Use of Petri Nets for Modeling an Agent-Based Interactive System: Basic Principles and Case Study

Several architecture models of interactive systems have been put forward by researchers over the past twenty years. Two main types of architecture can be distinguished: architectures with functional components: Langage (Foley & Vandam, 1982), Seeheim (Pfaff, 1985) and ARCH (Bass et al., 1991) and architectures with structural components: PAC (Coutaz, 1987), PAC-Amodeus (Nigay, et al., 1997), MVC (Goldberg, 1983), AMF (TarpinBernard & David, 1999), H4 (Guittet, 1995),.... The approaches currently used in interactive system design adopt a modular structuring aimed towards a better apprehension of the reactivity, flexibility, maintainability and re-use. Agent-based approaches are promising in this way. In the agent-based architecture proposed, we suggest using a division into three functional components: the application agents which handle the field concepts and cannot be directly accessed by the user; the interactive agents (or interface agents, or presentation agents) which , unlike the application agents, are in direct contact with the user (they can be seen by the user); the dialogue control agents which are also called mixed agents (Ezzedine & Trabelsi, 2005). Each agent therefore plays a role within its group; this role can be expressed in the form of the services it offers in the interactive system. We use so-called agent Petri Nets (PN) to model a priori the services offered by each interface agent: a service is defined as being a quadruplet S ={E, C, R, P}, with E: the event which triggers the service, C: the conditions to be met in order to perform this service, R: the resources necessary for the service to be performed, P: the property of this service, which can be either an operation concerning the agent alone (with or without a change of state for the interactive agents), or a call for the service of another agent. The succession of various calls for services gives rise to the succession of page-screens in the human-computer interface. This chapter begins with a state of the art about the use of Petri nets in Human-Machine Interaction. Then we explain the problem relating to agent-based architectures of interactive systems, and we propose a solution for the modeling of the interface agents of such architectures. Lastly, we illustrate our approach by a case study.