The Research On Negotiation Strategies And Coalition Mechanisms Of Agents

Agent negotiation and coalition are key technologies of multi-agent cooperative problem solving. It is also an effective solution method in multi-agent system (MAS). This paper summaries the progresses and exiting problems on researches on negotiation and coalition in multi-agent cooperative problem solving, and researches on logic model of agents, negotiation frameworks and coalition formation. Includes the following details: (1) Established a logic model of agents. To fit for the need of complex social actions, and synthetically consider the environment, mental states and decision utility of agents, we proposed an agent model combined with environment, mental states and utility. Detailedly depicted its mechanism and gave its syntax definition and semantic model. It realized the balance between logic rationality and utility rationality and partly improved the work of Russell, Wooldridge and Rao. (2) Gave two multi-issue negotiation frameworks of multi-agent separately based on decision utility and forgetting theory. We proposed a multi-issue negotiation of multi-agent based on sum-product decision utility, which could deal with negotiations between utility redeemable issues and irredeemable ones. Used argument, and made agents conferring the attitude of other agents. We introduced learning algorithm and dynamicly adjust the utility height to quicken the course of negotiation. Also, put forward a multi-agent negotiation based on forgetting theory, which combined with logic and game theory to achieve the balance between negotiation logic rationality and utility rationality. Dynamicly adjust utility and deal with irredeemable problems using logic and. It is a new solution of multi-issue negotiation. (3) Presented a mix ant colony and genetic algothrim to solve multi-agent coalition. We adopted biology evolution algorithm to find the optimal coalition. Syncretized and improved the basic ant colony algorithm and the basic genetic algorithm, made the advantage reciprocally, and proposed a multi-agent coalition formation method based on mix ant colony and genetic algorithm. It sufficiently exerted the advantage of resolving optimal problem using ant colony and genetic algorithm. It offsetted the shortages of traditional methods, that introduced restriction condition and pruning to produce an approximate solution. It can quickly fetch the optimal coalition.