Research On Negotiation-based Coalition Formation Oriented Agent Social Network

Because of limited resources and capacity of the single agent to solve task,the single agent can not solve the task or the efficiency is very low in Multi-Agent System(MAS).Then need Multi-Agent to solve a common task through collaboration,and coalition formation of Multi-Agent is an important collaboration way to improve the capacity and efficiency of solving tasks.Therefore the researches of coalition formation become an important research topic in MAS.In general,the research of Multi-Agent coalition formation include the computing of coalition value,generate of best coalition structure and the distribution of the coalition value.Mostly previous research of Multi-Agent ignored the differences of the Multi-Agent network topology,but assumed the environment of Multi-Agent is fully connected,which is not only makes the problem of multi-Agent coalition formation is an NP-hard problem,but also can not reflect the differences of the actual application environment.At the same time,previous research also ignores the heterogeneity of agent in collaborative attitude and collaborative resources,so the negotiation in coalition formation can not be orderly and can not set dynamic collaborative attitude,which is lead to coalition formation inefficient.In this thesis,the Collaborative-Degree Based Distributed Automatic Negotiation Coalition Formation Mechanism has been proposed on the basis of the distributed negotiation model aimed at the deficiency of existing research.Firstly,this thesis built a social networks-oriental distributed negotiation model on the basis of previous distributed negotiation protocol,guarantee the convergence of negotiation and not deadlock by control the agent's state and consistency of coalition formation by the communication protocol.At the same time,this model can reflects the social networks in the different applications because it is for different social networks,which is can to contrast analysis the differences of coalition formation in different environment.Thirdly,the Collaborative-Degree Based Distributed Automatic Negotiation Coalition Formation Mechanism has been proposed.The Agent is heterogeneous in this thesis,which is reflected by the negotiation tactic.The negotiation tactic includes the agent's collaborative attitude and agent's collaborative resources.Doing so will not only more consistent with the actual application background but also improve the efficiency of coalition formation by the dynamic collaborative attitude based on collaborative-degree and orderly negotiation.Finally,the experiment simulates the process of coalition formation for heterogeneous Agent under the different social networks environment.The experimental results show that the coalition formation mechanism proposed by this thesis has a better performance in small world social networks,which is reflect most of the applications,so the practicability of the mechanism has been verified.At the same time,the mechanism can improve the performance of negotiation rounds,success rate of negotiation and the average individual utility,which is verify the feasibility and validity of the mechanism.