:
a machine learning apparatus for determining an estimate placement to stack an article, of the plurality of articles, into the stacking area by the robot, wherein, during a learning phase, the plurality of articles were previously set in a setting area and moved by a worker to the stacking area, said machine learning apparatus including:
a label data acquisition section for acquiring label data of the article to be stacked, the label data including article placement data indicating a past placement of the article in the stacking area based on previously collected data, wherein the previously collected data is obtained while the article was previously set in the setting area,
a state observation section for observing a state variable representing a current state of the plurality of articles set in the setting area to be stacked by the robot, and an environmental state of the stacking area, the state variable including:
article property data indicating properties of the plurality of articles in the setting area to be stacked by the robot, and
stacking status data indicating a stacking status of the stacking area, the stacking status including stacking information of articles already placed in the stacking area,

a learning section for learning a correlation between: (i) the label data obtained in the label acquisition section, and (ii) the state variable including the article property data and the stacking status data, and
an estimation result output section configured to determine the estimate placement of the article, of the plurality of articles, to be stacked by the robot into the stacking area based on the correlation of the label data and the state variable learned by the learning section,

wherein the article stacking apparatus is configured to control the robot to stack the article, of the plurality of articles, into the stacking area based on the determined estimate placement.

2. The article stacking apparatus according to claim 1 , wherein the article placement data further indicates a temporary placement of the article.

3. The article stacking apparatus according to claim 1 , wherein the learning section includes
an error calculation section for calculating an error between an output of a correlation model and a correlation feature, the correlation model deriving the article placement data from the stacking status data and the article property data, the correlation feature being identified from teacher data prepared in advance, and
a model update section for updating the correlation model to reduce the error.

4. The article stacking apparatus according to claim 1 , wherein the learning section calculates the state variable and the label data using a multi-layered structure.

5. The article stacking apparatus according to claim 1 , wherein the estimation result output section is configured to output a result of the determination of the estimate placement of the article in the stacking area.

6. The article stacking apparatus according to claim 1 , wherein the machine learning apparatus is on a cloud server.

7. The article stacking apparatus according to claim 1 , wherein the estimation result output section is configured to determine the estimate placement of the article using the state variable and the label data acquired by the label data acquisition section and the state observation section from operations performed by workers with a plurality of article stacking apparatuses.

8. A machine learning apparatus for determining an estimate placement for a robot to stack an article, of a plurality of articles, into a stacking area, wherein during a learning phase, the plurality of articles were previously set in a setting area and moved by a worker to the stacking area, the machine learning apparatus comprising:
a label data acquisition section for acquiring label data of the article to be stacked, the label data including article placement data indicating a past placement of the article in the stacking area based on previously collected data, wherein the previously collected data is obtained while the article was previously set in the setting area,
a state observation section for observing a state variable representing a current state of the plurality of articles set in the setting area to be stacked by the robot, and an environmental state of the stacking area, the state variable including:
article property data indicating properties of the plurality of articles in the setting area to be stacked by the robot, and
stacking status data indicating a stacking status of the stacking area, the stacking status including stacking information of articles already placed in the stacking area,

a learning section for learning a correlation between: (i) the label data obtained in the label acquisition section, and (ii) the state variable including the article property data and the stacking status data, and
an estimation result output section configured to determine the estimate placement of the article, of the plurality of articles, to be stacked by the robot into the stacking area based on the correlation of the label data and the state variable learned by the learning section.

Assignees

• Fanuc Corp

Inventors

• Namiki Yuta

Classifications

• G05B2219/40006

  Placing, palletize, un palletize, paper roll placing, box stacking · CPC title

• B65G57/00

  Stacking of articles (B65G60/00 takes precedence; feeding, piling or stacking sheets B65H) · CPC title

• B25J15/00

  Gripping heads {and other end effectors (grippers used in machine tools B23Q7/04; gripping members fitted on cranes B66C1/42, B66C1/44; gripping means used for mounting electrical components H05K13/04; gripping means used in the manufacture of semiconductors H10P72/7602)} · CPC title

• B65G57/02

  by adding to the top of the stack · CPC title

• B25J9/1692

  Calibration of manipulator · CPC title