Cross-layer Resource Scheduling And Optimization In Ultra-dense Heterogeneous Networks

The advent of the 5G Rel-15 standard has greatly accelerated the pace of 5G development and construction.The Rel-16 version under study is dedicated to further improving 5G on supporting more application scenarios and improving system performance.Ultra-Dense Network(UDN)technology is considered to be an effective way to improve system throughput,and therefor becomes one of the key technologies essential for 5G.In ultra-dense networks,increasing the reuse of local spectrum can address the need for coverage and capacity growth.However,all base stations(BSs)use the same spectrum resource at the same time,Inter-cell Interference(ICI)becomes very strong,which will result in Signal to Interference plus Noise Ratio(SINR).The reduction in the system limits the overall throughput of the system.Therefore,interference management in ultra-dense deployments is especially important.The first part of this paper is to consider the ultra-dense network environment where the density of small base stations is larger than the absolute threshold constant in a certain coverage.Aiming at the shortcomings of the original clustering algorithm,a dynamic clustering algorithm is designed to reduce interference,increase throughput and reduce user's average service time.The initial clustering is carried out with the aim of maximizing intra-cluster interference,and then the cluster is continuously updated according to the real-time state of the system,the quality of service of the user is ensured,the flexibility of satisfying the user's demand in real time is improved,and the spectrum efficiency is further improved.In the second part,a Q-learning based resource scheduling(QLRS)algorithm is designed to maximize system capacity.The small base stations are firstly uniformly grouped into clusters based on geographical location.During each schedulingperiod,resources are scheduled for each cluster according to the number of users and reallocated to small cells with associated users within each cluster to optimize the whole system throughput and energy efficiency.The rewards after each iteration will be recorded in the Q table.When the Q table is converged after times of iteration the optimal resource allocation scheme is obtained.The simulation results show that compared with other resource allocation algorithms,the proposed algorithm further improves the system throughput under the conditions of ensuring energy efficiency and macrocell throughput.