Deep Reinforcement Learning Based Resource Allocation For Wireless Network

With the rapid development of wireless communication technology,the demand for the mobile communication increased rapidly in the last few years.The radio resource is more and more sparse.In order to solve these problems,new network architectures capable of coping with the current network environment is needed.Small cell can make up for the shortage of macro cell networks,and can solve the network coverage problem of blind areas and weak coverage scenarios,which has important research significance.Therefore,small cell networks have important research significance.In addition,the current survey shows that data services have replaced voice services as new mainstream services.Data traffic has grown rapidly and users' requests for data services have focused on a few hot contents such as online video and news headlines.In order to alleviate the transmission pressure of the wireless channel,the experts propose caching-enabled network architecture for the case where the same content is repeatedly transmitted.Therefore,the research on wireless resource allocation in a cloud-based caching-enabled small cell network will have important theoretical and practical significance.Based on the downlink of a cloud-based caching-enabled small cell network,this paper proposes a radio resource allocation algorithm based on deep reinforcement learning to maximize network throughput.First of all,the paper discusses the related concepts of CSCN and analyzes the network throughput of small base stations in the architecture.Then,the long short-term memory network is used to predict the user's mobile mode,and the conditions for selecting the user of the small base station are optimized according to the user's mobile mode.And then,the concept of game theory is introduced to model the problem of maximizing network throughput as a multi-agent non-cooperative game problem.Finally,in order to maximize the network throughput,a wireless resource allocation algorithm based on deep reinforcement learning is proposed,which enables small base stations to learn autonomously and select the wireless channel based on the network environment.The simulation results show that compared with the traditional random access algorithm,the proposed algorithm makes the network throughput significantly improved.