Conflict resolving negotiation for COoperative Schedule Management Agents (COSMA)

Abstract

A five component negotiation model which structures the interaction of agents along different levels of increasing complexity is investigated: The lowest level introduces the possible message types. The protocol level links those messages together to create communicational contexts. The decision layer needs some criteria to rank and compare different scheduling proposals. The usage of those criteria implements the negotiation strategy of the strategic level. As an outline we sketch how a cooperation level emerges within the agent society and enforces cooperative behavior of the agents. The model is presented at the problem of appointment scheduling, because it is an inherently distributed problem which necessarily involves communication, negotiation, and cooperation mechanisms to resolve possible conflicts. COSMA agents (COoperative Schedule Management Agents) are designed to act as personal assistants to maintain their user's calendar. They are provided with competence to negotiate about the scheduling of appointments with other users. After describing the message types, the negotiation protocol is presented. The decision criteria are derived from a time model based on the association of time intervals and preferences. The negotiation strategies are guided by a local time file, the priority of a meeting and the results of earlier negotiations, which is encoded in a simple partner model maintained by the agents. An example finally shows the overall behavior of the agent society and will reveal and answer several questions about the technical details.