Downlink Transmission Design For Reconfigurable Intelligent Surface Assisted MmWave Communication System

With the continuous development of wireless communication technology,the requirement of achievable rate for communication systems is getting higher and higher.In the Fifth Generation of Mobile Communication Systems(5G),Millimeter wave(mmWave)communication technology,as one of the key supporting technologies,has received wide attention in recent years.Millimeter wave frequency band resources are abundant and can significantly improve the performance of communication systems.However,the high mmWave frequency band,high transmission loss,and ease of being blocked by obstacles are not conducive to wide area coverage.The introduction of reconfigurable intelligent surface(RIS)is expected to solve the coverage problem of mmWave communication systems.The RIS consists of a large number of low-cost passive devices,and the parameters of these devices can be changed to achieve modulation of the reflected signal.Also,by adding an additional signal propagation link,the RIS helps to solve the obstacle occlusion problem.However,the key to the performance improvement of mmWave communication systems assisted by RIS lies in the proper design of the phase shift matrix of the reflective surface,i.e.,the passive beam assignment of the reflected signal.Existing schemes often use the joint design of active and passive beamforming,which is limited by the constant-mode constraint of reflected phase shift and has high algorithmic complexity and difficult practical deployment.To address the above problems,this thesis conducts research on the design of RIS phase shift in the downlink of mmWave system assisted by RIS with the optimization goal of maximizing the achievable rate.The specific research contents are as follows.First,based on the classical optimization theory,the phase shift design of the RIS is studied.Firstly,we establish the received signal and channel model of the system after the introduction of the RIS,and propose the optimization problem of the phase shift design of the RIS with the achievable rate as the optimization objective;then,for the problem that the form of the objective function is complex and difficult to optimize,we use the nature of the matrix eigenvalue to derive its lower bound and represent it as a function of the matrix trace,so as to convert the original optimization problem to optimize the lower bound of the matrix trace;then,the optimization problem is transformed and solved by two methods such as Successive Convex Approximation(SCA)for the nonconvex constraint of the phase shifted constant mode of the RIS;finally,the simulation results show that the proposed method can achieve a higher system achievable rate with very low complexity compared with the existing schemes in the literature.Second,based on the machine learning theory,we study the phase shift design problem of RIS.First,since the phase shifts of signals,channels and phase shift of RIS in communication systems are usually in the form of complex numbers,this thesis designs a targeted complex neural network,which can output the optimized phase shift coefficients of RIS based on the input channel state information;second,for the constant mode constraint problem of phase shifts,a specially designed activation function is introduced in the output layer of the complex neural network,which can realize the phase;then,the gradient expressions of the parameters of each layer in the backward propagation algorithm are derived;finally,the simulation results show that the proposed neural network converges quickly in training and can achieve a higher achievable rate compared with the classical optimization theory-based algorithm.