Building a Bridge Between AI and Robotics

About f i f teen years ago, the hand eye system was one of the e x c i t i n g research t o p i c s in a r t i f i c i a l i n t e l l i g e n c e . Several AI groups attempted to develop i n t e l l i g e n t robots by combining computer vision with computer control led manipulator. However, after the success of early prototypes, research e f f o r t s have been s p l i t t e d into general intel l igence research and real world oriented robotics research. Currently, the author fee ls there ex is ts a s i gn i f i can t gap between AI and r o b o t i c s in s p i t e of the necess i ty of communication and in teg ra t ion . Thus, without bu i l d ing a bridge over the gap, AI w i l l lose a f e r t i l e research f i e l d that needs real-time rea l world i n t e l l i g e n c e , and robot ics w i l l never acquire intel l igence even if i t works s k i l l f u l l y . In t h i s paper, I would l ike to encourage AI community to promote more e f f o r t s on rea l world robot ics , wi th the discussions about key points for the study. This paper also introduces current steps of our robot ics research that attempt to connect perception with action as i n t e l l i g e n t l y as possible.