Philosophy | Seminar in Logical Theory
P750 | 3393 | Barwise

Topic: Situated and Hybrid Reasoning Systems

Imagine building a simple robot whose mission in life is to deliver coffee to
people in a given building, using a video camera and a laptop computer as parts
of this robot.  (This describes an actual robot at the University of Toronto.)
Information learned about the surroundings is typically recorded in terms of
probabilities on a state space, while high level knowledge about who likes cream
in their coffee, who is in what office, and so forth is recorded in a high-level
symbolic language.  The problem arises as how to integrate these two very
different kinds of information.

In recent years, researchers in AI have discovered a need to integrate different
kinds of reasoning systems, especially to integrate visual and symbolic
information and have turned to so-called "hybrid systems".  Similarly, these
robots are typically "situated" in that their reasoning systems are built so as to
exploit regularities in their environment.  For example, they don't have to know
anything about rivers because there are none in the building.  If the robot lives
only on one floor, then it can exploit regularities involving level surfaces, lack of
stairs, elevators, and so forth.

In this seminar, we will do two things.  We will delve into the literature on hybrid,
situated reasoning systems, and we will study and push forward on a particular
such system being developed here at IU.  Students with mathematical, logical, or
programming experience are all encouraged to attend.