The multi-agent resource allocation problem corresponds to the negotiation of m resources among n autonomous agents, in order to maximize a social welfare function. Contrary to some former studies, the purpose is neither here to simply determine a socially optimal resource allocation nor to prove the existence of a transaction sequence leading to this optimum, but to find a practical transaction sequence among agents, for any type of contact networks. With this intention, we study various agent behaviors in order to identify which one leads the most often to an optimal resource allocation. The allocation that is reached, can be viewed as an emergent phenomenon that comes out from local interactions among the agents.
After a study of different transaction types, we show that, among the set of studied transactions, the so called "social gift" transaction, is the most efficient one for solving the resource allocation problem associated with the utilitarian social welfare.
Paru en juin 2008 , 20 pages