Agent-Based Middleware for Model Neutral Data Representation

The complexity of systems design has dramatically increased over the past decade. Developers are responding to the complexity by modeling such systems in order to further understand, characterize, or analyze their behavior. Any model-based application may actually utilize multiple models which potentially require common information. Thus, a benefit can be derived, if one model utilizes the seamless capture of information by another model. Unfortunately, information representations can be quite different and so simple information passing is impractical. Exchange between differing model representations will require mechanisms and structures to act as a "middleware" between the models. The design of an Agent-Based Middleware is proposed for Phase I which will allow models of differing representations to share information. Models will share by utilizing common communications protocols and common model-content structures through their associated Agent Interactions Wrappers. Common communication protocols will be extensions of current intelligent agent communication research. Common structures will be devised from generic object-based constructs, producing an exchangeable model information middleware. To assist the conversion of model content to the middleware, a user extendible ontology of concepts, entities and relationships will be devised. The final design will clearly define all structures and mechanisms involved in this open architecture.