Multi-Agent Systems optimization for distributed watershed management

Water systems are characterized by the presence of many and often conflicting interests as well as distributed independent decision-makers. The traditional centralized approach to water management, as described in much of water resources literature, should be therefore reconsidered in the light of a more realistic, decentralized representation of the decision-making structure in order to provide effective solutions for policy making. In this paper we use Multi-Agent Systems (MAS) to analyse the role of a hypothetical watershed authority, which has to deliver a management plan at the watershed level starting from an uncoordinated situation. MAS provide powerful modeling and analytical tools to tackle this problem as they naturally allow to represent a set of heterogeneous and self-interested agents (e.g., the operator of a reservoir or a diversion dam, or an environmental organization) acting in a distributed decision-making process at the individual water user or stakeholder level. MAS have been widely explored in the water resources field, however most of these works focused on the use of agents as modeling tools, whereas only few considered them as optimization units. In this paper we adopt MAS within both a Distributed Constraint Satisfaction and Distributed Constraint Optimization framework to characterize different strategies between the two extremes of a centralized and a fully distributed management. The proposed approaches are demonstrated on a steady state hypothetical watershed management problem, involving several active human agents and reactive ecological agents. Results show that the imposition of normative constraints on agents’ decisions through a mechanism design strategy allows the watershed authority to effectively design an efficient and socially acceptable distributed watershed management.