SettleBot: A Negotiation Model for the Agent Based Commercial Grid

Market-driven sharing of distributed computational resources requires coordination support that can be provided by distributed problem solving (software agent technology). Multiple-issue negotiation among autonomous software agents allows the efficient alignment of resource consumers’ demand profiles and the service capabilities of resource providers. To address the inefficiencies of negotiations on imperfect markets, the negotiation model suggested by the SettleBot research effort includes both self-interested negotiations driven by a heuristic strategy and a joint-gains approach to win/win-negotiations. While finding joint gains under imperfect information is a well-known problem with approaches relating to simulated annealing as common approximate solutions, self-interested negotiations in a dynamically evolving environment require intelligent agents that retrieve, process and leverage knowledge about the world state. Superior strategy solutions in given market scenarios are identified using a genetic learning algorithm.