Abstract
This paper demonstrates a prototype test bed for usability studies of a multi-modal language for directing robots, using a Sony AIBO and JULIAN, a grammar-based speech recognition engine. We developed a grammar for spoken commands based on Japanese in order to simplify command understanding and an event-driven system to interpret multi-modal directions. In our pilot studies, we confirmed our robot can properly react to spoken, non-verbal and multi-modal directions by humans. Results from recent experiments involving subjects unfamiliar with robots are also reported and discussed.
