Research On Cooperative Multipoint Transport Based On Reinforcement Learning

In order to meet the increasingly high requirements for wireless communication rate,network delay and service quality,the commercial application of the fifth generation(5G)mobile communication system is approaching.As a key technology of 5G communication,ultra-dense networking has become the focus of the industry.At the same time,cooperative multipoint transmission(CoMP)is a hot research topic in recent years in order to reduce the signal interference caused by ultra-dense deployment.Firstly,this paper studies the clustering problem of CoMP cooperative community,and according to the geographical location of the community and the distribution of mobile users in the community,it selects the appropriate community to form a cooperative cluster to provide services for users.Then,aiming at improving system capacity and reducing data loss rate,the energy collaboration between base stations and the energy allocation between base stations and users are studied.It includes the following two aspects.In terms of cooperative cell selection for multi-point cooperative transmission,this paper proposes a semi-dynamic clustering scheme based on reinforcement learning.In order to solve the situation that the system burden caused by dynamic clustering is too large,all cells are first divided into many pre-cooperative sets according to the geographical location of the cells.In order to solve the problem of user mobility and time-varying characteristics of the system,in each pre-collaboration set,users are dynamically allocated with service cells and power according to channel state and user location,which effectively improves the system throughput and service quality of edge users.In the aspect of energy management of multi-point cooperative transmission,this paper proposes an energy collaboration and allocation strategy based on deep reinforcement learning algorithm.First according to the situation of relay stations of battery capacity,energy capture and amount of data in the cache,Markov decision model is established,and then use the DQN algorithm to get the optimal coordination strategy,to maximize the system throughput,and through a collaboration of energy makes up for the green energy to capture the system caused by the uneven life is low,improve the network performance.