Divide-and-Conquer in Automated Negotiations Through Utility Decomposition

Abstract

The success of a negotiation depends largely on the actors and the negotiation do- main. It is common that negotiators rely on an agenda to simplify the process and reach better deals. This is particularly the case when the preferences of the nego- tiators are complex and when multiple issues are at stake. Using an agenda to ex- plore and decompose the interdependence relationships between the issues is one way to address this problem. In this paper, we propose to address this challenge by applying the classical divide-and-conquer approach to automated negotiations through means of utility decomposition and bottom-up agenda construction. The approach does not impose an agenda from the top level of the negotiations, but builds it bottom-up given the individual utility functions of the agents and the relationships between the issues. Our approach reduces the cost of exploring the utility spaces of the agents and the resulting bidding processes. We implement the approach in a novel protocol called the Decomposable Alternating Offers Proto- col (DAOP). The experimental results show that our divide-and-conquer algorithm makes a positive influence on the global performance of an automated negotiation system.