Teaching of assembly motion by demonstration: Artificial constrained motion primitives and its implementation using virtual small faces

"Teaching by demonstration" is a method to generate a robot program that makes a robot do the same task as the task which a human operator demonstrates. The author developed a "teaching by demonstration in VR" system which automatically generates a robot program to work in the real world after a task is demonstrated by an operator in the virtual world. This system deal with a task of controlling contact states between bodies like assembly task. The system generates a program which detects each contact state extracted from demonstration and realizes it by skilful motion primitives (SMP), and achieves the task. But when one body is not in contact with another, not only contact states but also other new states which can specify more general arrangement of the bodies are necessary in order to make the robot more skilful. Particularly, (1) specification of the arrangements of the bodies in a free space and (2) specification of the arrangements which should be achieved by preparation motion for a SMP, are important. In this paper, a virtual small face (VSF) and an artificial constrained motion primitive (ACMP) are introduced in order to specify these things and teach them to a robot by demonstration.