Industrial robot programming by demonstration

Abstract

With the increasing demands in robotized production, easy programming of industrial robots becomes a keen market requirement because the traditional machine-coding level programming methods are usually time consuming, complex to handle, and high technical skill demanding. In past years, the expectation in using artificial intelligence to make robot program to be easy has led many progresses in related technologies in this area. Among them, Programming by Demonstration (PbD) is outlined, by which the robot enables to learn from observing how human performs a task, and then translate as much information as possible from the demonstration into the movement of itself in an automatic way. It frees human from coding the robot through any machine language level programming. In this work, a framework for industrial robot programming by demonstration is introduced where the perception of assembly tree, action identification and segmentation, skill sequence formulation, and robot movement mapping and optimization are included. The encouraging results in the final experiments with LEGO blocks in RobotStudio™, an industrial robot simulation environment from ABB, verify the feasibility of the proposed PbD solution.