Research On Business-Driven Network Selection In 5G Heterogeneous Networks

With the rapid development of wireless communication technology,network traffic has increased rapidly.Especially since the commercial use of the 5^th Generation Mobile Communication System(5G),it is difficult for a single radio access technology(RAT)to cope with the increasingly complex and changeable business needs of users.Heterogeneous wireless networks(HWN)have been formed,which include multiple RATs such as 5G and the 4^th Generation Mobile Communication System(4G).Different RAT systems have certain differences in network coverage,available resources,mobility support,and quality of service(Qo S)guarantees.In such an environment,traditional network selection methods cannot accurately model user needs or adapt to user services.Most of the existing vertical handover solutions for HWN are not comprehensive enough,which may cause problems such as low handover efficiency.In the case of taking user Qo S and network efficiency into account,it is necessarily to study how to dynamically select the most matching network in HWN for users according to their business needs,so as to ensure the performance indicators of communication services and provide users with personalized business experience.In view of the above discussion,the main research contents of this article are as follows.(1)According to the relevant knowledge and technology of mobile cellular networks,a simulation experiment platform for heterogeneous cellular networks deployed jointly by 4G and 5G is built based on Python.The theoretical basis includes 4G and 5G wireless frame structure and air interface physical resources,the mathematical relationship of wireless air interface measurement indicators,the base station resource scheduling procedures,the calculation methods of channel propagation path loss and bit error rate.In the early stage,it is used as a basic experimental platform to obtain various data required by the proposed scheme.In the later stage,it is used as a verification platform to measure the performance gain of the proposed scheme through statistical performance indicators.(2)A cell attribute prediction method based on Long-Short Term Memory(LSTM)is studied.The LSTM model is trained with the data of each cell attribute(including throughput,delay,jitter,packet loss rate and cell load rate)obtained in the basic experimental platform.The trained model is used to predict the network attributes of the cell in time series,and its effectiveness is verified by simulations.(3)A business-driven network selection scheme is studied.Different types of communication businesses in mobile networks have different requirements for Qo S indicators.First,the utility function is used to model each network attribute(including Qo S indicators and cell indicators).Secondly,a subjective-objective-combined weighting method based on user preferences is adopted to obtain the combined weight of each network attribute.Then,combined with the previous network attribute prediction method based on LSTM,the comprehensive utility value of the network is calculated according to the combined weight,which characterizes the user's satisfaction of current communication business with the network performance and is used as the judgment criterion for network selection.Finally,the performance gain of this scheme is verified by simulation.