Accession Number : ADA294938

Title :   Everyday Believability.

Corporate Author : CARNEGIE-MELLON UNIV PITTSBURGH PA SCHOOL OF COMPUTER SCIENCE

Personal Author(s) : Pelton, Garrett A. ; Lehman, Jill F.

PDF Url : ADA294938

Report Date : MAR 1995

Pagination or Media Count : 14

Abstract : Believable Agents must be able to handle a large number of goals, opportunistically acting on them while maintaining a focus of purpose. They also must be able to operate over long life times without signifIcant performance degradations. We describe in this paper a method of managing an agent's goals so that the agent can make progress on as many goals as possible, without spending undue time considering goals that it can't make progress on. The key part of this method is a way of specializing and generalizing the set of perceptual cues that are used to bring a goal into consideration by the agent. The method is described via a simple example. (AN)

Descriptors :   *LEARNING MACHINES, *ARTIFICIAL INTELLIGENCE, ALGORITHMS, COMPUTERIZED SIMULATION, OPTIMIZATION, DECISION MAKING, DEGRADATION, RULE BASED SYSTEMS, LONG RANGE(TIME), OPERATING SYSTEMS(COMPUTERS), PATTERN RECOGNITION, SYSTEMS ANALYSIS, CONTROL THEORY, EXECUTIVE ROUTINES.

Subject Categories : Cybernetics

Distribution Statement : APPROVED FOR PUBLIC RELEASE