Since the paradigm of "agent-oriented programming" (AOP) was introduced by Yoav Shoham in 1993, many programming languages, tools and techniques have been developed for building software systems as collections of communicating "agents". In AOP, agents are designed and directly programmed in terms of high-level constructs based on aspects of human reasoning and communication, such as the symbolic representation of knowledge (or beliefs), proactive behaviour based on explicit goals and plans, and knowledge-level communication in terms of speech acts.
Proponents of AOP have long been describing this approach as a promising one for simplifying the design and construction of complex distributed systems, due to its correspondence with "natural" ways of human thinking and the ability to explicitly model the differing local knowledge and goals of the constituent agents. However, the use of AOP has not yet found its way into mainstream software development. In this talk I will discuss some possible reasons for this, and how agents might yet find their way into the mainstream.
In particular, I will discuss an approach for integrating agents into enterprise computing environments using concepts from the enterprise integration patterns of Hohpe and Woolf, and will describe the design of agent "endpoints" for the Apache Camel message routing and mediation engine. This approach will be illustrated using a case study involving the integration of agents programmed in the Jason agent platform with a database management system, a mail server, a message broker and the Apache ZooKeeper coordination server.
Last modified: Tuesday, 20-Aug-2013 14:51:03 NZST
This page is maintained by the seminar list administrator.