Self-organization and learning in multiagent based brokerage services

Abstract

Electronic Commerce is based on the exchange of information between involved stakeholders, using the underlying telecommunication infrastructure. Business-to-Consumer applications enable providers to propagate their offers, and customers to find an offer, which matches their demand. A huge collection of frequently updated offers and customers, the electronic marketplace, reveals the task of matching the demand and supply processes in a commercial mediation environment. The aim of the Abrose1 project (Agent based BROkerage SErvices for electronic commerce) launched by ACTS Domain V is to design a brokerage system based on adaptive multi- agent architecture. The first part of this paper explains the reasons of using this technology and gives a brief survey of related works. The second part describes the broker architecture and examines the underlying basic self-organization and learning mechanisms used. The third part concerns a description of the main tasks the global system has to perform. Then how it provides advanced electronic brokerage services in three business domains (tourism, trading and teleworking) is described. The paper ends on the current state of the Abrose system development.