Abstract of FKI-233-99

Document-Name:  fki-233-99.ps.gz
Title:		A Multiagent Framework for Planning, Reacting, and Learning
Authors:	Gerhard Weiss 
Revision-Date:	1999/09/01 
Category:	Technical Report (Forschungsberichte Künstliche Intelligenz)
Abstract:	There are two main approaches to activity coordination in 
                multiagent systems, plan-based coordination and reactive 
                coordination, each having its specific advantages and 
                disadvantages. This paper describes a framework called 
                M-Dyna-Q that aims at combining these two approaches such 
                that their advantages are retained while their disadvantages
                are avoided. The key idea behind M-Dyna-Q, which is a 
                multiagent variant of a single-agent framework known as Dyna-Q,
                is to achieve this combination through equipping the agents 
                with the capability to learn information that is relevant to 
                their planning and reacting activities. M-Dyna-Q thus offers 
                an integrated view of planning, reacting, and learning in 
                multiagent systems. 
Keywords:	(M-)Dyna-Q, multiagent systems, reactive coordination,
                plan-based coordination, learning
Size:		13 pages
Language:	English
ISSN:		0941-6358
Copyright:	The ``Forschungsberichte Künstliche Intelligenz''
		series includes primarily preliminary publications,
		specialized partial results, and supplementary
		material. In the interest of a subsequent final
		publication these reports should not be copied. All
		rights and the responsability for the contents of the
		report are with the authors, who would appreciate
		critical comments.

Gerhard Weiss