Learning a pick-and-place robot task from human demonstration

Abstract

Programming by human demonstration is an intuitive methodology to program robots. Since pick-and-place tasks are common, this paper proposes a framework to learn robot pick-and-place tasks by human demonstration. To recognize human motion, we adopt Gilbreth's therbligs to represent it. For each therblig, this work uses a corresponding robot motion primitive to implement it. Additionally, this work uses XABSL to describe the motion primitives and the sequence of them. The experiment showed that the human stacked three blocks in an order and the robot learned to stack the blocks in the same order by human demonstration.