Self-tuning Technology Of Avoiding Beam Collision On Communication Network

With the continuous development of wireless communication technology,the high-speed growth of mobile data services and massive access requirements have put forward more demands of high capacity,low power consump-tion and low delay for the new generation communication system.Aimed to further improve capacity,the wireless communication system gradually expands the antenna array scale and makes full use of beamforming techniques to achieve high-precision transmission gain.Unfortunately,owing to dense network deployment,new challenges for wireless communication network optimization appear.Especially for the multi-antenna and multi-cell com-munication system,beam collision is inevitable,bringing out strong interference and poor network performance.Generally,beam collision refers to the phenomenon of beam overlap when different beams cover a certain area at the same time.Without the accurate definition and mathematical model,beam collision elimination still rests on perception.In addition,considering the influence of user distribution and antenna array parameters on beam orien-tation and gain,the proper design of antenna array parameters should be paid more attention from the perspective of wireless communication network optimization.Therefore,this thesis focuses on the key technologies of beam management and methods in beam collision avoidance in different scenarios driven by interference model and data in the multi-antenna and multi-cell communication system.The main work of this thesis is as follows:Firstly,the multi-user MIMO(MU-MIMO)communication system is introduced,including transmission prin-ciple and precoding techniques.Then,the dimension extension evolution and merits and demerits of beamforming techniques are discussed.For sake of communication standard,the brand-new beam management mechanism and channel state information obtainment process of 5G New Radio(NR)system are elaborated in detail.Then,taking beam kinds and collision scenarios into account,related numerical analysis methods are utilized to deeply under-stand the effects of antenna array parameters and user distribution on beam transmission gain,which contributes to the main research content of this thesis.At last,the common learning frameworks on the intelligent communication optimization are briefly introduced to support the research methods in this thesis.Secondly,as for the multi-antenna and multi-cell communication system,the clear definition of beam collision event is given,and the strong correlation between beam collision parameters and signal-to-interference-plus-noise ratio(SINR)is deduced,thus constructing the implicit mapping between beam collision and interference.Since the time scale of antenna array parameters tuning is far outweigh correlated time of channel,the optimization problem of minimizing the average total beam collision parameters is established from the point of statistical concept.With the difficulty of beam collision parameters obtainment and direct optimization of beam collision,the original optimiza-tion problem is approximately transformed into the problem of minimizing the average sum rate.Then,nonlinear interior point method and genetic algorithm are used to find out the solution.Numerical results show that the former owns fast convergence speed while overly relying on iteration point locations and holding poor performance.The latter has a certain ability to optimizing,nevertheless,the accuracy is low.Crucially,neither of the two algorithms is suitable for dynamic communication optimization,which reflects the deficiencies of traditional methods in coping with beam collision and provides feasibility for the coming proposed intelligent optimization schemes.Thirdly,the optimization of multi-cell broadcast beam patterns in millimeter wave cellular network is inves-tigated.According to the correlation of influence of SINR on beam collision and network coverage performance,and then the optimization problem of minimizing the average total collision parameters of the broadcast beams is approximately transformed into the problem of maximizing the average amount of connected users.With the great correlation between broadcast beam deployment and user distribution,an adaptive optimization algorithm of broad-cast beam pattern based on modified deep Q network(DQN)is proposed under the framework of deep reinforcement learning(DRL),and different comparison schemes are designed to analyze the performance and complexity.Sim-ulation results show that the proposed algorithm has nearly optimal average network coverage performance in the high dynamic complex environment and has good universality for different user mobility modes.Finally,the optimization of base station cluster engineering parameters in 5G cellular network is explored.For one thing,it is tough to require the accurate channel state information(CSI)in the physical communication envi-ronment.For another,beam-space channel information is closely connected with SINR and spectrum efficiency.Therefore,the problem of minimizing average total collision parameters is approximately transformed into the prob-lem of spectral efficiency maximization.Especially,the introduction of the reference-signal-receive power(RSRP)following 5G NR standard enhances actual network optimization practicability.Then,under the DRL framework,the adaptive optimization algorithm of base station cluster engineering parameters based on deep deterministic policy gradient(DDPG)network is proposed,and different comparison schemes are designed to analyze the performance and complexity.Simulation results show that the proposed algorithm can adapt well to the dynamic complex inter-ference environment and has good trade-off between high-performance and low-complexity.