| In the new era of rapid development of the Internet,in order to meet people’s increasing network demand and the emerging Internet of Things industry with low latency,high reliability,wide coverage,a new generation of wireless network is also in full swing research.As one of the pivotal research directions of the next generation wireless communication system,extremely large-scale MIMO technology is expected to acquire higher spectral efficiency,wider coverage and provide lower latency and more reliable data networks.However,the use of extremely large-scale array will bring different channel characteristics: near-field effect and spatial non-stationary characteristics,which bring great challenges to the design of channel estimation scheme.Therefore,this paper puts forward a new channel estimation method for the near-field effect and spatial non-stationary characteristics of the extremely large-scale MIMO system.Firstly,we analyzed the near field effect and spatial nonstationarity of wireless channel in extremely large-scale array.When the antenna scale is large,the increase of Rayleigh distance will cause the scatterer and user to be located in the near field region of the large-scale array,resulting in the near field effect.In this paper,the parabolic wave which is spherical wave’s second-order approximate,is used to model the near-field effect in the extremely large-scale MIMO system.The model is simpler than the spherical wave model,but more accurate than the plane wave model.In addition,when the array scale is very large,the signal reflected by the scatterer cannot be reflected to all parts of the array,that is,different parts of the array may observe different terminals,resulting in spatial non-stationarity of the channel.For the spatial nonstationary characteristics of the extremely large MIMO system,we divide the extremely large array into several sub-arrays on average,and each sub-array has the same stationarity.The nonstationarity is described by multiplying a selection vector after the guide vector.The elements in the vector as 1 or 0,and set the corresponding part of the subarray visible to the scatterer to 1 and the rest of the invisible part to 0.Secondly,a subarray-based channel estimation scheme for spatial nonstationarity of the extremely large array is proposed under the near-field channel model.In the near field effect,the parabolic wave model is used in this paper.The channel guide vector is composed of two unknowns,angle and distance.Due to the special structure of the guide vector,it can be divided into two matrices,one of which is only related to angle and the other to angle and distance.Combined with the characteristics of the signal subspace,the angle is estimated through one-dimensional search,and then the estimated angle is substituted into another one-dimensional search to estimate the distance.In order to determine the stationary of channel,this article estimated angle value for every subarray,due to the same scatterer’s visible adjacent subarray angle has an increasing or decreasing order,and the adjacent subarray angle difference is not large,so we can judge scatterer visible subarray area according to the estimated angle,so as to determine the nonstationarity.The simulation results show that,the proposed algorithm can accurately estimate the angle and distance,and determine the non-stationarity of the channel,and the mean square error of the channel estimation is small.Finally,under the hybrid-field channel model,a pseudo-peak-removing channel estimation scheme is proposed.In this environment,part of the scatterers are located in the far field region and other part in the near field region.In order to simplify the model without losing generality,the parabolic wave model is still used for the modeling of far-field components,and the distance is set to infinity.The estimation of angle and distance is still the same as that of near field model.However,if there are two or more far-field scatterers in the hybrid-field channel,there will be false peaks in angle search.As to the existence of pseudo peak,one-dimensional search in estimated angle is put forward,we find out the possible existence of pseudo peak and real angle corresponding to the peak search.Because of the advent of pseudo peak is always in the middle of the two real angle,and has the certain function relation,so we can determine the real angle of channel between the pseudo peak corresponds to the angle and the real angle.Simulation results show that the pseudo-peak removal scheme proposed in this paper can get rid of the pseudo-peak and retain the real angel,and has a small mean square error in channel estimation in hybrid-field channel. |
