An agent framework for processing FIPA-ACL messages based on interaction models

Abstract

Interaction engineering is a key issue to effectively build Multi-Agent Systems. It requires software abstractions, components and control structures to manage interactions among agents and to improve infrastructures at runtime. We propose a framework for automatic processing of interactions generated using FIPA-ACL, a language widely accepted for agent platforms. This framework includes three elements: i) an agent interaction architecture to systematize interaction processing tasks, ii) interaction models to build re-usable validated code used to check different phases of interaction processing associated with message semantics, and iii) components and control structures implementing interaction architecture for a particular agent platform. The paper describes the implementation details of the proposed approach developed within the CAPNET agent platform and illustrates it by example. © 2008 Springer-Verlag Berlin Heidelberg.