Modelling Research On Massive MIMO Radio Channel Based On Measurements

Massive multiple input and multiple output(Massive MIMO),as a key technique for the emerging 5th generation(5G)wireless communications systems,presents a highly promising solution to meet the demanding requirements of spectrum efficiency and the energy efficiency.This thesis researches on the radio propagation characteristics of Massive MIMO.The realistic measurements are conducted in the typically indoor and outdoor,LOS and NLOS scenarios respectively with the bandwidth of 91 MHz,100MHz and 200MHz of the carrier frequency ranging from 1GHz to 6GHz.The channel propagation characteristics in spectrum and space domain are investigated by employing the high precision parameter estimation algorithm.Then a non-WSSUS(non-wide sense stationary un-correlated scattering)channel analytical model is proposed,followed with a simulation model of scattering cluster over the antenna array based on the Markov chain to build up the propability distribution of the scattering cluster birth-and-death process.The whole research work aims to contribute the radio channel modeling of the 5G Massive MIMO communication system.At first,the wideband Massive MIMO channel sounding is introduced.A multi-element virtual linear structure antenna array is employed and a rubidium clock standard trained by the GPS(Global Position System)is used to maintain the time synchronism.To improve the performance the antenna pattern is calibrated in anechoic chamber,at the same time,high-speed memory data acquisition card and real-time data analysis software are employed.At second,based on the channel sounding,we collect a rich supply of in-field data and extract the channel patameter,including the Power delay profile,RMS delay,angle of departure,angle of arrival and focusing on the spatial correlation properties.The results prove that the widely accepted WSSUS channel condition is not hold.The third,a survey of the communication channel modeling research is present.The emphasis is put on the technology and model of state of the art.With the conclusion of non-WSSUS,we develop a channel analytical model based on the traditional Kronecker model,fully considering the spatial correlation of the sub-channels.The novel model is supposed to be nonstationary within the whole antenna array,while stationary limited in one antenna set covered by same clusters.Since the whole antenna array is composed of antenna sets that the channel correlation matrix is piecewise stationary.The result is according to the theory of the visibility region of scattering clusters.The fourth,big data mining and machine learning is adopted to improve the efficiency and precision of the multipath clustering algorithm.Here we employ the KPowerMeans method.But we abandon the traditional mechanism of random initialization,and alternate with an arithmetic namely sequential search arithmetic,which helps to avoid the risk of convergence trouble that leads to locally optimal solution.In real propagation environment,due to the large physical size of the antenna array,the scatterer visibility region is not idealy cover the whole array,which leads to the non-statinary of the channel.Within the visibility region of one scatterer,the sub-channel can be treated as independent and identical distributed.The novel model is reasonable considering the correlation of sub-channel of the neighboring elements,which dovetail with the theory of the KPowerMeans algorithm.They all root in the channel matrix singular.At last,in the thesis,we build up a Markov chain to model the scattering cluster over the antenna array in Massive MIMO system,according to the theory of cluster's visibility region.We bring insight deeply into the cluster evolution over the antenna array axis.The thesis systematically demonstrates how does the spatial correlation influence the capacity of the communication system,and establishes a theoretical foundation for the research work of Massive MIMO.