ROBITA (Real-world Oriented BImodal Talking Agent) is a robot developed by Waseda University’s Kobayashi Lab in 1999 to participate in natural conversation with a group of people. Up until ROBITA’s development, robots were typically designed for one-on-one communication only (one human operator and the robot). As more people are introduced into the conversation, the task becomes much more complicated. In order to hold a group conversation, ROBITA needs to know who is talking, what was said, how to respond to what was said, and when it is appropriate to do so. A robot ill-equipped to deal with this situation will speak out of turn or miss it’s cue to speak entirely. It goes without saying that these skills will be essential if robots are ever going to support people.
ROBITA uses its two head-mounted microphones for sound localization (“who is talking?”). It then uses its stereo CCD cameras for face recognition to determine whether the person speaking is looking away from, or directly at the robot (“am I being spoken to?”). It uses speech recognition to determine what has been said, and can express when it is ready to listen by making eye-contact or when it is confused through its actuated brows and mouth. Additional expressions are made possible by its two arms and fully articulated 5-fingered hands as seen in the photo below.
(left) “OK!”, (center) thinking, (right) “Aha!”
Waseda University Kobayashi Lab