A Polynomial Kernel-Oriented Coalition Algorithm for Rational Information Agents

Abstract

Information agents behave like active intelligent front-ends of stand-alone information systems. The main purpose of such an agent is to gather intensionally relevant information in non-local domains. They may either work as individuals or effi cientlycooperate in order to satisfy their own set of given information search tasks. However, in particular the need to respect the database autonomy requirements and to cope with semantic heterogeneity hinders such a cooperation. In this paper we present a solution for handling the autonomy during decentralized information-gathering by rational cooperation. For this purpose, methods for terminological knowledge representation and inference, as well as for utilitarian coalition formation among the information agents, are used. The decentralized agent-utility calculation bases on the agent's productions, resulting from executing tasks of fi nding dependencies between local terminological information models. There is no prior need nor a possibility to browse through foreign database schemas in order to fi nd some possibly relevant data. The coalition algorithm proposed in this paper enables effi cient cooperation via the formation of Kernel-oriented stable coalitions among rationally cooperating information agents in polynomial time.