Study On Resource Allocation In Femtocell Networks Based On Game Theory

With the development of telecom industry and handheld mobile communication devices,the demand for high-rate wireless applications,specifically for indoor users,has remarkably increased.The traditional cellular network can not handle the increased demand for indoor wireless service,due to the intensive and unplanned deployment users.Femtocell is a promising technique for mobile operators to improve coverage and to provide high-data rate services in a cost-efficient manner.Meanwhile,the different access priorities and service requirements of macrocell users and femtocell users,together with the self-configuration and self-optimization characteristics of the two-tier femtocell networks,make the traditional resource allocation schemes for cellular networks not applicable.Thus,this study proposes various low cost,high effective and applicable algorithms for users under different service requirement and communication scenarioes.Firstly,the thesis studies the overlay spectrum sharing problem with the perfect channel information,and proposes a Cournot-based hierarchical game algorithm.Jointly considering the capability demand of MBS,priority of FBS,and dynamic spectrum price,this study proposes dynamic spectrum price function based on quadratic-cost and rare resource character.And a multi-period algorithm with dynamic price feedback is presented.And this guarantees the utility of primary users and secondary users to reach the maximum value in the corresponding tier,respectively.A dynamic trust value mechanism with needless overhead is designed.The secondary users which disturb the primary are punished by decrease the trust value.Thus,the spectrum sharing environment is improved.Secondly,this study proposes a hierarchical Stackelberg game for the multi-user multi-channel joint channel and power allocation problem under hierarchical Femtocell networks with perfect channel information.We specifically recommend a new power allocation function for MBS.The best response strategy of MBS can be directly obtained by using the proposed power-allocation function.As a result,costs are decreased.We also divide the strategy of MBSs into two substrategies,namely,subchannel selection strategy and power strategy.The two substrategies interact with each other in the game to improve the flexibility of the MBS strategy.A dynamic step-size mechanism is proposed to accelerate the algorithm.The proposed algorithm can improve the utility of both MBS and FBS,decrease the corss-tier interference.Then,this study deals with a joint channel and power allocation problem with multiple channels,users,constraints,and uncertain channel information in Femtocell networks.Specifically,a hierarchical robust Stackelberg game(RSG),which aims to achieving robust equilibrium,is first proposed for resource allocation with uncertainties.Considering the demand capacity of macrobase stations,an efficient fitness function in conjunction with particle swarm optimization-constriction factor(PSO-CF),is employed to yield the best response in the upper game to solve multiple constraints.Meanwhile,a stop protocol based on timing game is established for guaranteeing the convergence of hierarchical algorithm.This proposed algorithm can largely decrease the corss-tier interference,thereby improving the capability of FBS.Lastly,this study proposes two kind of game-based(unadversarial and adversarial)learning algorithms for the complete Femtocell networks which the channel gain is hard to obtain.The proposed algorithms do not need any prior channel gain information and direct information exchange among players.This study formulates this problem as multi-player MAB game.We introduce a concept of “virtual learning information” which can be obtained as the reward of the last actual played strategy to enrich the learning information and improve the learning ability.In the proposed algorithms,the players learn the actual played information in the upper learning;the “virtual learning information” is learned in the lower learning.The effective hierarchical learning manner significantly improves the learning ability and decreases the learning time.Meanwhile,a dynamic lower learning mechanism is proposed to avoid falling into an inadequate local extreme value.Due to the specific learning manner,the algorithm has high tolerance about delay and non-complete information.And the proposed learning algorithms show the high performance and adaptability in complex Femtocell networks.