MAGE: Multi-Agent Graphical Environment

Intelligent Agents in Cyberspace

Volume

Papers from the 1999 AAAI Spring Symposium

Contents

This paper describes our continuing research effort towards building a graphical development environment for rapidly creating, visualizing, and testing multi-agent software applications. Our system, Multi-Agent Graphical Environment (MAGE), addresses the need for enabling existing programs to be incorporated into agent-based software frameworks with minimal programming and the need for creating new agents and linking them to others. Users can create new agents and convert legacy programs to agents using the graphical user interface in MAGE and a messaging API at the source code level. Since MAGE handles all the agent-related bookkeeping and communication in a manner transparent to the application layer, users need not build an entire agent infrastructure for their applications. The graphical environment in MAGE enables the users to view and edit the configuration of their multi-agent systems. MAGE facilitates debugging by animating the message-passing among agents and allowing the users to execute their systems in a single-stepping mode. At the application layer, MAGE agents use KQML as their communication language. Below the application layer, MAGE employs CORBA to enable each agent to exchange messages with other agents. Currently, we are in the process of deploying MAGE on a web browser besides utilizing it to solve problems in domains such as information retrieval and data mining.