Abstract

Supply network is a special form of organization, where two paradigms are combined: market interaction and hierarchical structure. Because it includes different interdependent entities that should work together, supply network requires efficient tools to support coordination and improve the quality of interaction between these entities. In this thesis, we propose a multi-agent-based interaction protocol for supply network formation. The main characteristic of this protocol is its adaptability to agent behavior during supply network formation. Such an adaptability is needed to form the most suitable organization structure. The novelty of the proposed protocol is the unification of contracting, auction and negotiation in a three-steps solution. Contracting is a planning tool aiming to facilitate solving network formation problem through task decomposition. Auction enables participating agents to jointly search an agreement space and check the contract feasibility. Finally, negotiation is the last resort for agents to reach an agreement. It is applicable when agents fail to reach an agreement due to lack of knowledge about the existing constraints. In fact, negotiation is used to release constraints when no realizable solution is obtained under auction. The second contribution of this thesis is the consideration of supply network formation within a general framework of coordinated distributed problem solving. The proposed protocol and framework are simulated through a multi-agent system prototype serving as a proof of concept. The simulation results show the effectiveness of our three-steps solution protocol in terms of network formation success, customer satisfaction and the total gained rewards of the whole network.