Research On Wireless Resource Allocation Method Based On Deep Learning In C-RAN

In order to reduce system service delay and improve user experience,Cloud Radio Access Network(C-RAN)has become the core technology for real-time service in 5G networks.Although the C-RAN architecture has great advantages in terms of cost,capacity,and flexibility,it also brings some technical challenges.For example,issues such as computing resource management and user association under the C-RAN architecture need to be resolved.Therefore,how to manage wireless resources more intelligently and flexibly and realize the superiority of the C-RAN architecture is a very important research topic.In response to the above problems,this article did the following research work:First,it introduces the technical bottlenecks currently encountered by 5G networks,leads to the C-RAN scenario studied in this article,and explains the advantages of C-RAN architecture in 5G networks and the development process and advantages and disadvantages of C-RAN architecture.The related technologies under the existing C-RAN architecture are described in detail.Based on the above description,several C-RAN architecture problems have been proposed.Secondly,for the downlink C-RAN scenario with multiple users and multiple RRHs,using deep reinforcement learning algorithms,the problem of user association and computing resource allocation with service delay as the optimization goal is studied.Using the Markov model for modeling,the framework integrates the signal-to-dry ratio state,the BBU computing resource state,and the RRH cache state into the system state,and defines the optimization objective to minimize the "system service delay" as the reward function.Enter the system state into the deep reinforcement learning model to get the system action,and train the deep reinforcement learning model with the goal of maximizing the long-term cumulative reward of the system.When the system rewards tend to be stable,the model training is completed,and user association and BBU computing resource allocation are performed according to system actions.In order to speed up the training of deep Q network(Deep Reinforcement Learning,DRL)models,the models are improved using Dueling-DQN and Double-DQN.The simulation results verify the effectiveness of the proposed user association scheme and computing resource allocation strategy in the multi-user C-RAN system,which can effectively reduce the system service delay.Finally,for the C-RAN uplink transmission system with Mobile Edge Computing(MEC)server,using deep reinforcement learning algorithm,the user computing task offload and resources optimized for service delay and user equipment energy consumption are studied Distribution problem.The user calculation task can be divided into two parts according to the actual demand according to the proportion,one part is offloaded to the MEC server and executed according to the proportional parameter,and the other part is executed on the user equipment.In order to train the deep reinforcement learning model,the computing task size and the required computing resources are integrated into the system state,and the optimization objective to minimize the "system service delay and equipment energy consumption" is defined as the reward function.After the deep reinforcement learning model training is completed,the system state is output to the algorithm scheme to obtain the system action,that is,the calculation task offload scheme and the calculation resource allocation strategy.The simulation results verify the effectiveness of the proposed offloading scheme and resource allocation strategy in the multi-user MEC system,which can effectively reduce system delay and equipment energy consumption costs.