Research On Self-optimization Of Full-duplex Small Cell Network Based On Reinforcewment Learning

With the advantage of utilizing radio spectrum resources efficiently,small cell plays an important role in further communication network.The introduction of small cell in the communication improves network capacity,optimizes network coverage,and provides users with a better experience.Because the location of the small base station is random,and the small base may join or leave the network at any time.Therefore,the small base station network must have the ability to organize itself.Self-organization is divided into self-configuration,self-optimization and self-healing.This paper mainly studies the self-optimization of full duplex small base station network,especially the interference coordination.The use of full duplex transmission mode in small base stations is mainly to improve the utilization of spectrum.The main work of this thesis is divided into three parts:(1)Establish the interference model of the full duplex small base station network and analyze it.The interference coordination problem into ensure the QoS of macro cell and macro cell users and maximum capacity of small base station communication problems,solve the problem through channel allocation.The channel allocation process of the full duplex small base station network is divided into two stages: the user pairing of the small base station and the channel allocation.The two phases are closely combined to affect the performance of the system.(2)Considering the full duplex transmission mode of small base stations,two different users in the same small base station transmit simultaneously on the same channel.Different combinations of users produce varying degrees of co-channel interference.In this paper,a user pairing strategy based on Kuhn-Munkres is adopted to minimize intra-cell interference.(3)Allocation channel for macro users and paired small cell users based on Q-learning.Coordinate the cross-tier interference and inter-cell interference of the full duplex small base station network and maximum network communication capacity.On this basis,a cooperative distributed Q-learning algorithm is proposed,which improves the convergence speed of the algorithm and enables the system to achieve a balanced state faster.Experiments show that the user pairing method based on Kuhn-Munkres algorithm and the channel selection method based on Q-learning can effectively improve the communication capacity of full-duplex self-organized small cell network.