A cognitive-based agent architecture for autonomous situation analysis

We discuss the start of a project investigating a cognitive basis for intelligent agents that can approach the problem of situation understanding. We suggest that a practical system of intelligent agents can be build adapting existing agent modeling frameworks, ontologies from semantic web technology as well as a reasonable situation domain models. These can be brought together with a suitable cognitive architecture ACT-R which could be used to provide key roles in more human like situational awareness capability in emergency and disaster operations, especially where sensor information is harvested from semantically heterogeneous data sources. Existing situational ontologies and vocabularies can be supplemented by using DOLCE's formal ontology. This serves as a metalevel ontology that can relate different ontology modules and can generate new categories to extend an ontology (by agent learning) as needed, semantically-rich, conceptual level representations of real-world events. A Descriptions & Situations ontology provides a theory of ontological contexts capable of describing various types of context including non- physical situations, plans, beliefs, as entities so they can be communicated and understood between agents. We believe our system architecture provides a relatively good built-in infrastructure to meet fairly rigorous performance measurement requirements and has general applicability in a wide variety of situations.