An agent-based system may achieve its semantic richness through explicit ontologies, or domain-specific terminologies and theories. In order to detail the resulting ontology we introduce the DOD (Domain Ontology Description) phase in which we use a class diagram to describe the ontology of the domain representing the involved entities through classes.
PASSI is one of the few MAS design processes that include the design of ontology as an explicitly prescribed phase. This is the consequence of a precise choice since in our opinion, a good exploration of the domain is fundamental in order to deeply understand the problem and to have a good adherence between the application domain and the solution to be developed, particularly if the solution will be an agent-based one. For this reason the model we produce of ontology includes concepts (categories, entities of the domain), predicates (assertions on properties of concepts) and actions (that agents can perform in the domain).
In PASSI, it is important not only modeling the concepts of the domain but also the interactions of the agents with them. With interactions we mean the actions that agents can perform in the environment (using or affecting the constituting elements of the environment itself) and the predicates that can the knowledge of the agent on it.
In the next figur, a portion of the ontology used to describe the domain related to the PASSI toolkit chosen as a case study for this paper is reported. The diagram (an UML class diagram) represents concepts, predicates and actions with the following conventions: concepts are represented as classes with the concept stereotype and with the yellow fill color, predicates and actions are represented by classes with the predicate or action stereotype and with the blue or white fill color.
Concepts can be related using three UML standard relationships:
Exploring the figure, we can see that its elements define the pieces of knowledge of the agents and the ontology of their communications. For instance, we can there find the MAS concept that represents the MAS model instantiated by the designer during his work; it is composed by MAS elements (like agents, roles, tasks and all the other elements of the MAS meta-model discussed in subsection 3.1). Now let us consider this interaction occurring in a scenario involving the COD agent (User_interaction role) and the DOD agent (Ontology_server role). The first one (COD agent) needs a list of the already defined ontology elements in order to allow the definition of the referred ontology attribute of each communication. It sends a request to the DOD agent that answers with the MAS Description predicate; this latter describes the MAS model by listing its elements as it was requested. During the Communication Ontological Description phase, the designer could introduce a new communication between two agents and this corresponds to delegating the COD agent of performing the Add Element action described in Figure 13; it has two attributes: the actor and the receiver. As already said, this action will be performed by the COD agent that will be specified as the actor, the successful execution of the action will then notified to the agent that is responsible of ensuring the consistence among the different agents activities (the discipline agent responsible for the COD agent is Agent Society), this will be the receiver. The Add act specified in the action has two arguments: the MAS where the element will be added and the element to introduce in it.
Next phase: Communication Ontology Description
Previous phase: Task Specification (TSp)
Home: PASSI 2 homepage