Developing Intelligent Agents with Distributed Computing Middleware

Intelligent agents embody a software development paradigm that merges theories developed in Artificial Intelligence (AI) research combined with computer science. The power of agents comes from their intelligence and also their communication. Current agent development methodologies and resulting frameworks have been developed from an AI perspective. From a developer’s point of view they introduce new programming concepts and provide a specialised execution environment. Considerable emphasis is placed on hiding away the underlying complexity of how agents actually operate. However, the fact is; agent systems are inherently distributed software systems and this brings significant implications in their application and more importantly integration. This has been largely underestimated by the agent community resulting in increased development risk in large production systems. The Distributed Object Computing (DOC) development methodology on the other hand has been used to successfully build large scale distributed software systems using standards-based middleware. In this context Objects encapsulate behaviour and are inherently integrated with any system utilising compatible middleware. This paper explores the possibility of leveraging the power of both approaches through a proposed Agent Architecture Framework (AAF) that implements generic agent behaviours and algorithms with DOC middleware using well understood software design patterns.