Research On Resource Allocation Optimization Of 5G Ultra-Dense C-RAN Based On Deep Reinforcement Learning

With the rapid popularization and application of 5th Generation Mobile Communication Technology(5G)in the world,5G is becoming an indispensable part of People’s Daily life,work and entertainment.It also provides strong support for the further promotion and popularization of applications and concepts such as live ultra-high definition video,autonomous driving and meta-universe.Cloud Radio Access Network(CRAN)is the main networking mode for the future 5th generation mobile communication,and Ultra-Dense Network(UDN)is one of the core technologies of 5 G.In this paper,resource allocation optimization in 5G ultra-dense C-RAN architecture is studied.The demand for data traffic in areas with dense human traffic has always been a major challenge that mobile communication operators urgently need to solve.However,ultra-dense networking technology can greatly improve the capacity of 5G network and alleviate the demand for data traffic in areas with dense human traffic.Aiming at the optimization of the allocation of wireless spectrum resources in the downlink 5G ultradense network,this paper first analyzes and models the 5G ultra-dense networking mode of the macro station and micro base station,gives relevant concepts and formulas in the 5 G ultra-dense networking scenario,and models the 5G wireless resource allocation in the 5 G ultra-dense networking scenario.In terms of how to evaluate the performance of the network,this paper takes into account the two indexes of Energy Efficiency(EE)and Spectrum efficiency(SE),and carries out dynamic weighting of energy efficiency and spectrum efficiency.The problem of resource allocation at different time is expressed as a joint optimization problem of energy efficiency and spectral efficiency.On the problem of how to optimize 5G ultra-dense network resources,this paper puts forward a resource allocation scheme based on Advantage Actor-Critic,describes the wireless resource allocation problem as a Markov decision process,and defines the state space,action space and reward function based on deep reinforcement learning.Finally,the proposed scheme is verified by simulation in this paper,and compared with the resource allocation scheme based on DDQN algorithm and the resource allocation scheme based on random allocation strategy.The results prove that the resource allocation algorithm proposed in this paper based on Advantage Actor-Critic can effectively converge,and compared with other distribution schemes,The resource allocation scheme based on Advantage Actor-Critic has higher system performance and higher system throughput,and can guarantee the long-term performance of ultra-dense networking system.Aiming at the joint allocation optimization problem of CU-DU separated 5G C-RAN transmission link bandwidth resources and CU pool computing resources,this paper mainly focuses on guaranteeing the service quality of the three main businesses in 5G network.A 5G C-RAN architecture based on Network Slice technology and Network Functions Virtualization(NF V)technology was proposed and analyzed and modeled.Then,a resource allocation scheme based on Advantage Actor-Critic and DPPO is proposed respectively to jointly optimize the distribution of bandwidth resources of the middle link and computing resources in the CU pool.According to the system model and algorithm structure,the state space,action space and reward function of deep reinforcement learning are given.Finally,the two proposed joint allocation schemes are verified by simulation,and compared with the fixed allocation strategy scheme.The simulation results show that the two proposed joint allocation optimization algorithms based on deep reinforcement learning can converge effectively,and have better performance than the fixed allocation strategy scheme.