ONTOLOGY-DRIVEN DESIGN APPROACH TO DEVELOPMENT OF SITUATION AWARENESS SUBSYSTEM OF INTELLIGENT AGENTS

Abstract

An ontology-driven design approach to development of situation awareness subsystem (SAS) of intelligent agents is discussed. It is supposed that agents act in multi-agent competitive environments. The approach is based on the separation of domain independent and domain specific aspects. A set of domain independent models is proposed as a theoretical underpinning of the approach. Those models formalize different aspects of design and development of SAS including internal agent's representation of an environment, situation awareness calculation stages and SAS design stages. The models are used to build ontological and program components, used for SAS design and development. Two classes of ontologies are defined: SAS ontology and ontologies of specific domains. SAS ontology is used to provide domain specific program code integration into the domain independent SAS framework. Domain specific ontology is used to represent domain specific concepts on the base of SAS ontology concepts. SAS development process includes the following stages: domain specific ontology construction, automatized generation of domain specific code, integration of domain specific code into the SAS framework and implementation of domain specific functions within the framework. Automatized code generation is based on the proposed mappings of ontological classes and properties to the program structures. A toolset is proposed for building domain specific SAS. The toolset utilizes such languages from the Semantic Web stack as OWL, SPARQL and SPIN. The toolset is written in Java language. The approach is illustrated and evaluated on the simplified virtual soccer scenario.