An Evolutionary Cooperative Mechanism for Multi-agent System

Abstract

The optimal solution for the cooperation win-win task for Multi-agent system (MAS) was studied and a automatic negotiation reciprocal mechanism was build in this paper for the typical case of multi- agent foraging. The equivalences between agent in MAS and player in game theory, optimal solution in MAS and equilibrium in game theory, and cooperation task in MAS and multi-stage Markov game were then established. To the difficulties of equilibriums selection, a hawk dove game model, which is a uncorrelated asymmetry game, is build from the point view of evolutionary game theory. And a self-enforcing cooperative algorithm based on replicator dynamic was proposed to find the evolutionary stable strategy of hawk dove game. The simulation verified the effectiveness of the proposed approach.