Click here for full text:
An Internal Agent Architecture for Dynamic Composition of Reusable Agent Subsystems Part I: Problem Analysis and Decomposition Framework
Fonseca, Steven P.
Keyword(s): agent architecture; MAS framework; FIPAOS; JADE; Zeus
Abstract: The internal agent architectures that current MAS frameworks provide don't enable the rapid implementation of agents from reusable components. This is because of the underlying problem that the agent-oriented programming paradigm, defining how abstraction, decomposition, and modularity are achieved, is not sufficiently understood or developed. This paper presents an agent decomposition framework that offers agent-oriented programming and software engineering separation of concern guidelines and discusses the weaknesses of agent architectures currently supported by MAS frameworks. This is in preparation for proposing a more flexible and extensible internal agent architecture than current MAS frameworks provide. It is envisioned that this architecture will provide an infrastructure that supports constructing agents from reusable components at the subsystem level. Subsystems are added to an agent through a dynamic description, event, and ontology registration process. Once connected, the composite subsystems interact by an event-based software bus that acts as the central nervous system of an agent. This gives subsystems the ability to reason about the functionality and current state of their constituent parts and allows agents to be composed from industry wide best known components instead of building agents from a single MAS framework repository.
Back to Index