CONSORTS: ARCHITECTURE FOR UBIQUITOUS AGENTS

CONSORTS (an architecture for COgNitive reSOurce management with physically-gRounding agenTS) is a new kind of architecture for ubiquitous agents. It is designed to realize mass user support in addition to conventional personal assistance. The key concepts in CONSORTS are "semantic grounding" and "cognitive resources". By using sensory information brought through a ubiquitous environment, agents have grounding to the physical world and they are conscious of physical resources ( especially spatio-temporal resources) in a cognitive way, i.e., they can recognize, reorganize, and operate raw physical resources as cognitive resources. Services realized in CONSORTS include (1) extension of conventional personal services using information about the physical world, such as position; and (2) mass user support that provides information and social coordination for mass users beyond personal support.

In the architecture, we assume that users have mobile information devices such as PDAs or cellular phones, and that users' positions are captured by sensors, such as cameras or wireless LANs. We also assume that their history of moving is tracked by sensors and registered in a spatio-temporal reasoner. Service agents provide situation-based services that use information about a user's position and moving history. One such situation-based service provides information according to a user's position. For example, when a user happens to get near an attraction that he might be interested in, a navigation agent gives the user directions on how to get there.

Mass user navigation consists of two parts . The first part is personal service , which directs users to their favorite places according to their intentions and preferences. In other words, it maximizes the number of places they want to visit, and it minimizes moving distance and time while obtaining needed guidance information. The second part is social coordination service , which tries to decrease congestion and total moving distance and time for all users by making plans for all the users by coordinating their intentions and preferences. Another important part in this architecture is a user model which describes the user, i.e., (1) Intention : Goals that the user should achieve in a period, such as a day; (2) Preference: Goals that the user expects to visit in the period; and (3) Attribute: A static description about the user that can be used to retrieve suitable information.