Graphical multiagent decision model: Towards adaptability

Abstract

Multiagent decision making is a typical decision problem in the changing world. We propose a new framework, including multiply sectioned influence diagrams (MSID) and hyper relevance graph (HRG), to represent this large and complex decision problem involving multiple agents. An MSID is an agency graphical language for representing decision problems in a distributed fashion while an HRG expresses the organizational relationships in multiagent systems. MSID, together with HRG, is partly adaptive to the changing world. Extended from basic evaluation algorithms in influence diagrams, three evaluation algorithms for solving MSID are discussed and compared. Based on MSID methodologies, an evolutionary decision model and a strategy for evaluation algorithm selection are proposed to address decision problems in the changing world. An evolutionary decision model facilitates the communication between domain experts and decision engineers while the evaluation algorithm selection method enriches the strategy to solve decision models. These promising methods aid the representing and solving of decision problems in the changing world, however, much challenging work still exists. Copyright © 2002, American Association for Artificial Intelligence (www.aaai.org). All rights reserved.