Efficient Issue-Grouping Approach for Multi-Issues Negotiation between Exaggerator Agents

Katsuhide Fujita, Takayuki Ito, Mark Klein

Abstract: Introduction Negotiation is an important aspect of daily life and represents an important topic in the field of multi-agent system research. There has been extensive work in the area of automated negotiation; that is, where automated agents negotiate with other agents in such contexts as e-commerce(Kraus 2001), large-scale deliberation, collaborative design, and so on. Many real-world negotiations are complex and involve interdependent issues. To date, however, achieving high scalability in negotiations with multiple interdependent issues remains an open problem. We propose a new protocol in which a mediator tries to reorganize a highly complex utility space with issue interdependencies into several tractable subspaces, in order to reduce the computational cost. We call these utility subspaces “Issue groups.” The remainder of this paper is organized as follows. We describe a model of multiple interdependent issues negotiation and the strength of interdependency between issues, and the structure of interdependency graph. Next, we present a clustering technique for finding issue sub-groups. We present the experimental results, demonstrating that our protocol produces more optimal outcomes than previous efforts. Finally, we present our overall conclusions.