Towards robust agent behaviors in modeling and simulation: Situation filling in with commonsense knowledge

Abstract

We propose the use of commonsense knowledge for situation fill-in to achieve robust agent behaviors in modeling and simulation. We describe the representation and reasoning scheme required to support the said context of use; a knowledge structure for commonsense narrative and a spreading activation computation model for performing inference on the knowledge structure. A prototype was implemented and an exemplar demonstration case was run to show its potential use. The paper then rounds up with a discussion on how the commonsense knowledge retrieved for situation fill-in enables robust agent behavior in the simulation environment.