Research On Radio Reasource Allocation And Mobile Load Balancing For 5G RAN Slicing

With the rapid development of Internet,an increasing number of mobile users connect to mobile network through various devices,for instance,mobile phones,tablets and other terminal devices.The Fifth Generation Mobile Communication System(5G)will support even more diversified services with multiple requirements including throughput,latency,and reliability.Network slicing technology which is based on Network Functions Virtualization(NFV)is one of the key technologies to deal with the diversification of 5G network services.In the process of implementing 5G Radio Access Network(RAN)slicing,it is necessary to solve the challenges of inter-slice isolation and mobility management.This topic is derived from the China Unicom Network Technology Research Institute’s cooperation project "B5G Network System Key Technology Research Project".It mainly studies radio resource allocation and mobile load balancing algorithms for 5G RAN slicing,to improve radio resources utilization and users" experience.The main contribution of this study are listed as follow:(1)The researches on radio resource allocation and mobile load balancing of network slicing are reviewed.First,the network slicing technology is briefly introduced.Secondly,the relevant principles of RAN slicing design are clarified,and some challenges to the design of 5G RAN due to the introduction of network slicing are discussed.Finally,we summarize the research progress of RAN slicing radio resource allocation and mobile load balancing,and clarify the research motivation of this paper.(2)To address the problem of isolation between slices,a radio resource allocation algorithm for SLA contract rate maximization is proposed.First,the business parameters in Service Level Agreement(SLA)are mapped to the measurable network performance metrics,by analyzing user’s metrics and calculating SLA contract rate of slices to reflect performance of slices.Secondly,radio resources are allocated to network slices on the basis of the collected SLA requirements,optimal resource allocation scheme is provided for slices.Meanwhile,radio resources of slices that do not meet the requirements are dynamically updated without affecting the performance of slices which has met the SLA requirements,so as to ensure the isolation between slices and maximize the SLA contract rate of all slices.The simulation results show that the proposed algorithm can serve more users to achieve better slice SLA contract rate.When the slices’ demand of radio resources changes dynamically,the radio resource update strategy can restrain the decrease of SLA contract rate,keep the contract rate at a higher level,and ensure the isolation between slices.(3)Considering mobile load balancing for RAN slicing,a mobile load balancing algorithm based on Deep Reinforcement Learning(DRL)is proposed.First of all,we propose a system utility model to measure the satisfaction of UE’satisfaction.At the same time,a mobile load balancing strategy for RAN slicing is proposed,including slice-level load control realized by adjusting the proportion of radio resource allocated to the slice by Radio Remote Unit(RRU),and cell-level load balancing based on handoff.In order to improve the system satisfaction and maximize the system utility of mobile load balancing,the joint optimization problem of system satisfaction and user utility function was proposed in slice-level load control and cell-level load balancing respectively.The DRL is used to solve these optimization problems,and the optimal radio resource update of slice-level load control and handoff of cell-level load balancing are respectively realized.The simulation results show that the proposed algorithm can effectively reduce the total number of handoffs in the system and bring less balancing overhead.In addition,the proposed algorithm can effectively reduce the number of unsatisfied users and achieve higher system satisfaction.