Development and Evaluation of a Manually Guided Demonstration Tool for Robot Programming

Abstract

Learning from Demonstration is an intuitive technique for robot programming, which enables non-experts to quickly implement new robotic tasks. However, it is difficult to extract motion and force information by directly observing the human operator. Also so-called kinesthetic teaching, which records the motion and forces of the robotic arm, is difficult to be applied in an industrial environment. Therefore, there is a high demand for novel demonstration tools that simplify the programming process. In this thesis, a manually guided demonstration tool is developed that can be freely moved in the work space without being constrained by a robotic arm. The tool fulfills various requirements, which are pose tracking, ergonomic design including buttons and user feedback, integration of a 6D force-torque sensor, and software modules to interface the device with the programming framework. A prototype is evaluated regarding its capabilities to program contact-rich robotic tasks.