Machine learning apparatus, robot system, and machine learning method of learning state of tool
AssigneeFanuc Corporation
InventorsMikito Hane, Takashi Satou
Life of this patent
Estimating the term…
Abstract
A machine learning apparatus that can determine the state of a tool from a force applied from the tool to a robot while the robot performs a work using the tool. A machine learning apparatus for learning a state of a tool used for a work by a robot includes a learning data acquisition section that acquires, as a learning data set, data of a force applied from the tool to the robot while the robot causes the tool to perform a predetermined operation, and data indicating the state of the tool during the predetermined operation, and a learning section that generates a learning model representing a correlation between the force and the state of the tool, using the learning data set.
Claims14
A machine learning apparatus configured to learn a state of a tool used for a work by a robot, the apparatus comprising:
- a learning data acquisition section configured to acquire, as a learning data set,
- data of a force applied from the tool to the robot while the robot causes the tool to perform a predetermined operation, and
- label information indicating whether the state of the tool during the predetermined operation is an abnormal state in which the tool has damage or a normal state in which the tool has no damage;
- a learning section configured to generate a first learning model representing a correlation between the force and the state of the tool, using the learning data set;
- and
- a second learning model representing a correlation between image data of the tool and the abnormal state or the normal state of the tool,
- wherein the second learning model is configured to receive an input of the image data captured when the predetermined operation is performed, and output to the learning data acquisition section the label information corresponding to the input image data.
A machine learning method of learning a state of a tool used for a work by a robot, the method comprising, by a processor:
- acquiring, as a learning data set,
- data of a force applied from the tool to the robot while the robot causes the tool to perform a predetermined operation, and
- label information indicating whether the state of the tool during the predetermined operation is an abnormal state in which the tool has damage or a normal state in which the tool has no damage;
- generating a first learning model representing a correlation between the force and the state of the tool, using the learning data set;
- and
- controlling the operation of the robot in response to a result of determination for the state of the tool based on the generated first learning model,
- wherein a second learning model representing a correlation between image data of the tool and the abnormal state or the normal state of the tool receives an input of the image data captured when the predetermined operation is performed, and outputs the label information corresponding to the input image data.
Dependencies are inferred from the claim text.