Host: Japan Society for Fuzzy Theory and Intelligent Informatics
We are studying a robot vision system that has sleep and wake functions. Our proposed mathematical Activation-Input-Modulation (AIM) model, that can express consciousness states such as waking or sleep, controls a ratio between external and internal information processing. The purpose of this paper is to design a new memory architecture like a human's brain for an intelligent real-time robot vision system that can store environmental information in its memories efficiently. The system is equipped with working, short-term and long-term memories. Each memory is divided into explicit and implicit memories. The working memory stores visual information for a few seconds. Information stimulated by external sensors is stored in an explicit short-term memory. Information without a stimulus is stored in an implicit short-term memory periodically. Information organized during sleeping is stored explicit and implicit long-term memories. Operations of these memories, which run in parallel, are also controlled by the AIM model.
