Research On Adaptive Multi-service Intelligent Access In 5G Heterogeneous Networks

With the implementation and deployment of the fifth-generation mobile communication system(5G),ultra-dense heterogeneous networks(Het Nets)has become key technology that meets individual requirements of a large number of novel services on quality of services(Qo S).5G heterogeneous network architecture integrates multiple radio access technologies(RATs)such as 2G,3G,LTE-Advanced(LTE-A)and Wi-Fi,and supports diversified services by multiple base stations(BSs)and access points(APs)of different RATs.However,in the face of continuous evolution of network architecture and emergence of diversified services,the limited network resources make it difficult to support massive connections of the smart devices and meet increasing Qo S demands of various service applications.In addition,traditional network selection algorithms can not be flexibly applicable to the current complex network environment and service scenarios.Therefore,in 5G Het Nets,how to devise an intelligent and effective access selection mechanism has become a critical challenge on realizing user-centric and effectively solving wireless resource management problem.Combining with the latest researches on access selection mechanism,this thesis mainly studies the adaptive and intelligent network selection strategies for multi-service users in 5G Het Nets scenario.The main research works and contributions of this thesis are concluded as follows:Firstly,aiming at the problem that the existing access selection algorithms are difficult to improve system performance while ensuring personalized Qo S requirements,this thesis proposes a Nash Q-learning based intelligent network selection algorithm named MAQNS.By applying Analytic Hierarchy Process(AHP)and Gray Correlation Analysis(GRA)methods,MAQNS evaluates service preferences to guarantee that the personalized Qo S requirements are effectively considered in access selection.Subsequently,Markov Decision Process(MDP)and stochastic game are utilized to construct a multi-agent network selection model,then the MAQNS algorithm continuously learns Nash equilibrium solutions to find optimal access selection strategies.In this way,MAQNS can maximize long term performance of 5G network system on the premise of ensuring Qo S requirements of multi-service users,and balance network loads.Finally,evaluation results show that,compared with existing algorithms,the proposed MAQNS algorithm can not only significantly enhance system throughput but also effectively reduce user blocking,and the performance of user Qo S also can be optimized to some extent.Secondly,for the frequent network handover problem caused by the random motion of users,this thesis proposes an adaptive heterogeneous network selection algorithm named REMNS.REMNS firstly analyzes differentiated service requirements efficiently by designing a comprehensive preference evaluation framework from subjective and objective perspectives.Afterwards,by taking advantage of fuzzy logic theory that can quickly respond to the varying conditions,with considering the movement speed of users and received signal strength of networks,a network pre-assessment mechanism is devised to filter available networks and avoid frequent network handover.Meanwhile,benefiting from relative entropy theory,the optimal objective to maximize the quality of experience(Qo E)of multi-service users is proposed to make appropriate access decisions.Finally,evaluation results show that,compared with mainstream algorithms,the proposed REMNS algorithm can significantly improve Qo E of users requesting different types of services,effectively reduce user blocking.Therefore,the REMNS algorithm solves the access selection problem for users in any motion states,which is extensible in 5G network scenario that pays more and more attention on user experience.