Shovel and shovel assist system

Patent metadata

Field	Value
Publication number	US-12320096-B2
Application number	US-202117241261-A
Country	US
Kind code	B2
Filing date	Apr 27, 2021
Priority date	Oct 31, 2018
Publication date	Jun 3, 2025
Grant date	Jun 3, 2025

Abstract

A shovel includes a processor, and a memory storing program instructions that cause the processor to obtain environmental information around the shovel, and perform determination related to an object around the shovel based on the obtained environmental information, by using a learned model on which machine learning has been performed. The learned model is updated to an additionally learned model on which additional learning has been performed based on teaching information generated from the obtained environmental information. In a case where the learned model is updated, the processor performs the determination based on the obtained environmental information, by using the updated learned model.

First claim

What is claimed is:

1. A shovel comprising:
an actuator;
a processor; and
a memory storing program instructions that cause the processor to:
obtain environmental information around the shovel; and
perform determination related to an object around the shovel based on the obtained environmental information, by using a learned model on which machine learning has been performed, the object including a person;
wherein the learned model is updated to an additionally learned model on which additional learning has been performed based on teaching information generated from the obtained environmental information;
wherein, in a case where the learned model is updated, the processor performs the determination based on the obtained environmental information, by using the updated learned model,
wherein the actuator is controlled based on the determination,
wherein a plurality of shovels including the shovel include the identical learned model and perform the determination by using the learned model,
wherein the learned model included in each of the plurality of shovels is updated to the additionally learned model, and
wherein the teaching information is generated from additional environmental information selected among the obtained environmental information based on a result of the determination.

2. The shovel as claimed in claim 1 , wherein the learned model is updated to the additionally learned model on which the additional learning has been performed based on the teaching information generated from the environmental information obtained by at least one of the shovel or another shovel different from the shovel.

3. The shovel as claimed in claim 1 , wherein the learned model is updated to the additionally learned model on which the additional learning has been performed based on the teaching information generated from the environmental information around the plurality of shovels, the environmental information being obtained at the plurality of shovels.

4. The shovel as claimed in claim 1 ,
wherein the instructions cause the processor to further record the environmental information obtained at a predetermined timing, and
wherein the learned model is updated to the additionally learned model on which the additional learning has been performed based on the teaching information generated from the recorded environmental information.

5. The shovel as claimed in claim 4 , wherein the predetermined timing is when the shovel turns or the shovel travels.

6. The shovel as claimed in claim 4 ,
wherein the processor performs the determination related to detection of the object around the shovel, and
wherein the predetermined timing is when the processor determines that the object around the shovel is detected.

7. The shovel as claimed in claim 1 , further comprising:
a controller configured to control the actuator based on the determination.

8. The shovel as claimed in claim 1 , further comprising:
an operating device,
wherein the actuator is driven based on an operation performed on the operating device, and
wherein, in a case where it is determined that the object is present within a predetermined range around the shovel based on the determination, the actuator is not driven even if the operating device is operated.

9. The shovel claimed in claim 1 , wherein the processor performs the determination related to a motion content of the object by using the learned model based on the environmental information obtained sequentially in time.

10. The shovel as claimed in claim 1 , wherein the teaching information is generated by identifying that the result of the determination performed by the learned model based on the environmental information is incorrect.

11. The shovel as claimed in claim 1 , wherein the teaching information is generated from the additional environmental information selected among the obtained environmental information by an operator of the shovel based on the result of the determination.

12. The shovel as claimed in claim 1 , wherein the determination relates to states of the person, the states including sitting, standing, or lying.

13. The shovel as claimed in claim 1 , wherein the determination relates to a state of a part of the person.

14. A shovel assist system, comprising:
a first shovel;
a second shovel; and
an external device configured to communicate with the first shovel and the second shovel,
wherein the first shovel includes
an actuator;
a processor; and
a memory storing program instructions that cause the processor to:
obtain environmental information around the first shovel; and
perform determination related to an object around the first shovel based on the obtained environmental information around the first shovel, by using a learned model on which machine learning has been performed, the object including a person;

wherein the second shovel includes
a processor; and
a memory storing program instructions that cause the processor to:
obtain environmental information around the second shovel;
record the obtained environmental information around the second shovel; and
transmit the recorded environmental information to the external device; and

wherein the external device includes
a processor; and
a memory storing program instructions that cause the processor to:
generate teaching information based on the environmental information around the second shovel, received from the second shovel;
perform additional learning on a learned model identical to the learned model used to perform the determination at the first shovel, based on the generated teaching information, to generate an additionally learned model; and
transmit, to the first shovel and the second shovel, the additionally learned model on which the additional learning has been performed,

wherein the learned model is updated to the additionally learned model received from the external device at the first shovel,
wherein the determination is performed, in a case where the learned model is updated, by using the updated learned model, based on the obtained environmental information around the first shovel,
wherein the actuator is controlled based on the determination at the first shovel,
wherein the learned model is updated to the additionally learned model received from the external device at the second shovel, and
wherein the teaching information is generated from additional environmental information selected among the obtained environmental information based on a result of the determination.

15. The shovel assist system claimed in claim 14 , wherein the first shovel is the second shovel.

16. The shovel assist system claimed in claim 14 , comprising a plurality of said first shovels, wherein the external device communicates with the plurality of first shovels.

17. The shovel assist system claimed in claim 14 , comprising a plurality of said second shovels, wherein the external device communicates with the plurality of second shovels.

18. The shovel assist system claimed in claim 14 , wherein the processor performs the determination related to a motion content of the object by using the learned model based on the environmental information obtained sequentially in time.

19. A machine learning device comprising:
a processor; and
a memory storing program instructions that cause the processor to:
perform machine learning on a learning model to generate a learned model based on a data set, the data set being a combination of environmental information in a work site where a shovel is disposed and a correct answer t

Assignees

• Sumitomo Construction Machinery Co Ltd

Inventors

• Kurosawa Ryota

Classifications

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06T7/0004

  Industrial image inspection · CPC title

• E02F9/26

  Indicating devices {(E02F5/145 takes precedence)} · CPC title

• E02F9/24

  Safety devices {, e.g. for preventing overload (E02F9/226 takes precedence)} · CPC title