MAPS: A Multi-Agent Problem Solving System

P. Kefalas and D.Pratikaki

Abstract: The Multi-Agent Problem Solving system (MAPS) described in this paper, consists of a number of identical integrated agents which work independently within the same world. MAPS can be viewed as a framework within which highly distributed applications for problem solving could be developed. Given a goal, each agent uses its beliefs about the world in order to generate plans which best reflect its abilities. In the presence of other agents in the same environment, who continuously modify the world, the accomplishment of tasks may not be straightforward. It is, therefore, necessary that an agent in MAPS should engage communication with other agents, whenever exchange of information is required. The abstract software architecture of MAPS agents contains a number of modules which contribute to an efficient overall problem solving process. An example of MAPS effectiveness is demonstrated by implementing a simulator over a simple domain. This leads towards an experimental evaluation of the system which is finally discussed.

Keywords: Integrated Agent Abstract Architecture, Distributed Artificial Intelligence, Multi-Agent Systems

Appeared in: Proceedings of the 6th Panhellenic Conference on Information Techology, Athens, December 1997.

Available: Hardcopy on request from the authors. .


Back to Dept. of Computer Science, CITY Liberal Studies
Back to Petros Kefalas home page