Roy M. Turner
Department of Computer Science
Kingsbury Hall
University of New Hampshire
Durham, NH 03857 USA
E-mail: rmt@unh.edu
Abstract To operate successfully in a complex world, intelligent agents must exhibit
context-sensitive behavior. Context impacts the appropriateness of virtually
all aspects of an agent's behavior, yet most existing reasoning approaches pay
little if any attention to explicitly recognizing, reasoning about, and making
use of knowledge about the current context. We have developed a mechanism as
part of our work on schema-based reasoning that uses
(c-schemas) to explicitly represent contexts an agent
may encounter. The agent's context manager retrieves the best c-schemas
from its memory based on features of its current situation, then merges them
to form a view of the current context, the
. This is then used to set behavioral parameters,
initiate and terminate context-specific actions, focus its attention on
appropriate goals to achieve, select actions for achieving them, and rapidly
and appropriately handle unanticipated events. An early version of this
approach was implemented in MEDIC, a schema-based medical consultant;
currently, we are developing the approach in ORCA, a schema-based
controller for autonomous underwater vehicles. We are also extending it for
use in cooperative distributed problem solving systems.
Context-sensitivity is fundamental to intelligent behavior. An organism's
context conditions what stimuli it is receptive to, what interpretation it
places on them, and its responses. Context modulates behavior by affecting
the actions used to achieve goals as well as the timing and manner in
which those actions are carried out. By paying attention to its context, an
intelligent agent can more quickly select appropriate behavior to achieve its
goals, and it can more effectively focus its attention and respond to
unanticipated events.
Unfortunately, context-sensitive reasoning has received little attention in the artificial intelligence literature. In most existing reasoning approaches, any contextual knowledge that is present is generally spread throughout the knowledge base as applicability conditions on operators. Nowhere is context explicitly represented or reasoned about.
For several years, I have been developing a mechanism for automatic context-sensitive reasoning. The result, part of an approach to adaptive problem solving called schema-based reasoning, uses frame-like contextual schemas, or c-schemas, to represent prototypical contexts an agent knows or has been told about. The agent assesses its current situation by retrieving one or more c-schemas from its memory based on features of the situation [\protect\citenameTurner, 1992]. If more than one are applicable, they are merged to give an overall picture of the current context. The contextual information is then used as predictive and prescriptive information to: