Analysis of Negotiation Protocols by Distributed Search

Negotiation is one of the main mechanisms for coordination and cooperation in multiagent systems. However, most negotiation protocols are complex and their features are difficult to characterize. In this paper, we propose a general experimental approach to analyzing negotiation strategies using distributed search. In this approach we first formulate the problems that negotiation protocols intend to solve as distributed constraint satisfaction/optimization problems, and then capture the negotiation protocols as distributed search algorithms. By analyzing the derived search algorithms, we can characterize many important properties of the negotiation protocols. In this paper, we are particularly interested in the properties of a newly developed negotiation protocol, which is motivated by distributed sensor network applications, including its completeness, complexity, convergence rate, and scalability. Although the idea of viewing negotiation as distributed search is not completely new, in this research we not only view negotiation as distributed search, but directly apply a search algorithm to reveal the essential features of a negotiation protocol and analyze its performance.