Robot control device, robot system, and robot control method

1. A robot control device comprising:

a trained model which is built by being trained on work data when a human operates a robot so that the robot performs a series of operations, the work data including input data and output data, the input data being the state of the robot and its surroundings, the output data being the corresponding human operation or the operation of the robot by the human operation;

a control data acquisition section which acquires control data of the robot to make the robot perform the work, in the case the input data concerning the state of the robot and its surroundings is input to the trained model, by acquiring output data concerning the human operation or the operation of the robot predicted accordingly from the trained model;

base trained models, each of the base trained models being built by being trained on work data when a human operates the robot so that the robot performs an operation decomposed from the series of operations, and being built for each of a plurality of operations decomposed from the series of operations, the work data including input data and output data, the input data being the state of the robot and its surroundings, the output data being the corresponding human operation or the operation of the robot by the human operation;

an operation label storage section which stores operation labels, each of the operation labels including information expressing the operation and being stored in correspondence with the base trained model;

a base trained model combination information acquisition section which acquires combination information when the trained model is represented by a combination of a plurality of the base trained models, by acquiring a similarity between the trained model and the respective base trained models; and

an information output section that outputs the operation label corresponding to each of the plurality of the base trained models which are combined to represent the trained model based on the combination information.

2. The robot control device according to claim 1 , wherein

the base trained model combination information acquisition section acquires a relationship which indicates chronologically earlier or later of the base trained models when the trained model is represented by a combination of the plurality of the base trained models, and

the information output section outputs the chronological relationship of the plurality of the base trained models using the operation labels.

3. The robot control device according to claim 1 , wherein

when the input data is input to the trained model, the trained model outputs the corresponding output data and is capable of outputting a degree of progress indicating to which degree of progress of the series of operations the output data corresponds, and

the base trained model combination information acquisition section is capable of acquiring the relationship which indicates chronologically earlier or later of the base trained models based on the degree of progress.

4. The robot control device according to claim 1 , wherein

the trained model and the base trained models are capable of clustering the work data to be trained according to the features, and

the base trained model combination information acquisition section determines a similarity between the trained model and the respective base trained model by comparing a result of the clustering in the trained model and a result of the clustering in the base trained model.

5. The robot control device according to claim 1 , wherein

the base trained model combination information acquisition section determines a similarity between the trained model and the respective base trained model by comparing data output by the trained model and the base trained model for the same input data.

6. The robot control device according to claim 1 , comprising:

a comparison section, when the data of the state of the robot and its surroundings which is included in untrained work data is input to the trained model, acquiring a degree of concordance by comparing the output data output by the trained model and the output data included in the untrained work data; wherein

the information output section outputs the degree of concordance in relation to a time series of the operation of the robot.

7. The robot control device according to claim 6 , wherein

the untrained work data is generated in real-time, as the result of operating the robot by a human so that the robot performs at least a part of the series of operations.

8. The robot control device according to claim 1 , wherein

by being specified a chronological position of an operation among the series of the operations, the trained model is capable of deleting a part corresponding to the specified operation in the trained model.

9. The robot control device according to claim 1 , wherein

by being specified a chronological position of an operation among the series of the operations and additional work data for training an operation to replace, the trained model is capable of replacing a part corresponding to the specified operation in the trained model with the result of training on the additional work data.

10. The robot control device according to claim 1 , wherein

when the trained model outputs the output data in response to the input of the input data, the trained model is capable of outputting a degree of confidence that indicates certainty of the prediction, and

the information output section outputs the degree of confidence along with the operation label.

11. A robot system comprising:

the robot control device according to claim 1 ; and

the robot.

12. A robot control method using a trained model which is built by being trained on work data when a human operates a robot so that the robot performs a series of operations, the work data including input data and output data, the input data being the state of the robot and its surroundings, the output data being the corresponding human operation or the operation of the robot by the human operation,

and using base trained models, each of the base trained models being built by being trained on work data when a human operates the robot so that the robot performs an operation decomposed from the series of operations, and being built for each of a plurality of operations decomposed from the series of operations, the work data including input data and output data, the input data being the state of the robot and its surroundings, the output data being the corresponding human operation or the operation of the robot by the human operation,

the method comprising:

a control data acquisition step for acquiring control data of the robot to make the robot perform the work, in the case the input data concerning the state of the robot and its surroundings is input to the trained model, by acquiring output data concerning the human operation or the operation of the robot predicted accordingly from the trained model;

an operation label storage step for storing operation labels, each of the operation labels including information expressing the operation and being stored in correspondence with the base trained model;

a base trained model combination information acquisition step for acquiring combination information when the trained model is represented by a combination of a plurality of the base trained models, by acquiring a similarity between the trained model and the respective base trained models; and

an information output step for outputting the operation label corresponding to each of the plurality of the base trained models which are combined to represent the trained model based on the combination information.