| With the rapid development of wireless communication,emerging wireless technologies not only improve the performance of wireless network,but also enrich the application scenarios of wireless communication.At present,the commercialization process of 5G has started,and the research on B5G and 6G has also begun.6G will be a minimalist network with endogenous intelligence.A complete system-level simulation platform can not only assist in evaluating and optimizing the performance of 5G network,but also help to further explore and analyze the key technologies of B5G and 6G Meanwhile,as one of the important factors affecting system performance,resource scheduling algorithm is a research focus in every technical scenario.The corresponding scheduling module is also an important part of the system-level simulation platform.Consequently,this thesis introduces the built 5G/B5G system-level simulation platform,the proposed scheduling algorithms and corresponding module implementation.The main work is as follows:Firstly,based on the open source project K-SimSys,this thesis supplements and improves the functions to build a complete system-level simulation platform.The main construction work includes:1)Add a network topology model of circular region and a multi-antenna detection method based on the zero forcing(ZF)algorithm.2)Extend two kinds of business traffic models,the Bernoulli traffic model and the file transfer protocol(FTP)traffic model.3)In the wireless resource management module,a transport block(TB)mechanism is designed to simulate transmitted data.Meanwhile,the HARQ mechanism based on the TB form is implemented.In addition,three classical resource scheduling algorithms are implemented in the resource scheduling module,namely round robin(RR)scheduling algorithm,max C/I scheduling algorithm and proportional fair(PF)scheduling algorithm.Secondly,for the technical scenarios of sparse code multiple access(SCMA),this thesis studies the fairness scheduling algorithm for guaranteed bit rate(GBR)services.This thesis proposes two resource scheduling algorithms based on Hungarian algorithm,and carries out a simulation compared to the algorithm SCMA-PF.The simulation results show that,with relatively adequate resources,the proposed algorithm can improve the resource utilization,because it can make more users meet GBR requirement and increase the total throughput of the system.Although the fairness index of the proposed algorithm decreases,it can also achieve better fairness.Finally,in view of the requirement of endogenous intelligence in 6G,this thesis takes the scheduling module as an example to explore the combination of traditional system-level simulation platform and artificial intelligence(AI)algorithm.In this thesis,a distributed deep reinforcement learning algorithm framework is designed.After that,the overall architecture is designed to embed the intelligent algorithm module into the traditional system-level simulation platform.Lastly,an intelligent scheduling algorithm based on deep deterministic policy gradient(DDPG)algorithm is implemented.The effectiveness of distributed framework and the feasibility of intelligent scheduling scheme are verified by simulation.Simulation results show that the intelligent scheduling scheme can achieve similar throughput with PF algorithm. |
