Using multiagent negotiation to model water resources systems operations

The operations of water resources infrastructures, such as dams and diversions, often involve multiple conflicting interests and stakeholders. Among the approaches that have been proposed to design optimal operating policies for these systems, those based on agents have recently attracted an increasing attention. The different stakeholders are represented as different agents and their interactions are usually modeled as distributed constraint optimization problems. Those few works that have attempted to model the interactions between stakeholders as negotiations present some significant limitations, like the necessity for each agent to know the preferences of all other agents. To overcome this drawback, in this paper we contribute a general monotonic concession protocol that allows the stakeholders-agents of a regulated lake to periodically reach agreements on the amount of water to release daily, trying to control lake floods and to supply water to agricultural districts downstream. In particular, we study two specific instances of the general protocol according to their ability to converge, reach Pareto optimal agreements, limit complexity, and show good experimental performance.