Research On Channel Estimation For Mm Wave Massive Mimo

With the rapid growth of wireless users'number and the development of Internet of things,artificial intelligence and the other technologies,there have been more potential challenges for improving the performances of wireless communication system on transmission rate,capacity and delay.To deal with these challenges,the fifth generation mobile communication system?5G?adopts millimeter-wave massive multiple input multiple output?MIMO?technology to enrich the frequency resources,improve the spectrum utilization and increase the system capacity.Moreover,for taking full advantages of the gains from the millimeter-wave massive MIMO system,the channel state information needs to be obtained by the tranceivers quickly and accurately.However,the increase of antenna size and the use of hybrid precoding technology have brought new challenges to channel estimation.Thus,this thesis focuses on the researches of the channel estimation for the millimeter-wave massive MIMO systems.Firstly,research background and the state of art of the millimeter-wave massive MIMO channel estimation are investigated,followed by the improvement directions of the current technologies.The background of millimeter-wave channel characteristics,compressed sensing and neural network theory is introduced,which lays a foundation for the subsequent researches of the channel estimation algorithms for the millimeter-wave massive MIMO system.Secondly,based on the sparsity of the massive MIMO channel in the angular domain,the system transmission model and the parameterized channel model are established.Based on these models,the implementations of both traditional and compressed sensing based channel estimation algorithms are studied.Simulation results show that the channel estimation algorithm based on compressed sensing is more suitable for millimeter-wave massive MIMO system.Thirdly,an l_1/2-regularization based millimeter-wave massive MIMO channel estimation algorithm is proposed.The algorithm constructs an objective function consisting of the weighted sum of an l_1/2-regularization term and an error constraint term,and then iteratively optimizes the parameters in the objective function with the gradient descent method.Finally,a super-resolution based channel estimation result can be achieved.At the same time,a pruning operation is introduced to reduce the computational complexity.Simulation results show that the algorithm can obtain lower channel estimation error and have better anti-noise performance than the existing algorithms.Finally,a CAMP-Net based millimeter-wave massive MIMO channel estimation method is proposed.Considering the advantages of deep learning,the iterative process of traditional approximate message passing algorithm is expanded into a CAMP-Net,which is a complex-valued deep network structure,and the piecewise linear shrinkage function is used to optimize the fitting results.In the simulation experiment,the network is trained and tested,and the results show that it is superior to the compressed sensing-based channel estimation algorithm in both accuracy and efficiency.