1 Introduction
The control and optimization problems with a large number of agents have attracted lots of authors' interests because they appear in many fields, such as engineering, social, economic, biology and network communication([1] [2] [3]). In these areas, the mean field models which each agent interacts with the average effect of others and the individual has negligible effect on the overall population have attracted extensive attentions([4] [5][6] [7] [8]). Game theoretic solutions, the centralized control solutions and the decentralized control solutions have been developed in many papers([4]℃[16]).