Research On Low-cost Massive MIMO Wireless Transmission Theories And Schemes

Mobile communications have been advanced to penetrate into every aspect of daily life.The boom of mobile social networks and mobile multimedia services raises the demand for higher data rate and various service types.In order to satisfy the increasing requirement,massive multiple input multiple output(MIMO)technique was proposed.By deploying a large number of antennas at the base station,massive MIMO fully exploits the spatial degrees of freedom and improves spatial diversity and multiplexing,thereby significantly enhancing the spectrum efficiency.Massive MIMO has been proven its effectiveness and put into primary commercial use.However,before its large-scale deployment,massive MIMO still faces the problem of high-cost which is resulted from,for instance,massive expensive radio-frequency chains,dense employment of massive MIMO base stations,high overhead of channel estimation,and timeconsuming estimation algorithms.In this paper,we address the above problems by reducing the hardware cost,resource overhead,and complexity of algorithms,and study the low-cost design of massive MIMO transmission schemes.First,we study the discrete Fourier transform(DFT)-based hybrid beamforming multiuser system,which tackles the problem of massive expensive radio-frequency chains by reducing the hardware cost.We aim to design the analog beam selection schemes.The approximations of uplink ergodic achievable rates when using the zero-forcing and maximum-ratio combining receivers and those of using zero-forcing and maximum-ratio transmitting precoders are initially derived under the assumption that the beams are fixed.Based on the analytical results,the optimal beams are selected to maximize the ergodic achievable rate.Specially,a two-step beam selection algorithm is proposed to find the near-optimal beams within a short time.Numerical results confirm the tightness of the approximations of the ergodic achievable rates and reveal the effectiveness of the proposed two-step beam selection algorithm.Second,we study the large intelligent surface(LIS)-assisted massive MIMO system,which tackles the problem of dense employment of massive MIMO base stations by reducing the hardware cost.The transmission scheme of base station is initially studied in order to maximize the ergodic spectral efficiency of the LIS-assisted massive MIMO system.By formulating tight upper bounds of the ergodic spectral efficiency when using different precoders,we conform the optimality of targeting power to both the original and the supplementary links.Then,the phase shift scheme of LIS is studied.In particular,we propose an optimal phase shift design based on the upper bound of the ergodic spectral efficiency and statistical channel state information.Furthermore,we derive the requirement on the quantization bits of the LIS to promise an acceptable spectral efficiency degradation.Numerical results verified that 2-bit quantization is enough to guarantee the spectral efficiency degradation within 1 bit/s/Hz.Afterwards,we investigate the frequency-division-duplex(FDD)multi-antenna system and propose an uplink-aided downlink channel reconstruction scheme,which tackles the problem of high overhead of channel estimation by reducing the resource overhead.On basis of the spatial reciprocity between uplink and downlink channels,frequency-independent parameters,i.e.,delays and angles of paths,are estimated during uplink sounding.Then,the necessity to estimate the attenuation factors of paths in the downlink is verified.We estimate the attenuation factors through downlink training and feed them back to the base station with limited pilot and feedback overhead.We introduce and enhance the Newtonized orthogonal matching pursuit(NOMP)algorithm to detect the delays and gains in a multi-antenna multi-subcarrier condition.The results of our analysis show that the enhanced NOMP algorithm achieves high estimation accuracy.Simulations and over-the-air tests are performed to assess the performance of the uplink-aided downlink channel reconstruction scheme,and results show that the reconstructed channel is close to the practical channel.Next,we study the application of the uplink-aided downlink channel reconstruction scheme in FDD 3-dimensional(3D)multiuser massive MIMO systems.An efficient downlink training scheme for 3D multiuser systems is proposed to tackle the problem of high overhead of channel estimation by reducing the resource overhead.We propose an e NOMP algorithm to extract the frequency-independent parameters,including the delays,downtilts,and azimuths,simultaneously.An efficient downlink training scheme is then proposed for the estimation of downlink attenuation factors of multiple users.This scheme uses cell-common pilots instead of userdedicated pilots by adopting a codebook,thereby offering an acceptable estimation error rate of the attenuation factors with a limited pilot amount and guaranteeing the low-cost reconstruction of multiuser downlink channel.Numerical results verify the precision of the e NOMP algorithm and demonstrate that the spectrum efficiency of the system using the reconstructed downlink channel can approach that of the system using perfect CSI.Thereafter,we face the FDD multiuser massive MIMO system and propose a downlink time-varying channel tracking scheme,which tackles the problems of high overhead of channel estimation and time-consuming receiver algorithm simultaneously by reducing the resource overhead and the complexity of algorithm.In the uplink,with the tracing results in previous instances,we can efficiently trace the parameters of the existing paths,find paths that disappeared,and detect new paths from sparse uplink sounding reference signals.In the downlink,the downlink attenuation factors for each user can be estimated with a limited amount of pilots.The K-means clustering algorithm is utilized to design the downlink pilots and restrict them within a predefined amount.Numerical results validate the effectiveness and robustness of the proposed tracking scheme,and suggest that we should frequently apply this scheme to guarantee a high multiuser sum spectral efficiency performance.Finally,we face the FDD extremely large-scale massive MIMO system and propose a model-driven deep learning-based downlink channel reconstruction scheme,which tackles the problem of time-consuming receiver algorithm by reducing the complexity of algorithm.Considering the spatial non-stationarity in extremely large-scale massive MIMO systems and the acceleration of channel reconstruction,we introduce a powerful neural network for object detection to enable the fast detection of angles and delays and the identification of spatial non-stationarity.A low-complexity algorithm-based refiner further refines the estimates toward high accuracy.Numerical results show that the proposed scheme can rapidly and accurately reconstruct the non-stationary downlink channel.When reduced to stationary scenarios,the proposed scheme is proven to achieve comparable reconstruction accuracy as the enhanced NOMP-based method with one hundred percent of time consumption.