Research On Detection Algorithm And System Verification For Intelligent MIMO Reflection System

With the development of modern society,wireless communication technology has entered the 5G area.Massive multiple-input multiple-output(MIMO)technology is the key technology in the 5G area.For massive MIMO,the number of transmit and receive antennas is relatively large,which brings performance advantage and increases the complexity of signal detection.In the traditional detection algorithms,the maximum likelihood(ML)signal detection algorithm with the best performance has high computational complexity.With the increase of the number of transmit antennas and the modulation dimension,the complexity of the ML detection algorithm increases exponentially.Some linear detection algorithms with low complexity are affected by complex channel environment,channel estimation error,and a large number of nonlinear factors.The performance is far from ML detection algorithm.Intelligent reflecting surface(IRS)is a new communication method for post-5G area.Configuring smart reflectors in MIMO communication systems is a new solution to improve the performance requirements of wireless networks.An intelligent reflecting surface is a planar array composed of a large number of reconfigurable passive elements.Through an intelligent controller,each element can independently generate a certain amount of phase shift on the incident signal,thereby changing the propagation of the reflected signal.Although the intelligent reflecting surface-assisted MIMO communication has potential performance improvement and application advantages,because of the introduction of the intelligent reflecting surface phase matrix,the parameter dimension is larger and the complexity of the detection problem of intelligent reflecting surface-assisted communication is higher.Deep Learning(DL)methods utilizing neural networks have been successfully applied in various scenarios.Deep learning aims to obtain the feature information of each layer through a layered network to solve the important problem that requires manual design in the past.It is a method based on representation learning of data.With its unique training principle,deep learning does not need to understand the specific working mechanism which only needs a suitable network and enough training data to solve the problem.It has shown great potential in dealing with nonlinear problems.In this thesis,the deep learning method is considered for signal detection of MIMO communication system and intelligent reflecting surface-assisted MIMO communication system.In addition,the demonstration platform of the new wireless communication system based on intelligent reflector is implemented.Firstly,this thesis studies the classical MIMO communication system.The second chapter of the thesis introduces the classical MIMO detection algorithm.The thesis compares and analyzes the performance and complexity.Then,the thesis introduces the basic principles of deep learning and neural networks,explaining the gradient optimization algorithm of neural networks and common neural networks.The third chapter of the thesis studies the intelligent MIMO detection problem based on meta-learning,which introduces the black-box-based deep learning detection and the deep learning-based sphere decoding algorithm,and analyzes the simulation results.In addition,the thesis introduces the idea of meta-learning.For the K-best algorithm of the near-optimal tree search detection algorithms,the fixed value of K in each layer is canceled,and the value of K is set as a variable.The thesis designs the fitting function form of K according to the law of change.On this basis,a meta-learning parameter learning network based on network fusion is designed to learn the parameters of the fitting function,which solves the problem that K is difficult to learn directly.The deep learning network is used to replace the path selection in the classical algorithm.The simulation results verify the advantages of the proposed signal detection network compared with traditional algorithms in terms of performance and complexity.Secondly,this thesis studies the intelligent reflecting surface-assisted MIMO communication system.The fourth chapter introduces the classical iterative expansion detection network and its improvement,analyzing the simulation results.Combined with the characteristics of the intelligent reflecting surface-assisted MIMO communication system,a new deep learning detection network for intelligent reflecting surface-assisted communication is constructed through the idea of the gradient descent algorithm.Different from the traditional fully connected unit structure,the detection network strengthens the influence of some inputs by adding a shortcircuit direct connection structure,solves the problem of gradient divergence,and reduces the complexity of the detection network by adjusting the weight coefficients of different units.The simulation results analyze and compare the performance of the deep learning detection network and traditional algorithms.In addition,the thesis illustrates the advantages of the detection network in terms of complexity and network structure.Finally,this thesis also studies a demonstration platform for implementing a new wireless communication system based on intelligent reflecting surface using VIVADO FPGA.The fifth chapter of the thesis uses an IRS to replace the role of the transmitter radio frequency and antenna in the traditional communication system,and designs a new wireless communication system with intelligent reflective surface.The thesis firstly introduces the hardware modules of the software radio platform,and analyzes the performance parameters of each module.Then,the communication process in the communication system is introduced,the physical diagram of the demonstration platform of the new wireless communication system based on the intelligent reflecting surface is given,and the video transmission process is introduced.Finally,the function of the new wireless communication system based on the intelligent reflecting surface is verified through numerical simulation and actual environment test,and the feasibility of video transmission based on the communication system based on the intelligent reflecting surface is proved.