Track:
Contents
Downloads:
Abstract:
One of the dangers in building cognitive architectures is modeling biological phenomena that do not exercise a large portion of the cognitive machinery. Focusing on situation awareness or situation estimation avoids this problem, since it requires iterating between sensory processing and rea-soning. A system doing situation estimation must include the ability to explain and predict the events produced by sensory processing. Prediction requires that perceived information not influencing the principal actors be ignored. However, novel and unanticipated sensations cannot be safely ignored by situation estimation and must trigger a switch in the reasoning frame in which predictions are being formed. While extraneous information needs to be ignored, the missing portions of objects or activities supporting anticipated situations must be filled-in. Internal world models are needed for recognizing objects and events, filling-in unseen portions of the world, and predicting how executing a behavior will affect the world.