Reaching Agreements for Coalition Formation through Derivation of Agents' Intentions
Résumé
This paper addresses the coalition formation problem in multiagent systems. Although several coalition formation models exist today, coalition formation using these models remains costly. As a consequence, applying these models through several iterations when required becomes time-consuming. This paper proposes a new coalition formation mechanism (CFM) to reduce this execution cost. This mechanism is based on four principles: (1) the use of information on task relationships so as to reduce the computational complexity of the coalition formation; (2) the exploitation of the coalition proposals formulated by certain agents in order to derive their intentions, (this principle makes the search for solutions easier, which in turn may result in earlier consensus and agreements-the intention derivation process is performed on a new graph structure introduced in this paper); (3) the use of several strategies for propagating the proposals of the agents in the coalition formation process; and (4) the dynamic reorganization of previous coalitions.