Experimental analysis of augmented reality interfaces for robot programming by demonstration in manufacturing

Augmented Reality (AR) technology has been effectively utilized to support various manual operations in the manufacturing industry. An important application is serving as a user interface for human-robot collaboration. This paper presents an experimental study on the feasibility of robot programming by demonstration (PbD) through AR interfaces in the context of manufacturing. Our focus is on comparing the pointing and line tracing processes using three input methods provided by an AR headset: hand ray, head gaze, and eye gaze, based on both objective and subjective measures obtained from the experiment. The hand ray method performs the best in terms of accuracy, precision, and completion time in most experimental conditions. The SUS and NASA-TLX scores indicate acceptable usability for the hand ray method but low usability for the others. A prototyping AR tool using the hand ray as non-contact input in the real world is developed for motion planning of an industrial robotic arm. A test case of tire mold welding verifies the feasibility of the AR tool while also showing its limited capability in precision manufacturing. This work demonstrates a new approach for robot PbD on tangible objects enabled by AR.