Robot programming using the paradigm of learning by demonstration

This paper presents the appl ication of the paradigm of learning by demonstration for robot programming. Algorithms use bio-inspired techniques to extract relevant information accompanying the demonstrator’s action. A visuo-motor map relates visual inputs to motor commands necessary to imitate a behavior or a task. The system was evaluated qualitatively using a survey, and quantitatively by specifi c metrics to score the quality of the imitation of a group of four gestures. Thus, the learning by demonstration potential for robot programming is corroborated, since the system was able not only to make their own interpretations of the gestures to be taught, but to use the skills learned in conducting novel gestures.