Investigation On Precoding Algorithm Designs For Millimeter Wave Communication Systems

In order to achieve mobile communications with higher data rate and low-er latency,academia and industry have conducted extensive and deep research on the fifth generation communication networks(5G).As key technologies of 5G in physical layer,millimeter wave(mmWave)communications and massive multiple-input multiple-output(MIMO)can significantly improve the spectral efficiency of communication systems,and precoding plays an important role in the combination of these two technologies.Analog precoding and hybrid precoding are general precoding methods for mmWave communications.For precoding algorithms,performance and complexity are contradictory each oth-er.As a result,it is necessary to make a trade-off between these two factors.Currently,the study of mmWave precoding algorithms is inadequate.For in-stance,some algorithms are able to achieve near-optimal performance,but the complexities of them are relatively high;some precoding algorithms can on-ly be applied in specifically scenarios;and some algorithms provide mediocre spectral efficiencies,which may not satisfy the requirements of ultra-high-speed wireless communications,etc.Therefore,it is necessary to investigate the de-sign of mmWave precoding algorithms further.According to the characters of mmWave and massive MIMO,this thesis provides several designs for differ-ent precoding systems.Our works achieve certain progress in improving the spectral efficiency or reducing the complexity.The research contents of this thesis are shown in the follows.Firstly,to reduce the number of RF chains in mmWave communication sys-tem,phase shifters are used for analog precoding.Since phase shifters can only adjust the phase of signals,an extra unit-modulus constraint is introduced into precoding problem.Thus,based on single-stream communication system,we research the solution for analog precoding problem,and propose dominant vec-tor approaching(DVA)optimization method.Then,we prove its convergence and local optimality,and show the simulation results.Secondly,we investigate the analog precoding designs for multi-stream mmWave communication systems.The optimizaion problems are formulat-ed based on analog precoders with the fully-connected Structure(FS)and the partially-connected Structure(PS),respectively.In the optimization process,the non-convex precoding problem about large scale matrix is decomposed in-to a series of similar sub-problems,each of which only focuses on one vector in analog precoding matrix.Then we use DVA method to solve the problem.Thus,DVA-FS and DVA-PS algorithms are proposed as the precoding design-s.Through simulations,we show that the proposed algorithms achieve good performance,and they work well in the scenarios with imperfect channel state information or finite resolution phase shifters.Considering the complexity and performance together,we find that compared with classical orthogonal match-ing pursuit(OMP)algorithm and successive interference cancelation(SIC)al-gorithm,the proposed algorithms can achieve better performance with similar complexity.Thirdly,based on the hybrid precoding system with the fully-connected structure,we study the design of hybrid precoding algorithms.By analyzing the existing hybrid precoding algorithm based on manifold optimization,we point out that there are some redundancies in the computations,which generate high complexity.Thus,using the idea of manifold optimization and conjugate gradient search,we design an improved hybrid precoding algorithm with rel-atively low complexity.The simulation results and complexity analysis show that compared with the original algorithm,the proposed algorithm can achieve similar performance with lower computational complexity.Then,in the opti-mization,we express the hybrid precoding matrix as the sum of a series of sub-matrices,wherein each independent sub-matrix is the product of a column in the analog precoding matrix and the corresponding row in the digital precoding ma-trix.Thus,the hybrid precoding problem is transformed into the optimization problems about sub-matrices.Using proposed phase pursuit method,the sub-problems can be solved easily.By performing iterations,we obtain the com-plete steps of hybrid precoding algorithm based on phase pursuit method.The simulation results show that the proposed algorithm can achieve near-optimal performance with low complexity,and it provides a recommendable trade-off strategy between system performance and computational complexity.Finally,for the hybrid precoding system with the partially-connected struc-ture,we propose two algorithms.One of them is biased towards performance.Under the partially-connected structure,the difference between the hybrid pre-coding matrix and the optimal matrix is relatively large.To achieve better per-formance,we intend to optimize the analog precoding matrix and the digital matrix in two stages.For the analog precoder,we can neglect zero elements.Thus,the problem is decomposed into some sub-problems,each of which aims to find a sub-vector to maximize the values of quadratic forms.They can be solved iteratively.With the assist of equivalent channel,the digital precoding matrix can be obtained.The simulation results demonstrate that the proposed algorithm is able to achieve good performance compared with similar type of algorithms.The other one algorithm is more focused on complexity.In the analog precoding matrix under the partially-connected structure,the effective sub-vectors are arranged diagonally.Therefore,we divide the analog precod-ing matrix into several blocks.Thus,the optimization about large scale matrix is converted into the optimizations about smaller blocks,which are easier to be solved.The proposed algorithm achieves general performance,but its com-plexity is reduced significantly.Hence,it has higher computational efficiency.