| The traditional communication network is designed as a single network architecture,which provides all network functions through vertically integrated network elements,but it can not adapt to 5G diversified and differentiated service scenarios.In order to support multiple service scenarios with different performance requirements on the same infrastructure,network slicing(NS)came into being.It divides a physical network into multiple virtual logical networks through virtualization technology.Each network slice is logically isolated to adapt to various types of services and meet the different needs of users.Network slicing usually includes access side slice(including radio access and fixed access)and core network slice.This thesis mainly studies the slicing technology of radio access network,focusing on the two-layer resource scheduling method of inter-slice and intra-slice,which can effectively allocate radio resource according to the slice state,and meet the needs of different services and users at the same time.The main research contents of this thesis are as follows:(1)For the radio access network slicing scenario under a single base station,in order to give consideration to isolation and resource efficiency in high load scenarios,this thesis proposes an efficient,flexible and low complexity two-layer radio resource scheduling algorithm suitable for practical systems.The algorithm first allocates resources to enhanced mobile broadband(e MBB)slice,ultra reliable low latency communication(u RLLC)slice and massive machine type of communication(m MTC)slice according to the slice’s traffic load,priority and set isolation level,and then adopts different resource management strategies for each type of user demands within the slice.Simulation results show that compared with the existing radio access network(RAN)slicing algorithm,this algorithm can meet the throughput requirements of e MBB users and the delay requirements of u RLLC users under high load;At the same time,the algorithm has low complexity and high flexibility.Different isolation levels and priorities can be set according to different scenarios and needs.(2)For the slicing scenario of radio access network under multi base stations,the previous resource allocation methods can not meet the requirements of slicing when the number of slices changes,and are only applicable to specific scenarios.To solve this problem,this thesis proposes a method to realize the optimal resource allocation independent of the number of slices.This method first uses Ape-X method(a deep reinforcement learning method)to allocate resources to slices,and then uses the resource mapping from slices to base stations and user resource allocation to meet user needs.This method manages resources in a slice-agent way,that is,a agent controls a slice.In addition,the requirements of slicing can be met without over allocating resources.Simulation results show that the proposed method can allocate resources according to the state and demand of slices,allocate necessary resources to meet the demand of slices,and is not affected by the change of the number of slices.At the same time,this method also has high general performance and scalability. |
