Multi-Agent Argument Negotiation Strategy Based On Differential Evolution Algorithm

Nowadays, more and more researchers focus on studying business negotiation,and the efficiency of negotiation has been improved rapidly. However, in multi-Agentbased negotiations, most negotiations use liner function as the objective function, inwhich the measuring criterion is total utility alone. In this circumstance, the optimalproposal tends to lean to the boundaries, due to the degeneration of negotiation.Although the optimal proposal could achieve the maximum utility, the negotiationagents find the proposal less satisfied, because the satisfaction degree of individualissue is also an important factor, so that negotiation agents would prefer the proposalwith balanced issues.In this dissertation, an argument based multi-agent negotiation based ondifferential evolution is proposed. This strategy could be described as the following.In order to solve the bi-polar problem of initiate proposal, the single-objectivefunction has been improved into multiple-objective function, in which a betterbalanced proposal would be generated through Differential Evolution Algorithm. Inthis dissertation, the standpoint of balanced proposal is introduced, which is usingmultiple-objective function as the fitness function to improve the quality of initiateproposal. Meanwhile different types of proposals could be generated, according to thepreference of negotiators. Finally, editing and running the algorithm in Matlab,through which the validity of this negotiation strategy has been proved.When the negotiation reaches a deadlock, argument would be introduced.Argumentation based on Multi-agent lead the argumentation theory into negotiation,which could improve the flexibility of agents. In order to make the negotiation moredirect, an argument negotiation strategy based on proposal classification has beenproposed. In this strategy, some quotas are used to judge the proposals, which are thesatisfaction degree of total utility and satisfaction degree of single issue utility. Thendivide the proposals into5types, for each of which the corresponding argument isgenerated and sent. Repeat this argument process until an agreement is reached or thenegotiation failed, which would accomplish the whole negotiation process.After realization the aforesaid argument negotiation strategy, the validity of thisstrategy has been proved. Meanwhile, an experiment in the environment of JADE hasbeen performed to prove the effectiveness of this negotiation strategy. This argumentbased negotiation strategy is meaningful in both theory research and implementation.