Skill transfer mechanical apparatus

A computer readable non-transitory storage medium that stores a program, the program including a learning model, the program when executed causing a computer with a memory to control operation of an operating part that controls a working part and execute steps comprising:

• outputting a basic motion instruction for causing the working part to carry out a basic motion by the operating part in accordance with the program;
• receiving, from an inputter, a manual motion correcting instruction according to an operation input by the operator on the inputter;
• receiving, from a detector, motion information regarding the operation of the operating part corresponding to the motion of the working part, the motion information comprising force data indicative of a force applied to a work environment by the working part, and position data indicative of a position of the working part during operation;
• inputting the motion information to the learning model and outputting an automatic motion correcting instruction by using an output result of the learning model;
• controlling the operation of the operating part to cause the operating part to move the working part in accordance with the manual motion correcting instruction and the automatic motion correcting instruction;
• adding the manual motion correcting instruction to the automatic motion correcting instruction to generate a combined motion correcting instruction;
• and
• storing the combined motion correcting instruction and the motion information in the memory,
• wherein by machine learning, the program updates the learning model by causing the learning model to carry out the machine learning of the combined motion correcting instruction stored in the memory by using the motion information stored in the memory, and
• wherein after the machine learning of the learning model, the program inputs the motion information to the updated learning model during the operation of the operating part, and outputs an updated version of the automatic motion correcting instruction by using an output result of the updated learning model.

A method of controlling a computer with a memory that stores a program to control operation of an operating part that controls a working part, the method comprising:

• outputting a basic motion instruction for causing the working part to carry out a basic motion by the operating part in accordance with the program;
• receiving, from an inputter, a manual motion correcting instruction according to an operation input by the operator on the inputter;
• receiving, from a detector, motion information regarding the operation of the operating part corresponding to the motion of the working part, the motion information comprising force data indicative of a force applied to a work environment by the working part, and position data indicative of a position of the working part during operation;
• inputting the motion information to a learning model and outputting an automatic motion correcting instruction by using an output result of the learning model;
• controlling the operation of the operating part to cause the operating part to move the working part in accordance with the manual motion correcting instruction and the automatic motion correcting instruction;
• adding the manual motion correcting instruction to the automatic motion correcting instruction to generate a combined motion correcting instruction;
• and
• storing the combined motion correcting instruction and the motion information in the memory,
• wherein by machine learning, the program updates the learning model by causing the learning model to carry out the machine learning of the combined motion correcting instruction stored in the memory by using the motion information stored in the memory, and
• wherein after the machine learning of the learning model, the program inputs the motion information to the updated learning model during the operation of the operating part, and outputs an updated version of the automatic motion correcting instruction by using an output result of the updated learning model.

A method of controlling a computer with a memory that stores a program to control operation of an operating part that controls a working part, the computer being integrated with a mechanical system including machinery comprising a machinery body as the operating part and having an end effector part as the working part and configured to move the end effector part so that the end effector part performs work on a work target object, the computer being configured to control operation of the machinery body, the machinery further including an inputter, the method comprising:

• outputting a basic motion instruction for causing the end effector to carry out a basic motion by the operating part in accordance with the program;
• receiving, from the inputter, a manual motion correcting instruction according to an operation input by the operator on the inputter;
• receiving, from a detector integrated with the machine body, motion information regarding the operation of the operating part corresponding to the motion of the working part, the motion information comprising force data indicative of a force applied to a work environment by the working part, and position data indicative of a position of the working part during operation;
• inputting the motion information to a learning model and outputting an automatic motion correcting instruction by using an output result of the learning model;
• controlling the operation of the operating part to cause the operating part to move the working part in accordance with the manual motion correcting instruction and the automatic motion correcting instruction;
• adding the manual motion correcting instruction to the automatic motion correcting instruction to generate a combined motion correcting instruction;
• and
• storing the combined motion correcting instruction and the motion information in the memory,
• wherein by machine learning, the program updates the learning model by causing the learning model to carry out the machine learning of the combined motion correcting instruction stored in the memory by using the motion information stored in the memory,
• wherein after the machine learning of the learning model, the program inputs the motion information to the updated learning model during the operation of the operating part, and outputs an updated version of the automatic motion correcting instruction by using an output result of the updated learning model, and
• causing the machinery body to move the end effector part according to the automatic motion correcting instruction, and the manual motion correcting instruction,
• detecting, with the detector, the motion information of the end effector part during the operation of the machinery body, and outputting the motion information,
• storing the motion information of the end effector part in the memory,
• selectively connecting and disconnecting a communication path of the manual motion correcting instruction from the inputter to the computer, and
• in response to disconnection of the communication path, accepting training of the operation of the operator by using the inputter.