Robot control device, robot system, and robot control method

A robot control method using a learned model that denotes a model created through learning work data composed of input data and output data, the input data including states of a robot and surroundings of the robot in a case where a human operates the robot to cause the robot to perform a series of works, the output data including a human operation corresponding to the case or a movement of the robot caused by the human operation, the robot control method comprising:

• a control data acquisition step of acquiring control data on the robot used to make the robot perform the works, by obtaining output data related to a human operation or a movement of the robot from the learned model, the human operation or the movement of the robot being presumed in response to and in accordance with input data received by the learned model, the input data being related to states of the robot and surroundings of the robot;
• a completion rate acquisition step of acquiring a completion rate indicating to which progress level in the series of works the output data outputted by the learned model corresponds;
• and
• a certainty factor acquisition step of acquiring a certainty factor indicating a probability of the human operation or the movement of the robot being presumed in a case where the learned model outputs the output data in response to reception of the input data, wherein:
• control data on the robot used to make the robot perform the works is acquired based on the output data outputted by two or more learned models that are created each corresponding to each of different series of works,
• the robot is controlled such that the robot performs coupled works in which the different series of works are chronologically coupled, by successively performing controls of the robot based on the output data outputted respectively by the two or more learned models, and
• a learned model of the two or more learned models is switched on a condition that the certainty factor is a predetermined level or higher.

A robot control device comprising:

• a learned model created through learning work data composed of input data and output data, the input data including states of a robot and surroundings of the robot in a case where a human operates the robot to cause the robot to perform a series of works, the output data including a human operation corresponding to the case or a movement of the robot caused by the human operation;
• and
• a processor that is configured to:
• acquire control data on the robot used to make the robot perform the works, the acquisition being made by obtaining output data related to a human operation or a movement of the robot from the learned model, the human operation or the movement of the robot being presumed in response to and in accordance with input data received by the learned model, the input data being related to states of the robot and surroundings of the robot;
• acquire a completion rate indicating to which progress level in the series of works the output data outputted by the learned model corresponds;
• and
• acquire a certainty factor indicating a probability of the human operation or the movement of the robot being presumed in a case where the learned model outputs the output data in response to reception of the input data, wherein:
• the learned model is configured to express a state transition, and configured to output a movement corresponding to each state,
• the processor is configured to acquire a soundness factor that is a numerical value of an evaluation of the robot's behavior based on an evaluation given by the user,
• the processor is configured to output a movement log, the movement log including the input data received by the learned model and data related to a movement of the robot based on the output data outputted by the learned model,
• with respect to a movement of the robot described in the movement log, the user is allowed to give an evaluation for each movement corresponding to a part of the series of works, and
• based on the evaluation given by the user, the processor adjusts the soundness factor corresponding to the state in the learned model.

A robot control method using a learned model that denotes a model created through learning work data composed of input data and output data, the input data including states of a robot and surroundings of the robot in a case where a human operates the robot to cause the robot to perform a series of works, the output data including a human operation corresponding to the case or a movement of the robot caused by the human operation, the robot control method comprising:

• a control data acquisition step of acquiring control data on the robot used to make the robot perform the works, by obtaining output data related to a human operation or a movement of the robot from the learned model, the human operation or the movement of the robot being presumed in response to and in accordance with input data received by the learned model, the input data being related to states of the robot and surroundings of the robot;
• a completion rate acquisition step of acquiring a completion rate indicating to which progress level in the series of works the output data outputted by the learned model corresponds;
• and
• a certainty factor acquisition step of acquiring a certainty factor indicating a probability of the human operation or the movement of the robot being presumed in a case where the learned model outputs the output data in response to reception of the input data, wherein:
• the learned model is configured to express a state transition, and configured to output a movement corresponding to each state,
• a soundness factor that is a numerical value of an evaluation of the robot's behavior based on an evaluation given by the user is acquired,
• a movement log is output, the movement log including the input data received by the learned model and data related to a movement of the robot based on the output data outputted by the learned model,
• with respect to a movement of the robot described in the movement log, the user is allowed to give an evaluation for each movement corresponding to a part of the series of works, and
• based on the evaluation given by the user, the soundness factor is adjusted corresponding to the state in the learned model.