KODAMA Project: from Design to Implementation of a Distributed Multi-Agent System

摘要

This paper describes the design and implementation of a worldwide distributed multi-agent system, called KODAMA (Kyushu University Open & Distributed Autonomous Multi-Agent). Historically, research on multi-agent systems is a natural extension of the evolution of programming, as well as communications and control technologies. Through our work on the KODAMA project, we concentrate on enabling technologies for building scalable, flexible and secure multi-agent systems, which can operate in a worldwide computing environment. The innovative points of KODAMA are the approaches it takes to formalizing a global distributed computing architecture based on agent-oriented programming. The core concept behind our design is a separation principle that mandates the separation of application-level logic from agent-level logic and the separation of agent-level logic from network-level logic. In this way, the inherent distribution and complexity that the worldwide computing environment involve can be abstracted, encapsulated and tackled appropriately at different levels. In line with the separation principle, this article gives a general introduction to the key abstraction models that form the basis of KODAMA.