A Framework for Robot Programming via Imitation

Abstract

In this paper, a framework is described to program industrial robots through human task demonstrations. The framework utilizes the imitation of the human demonstrated motion for robot trajectory generation. The imitation principle is based on the kinematic model-based mapping strategy between the human arm and the robot. The framework also utilizes a state-of-the-art probabilistic model-based generalization architecture to eliminate inconsistencies in demonstrations and retrieve an executable trajectory for the robot. The implementation of the approach is shown for spray painting operation in a six-degrees-of-freedom industrial robot. The experimental results reflect that the framework is capable of generating accurate trajectories for complex applications.