A robust strategy for automated negotiations

摘要

The ubiquity of electronic commerce has resulted in considerable interest in automating the mechanisms it employs. Ideally, we would like to only leave closing a deal to the human user, and have software negotiate price. This paper propose a negotiation strategy that focuses on 1) beginning the negotiation process at a realistic and competitive price, 2) encouraging the continuation of a round of negotiations once it has begun, and 3) keeping the computation and communication load low so that the bulk of marketplace resources are spent on the negotiation process itself. We offer comparison with respect to the negotiated price, negotiation time and communication overhead against Red Agent and show favorable results. This work also presents the design of a marketplace where agents would meet and negotiate through a protocol, read product information from a central ontology, and be able to fetch negotiated prices for a product from the central history database. The work therefore demonstrates feasibility for a negotiation support system comprising both the negotiation strategy and a negotiation infrastructure.