Complete and Incomplete Knowledge for Homer
Complete and Incomplete Knowledge for Homer
The natural language knowledge in the basic agent's lexipedia is completely pre-coded.
Thus, the agent does not learn new words or meanings. In addition, the
agent assumes complete knowledge, at some level of abstraction, of its
domain.
The episodic memory allows
learning of environmental details, but underlying principles involving
domain interaction are all pre-coded. The agent's knowledge satisfies the
99%
rule, so it is allowed to make the occasional knowledge-based mistake.
However, while Homer's domain knowledge is assumed to be complete,
Homer is only able to
directly perceive
objects in a limited field of view.
Objects which are out of range are assumed immobile and not moving.
If Homer is unable to produce a
plan because of this incomplete knowledge, it simply
fails instead of requesting the needed information.
Return to the top of this architecture.
Go to a discussion of this environment
for multiple architectures.
Current Location: Homer - Environment - Complete Knowledge