Research On Channel Modeling And Channel Prediction In Massive MIMO System

Massive MIMO is the technology to be used in the next generation wireless communication system(5G),which has the advantages of high spectral efficiency and high throughput.As one of the hotspots in the Massive MIMO communication system,the ability of channel transmission information can be obtained through the channel model.In the actual communication system,the channel state information at the transmitters is outdated due to channel time-varying or relative motion.Channel prediction technology can effectively solve the above problems.Therefore,channel prediction technology has became another research hotspot in Massive MIMO systems.In this thesis,the fundamental structure,channel modeling and channel prediction technology of Massive MIMO system are studied.The main work and results are as follows:1.The existed channel models in MIMO systems have been studied.We have studied the new characters in the Massive MIMO system,such as different receiving power,offset angle of the antenna element at the transceivers,and the far-field assumption has no longer been approved due to large-scale antennas are used in the Massive MIMO systems.The traditional MIMO channel models are not the same with Massive MIMO.Then,a channel model based on power attenuation matrix considering the size of the antenna array is proposed.We have applied this power attenuation matrix to the existed 3GPP 3D channel model to representing the Massive MIMO channel model.And the proposed channel is verified by simulation.The correctness of statistical properties such as spatial cross-correlation,time autocorrelation and condition number of channel matrix of Massive MIMO channel model have been verified by MATLAB simulation.2.A Massive MIMO channel prediction scheme based on multi-task learning is proposed.In the Massive MIMO communication systems,the channel will be presented frequency selectivity and time-varying characteristics due to multipath propagation phenomena and Doppler spread,resulting in outdated CSI information at the transmitters.In order to resolve this prblem,a Massive MIMO channel prediction scheme based multi-task is proposed.The multi-antennas used in Massive MIMO channel and the multi-tasking framework can be perfectly matched.By using the similarity of CSI information between different pairs of transmitting and receiving antennas,the CSI of transceiver pairs can be combined and studied together,so that the channel prediction performance is better.In this chapter,the performance of channel prediction scheme based on multi-task learning for single-step prediction and multi-step prediction is investigated respectively.The simulation results show that the scheme can have lower prediction error under the same experimental conditions.3.Massive MIMO channel prediction scheme combined spatial correlation is studied.In Massive MIMO systems,due to the limitation of spatial distance between antenna pairs caused by large-scale antenna array,the spatial correlation between antenna pairs is introduced into the channel prediction model.The performance of channel prediction is improved through the mutual promotion of spatial correlation among various antenna pairs.Based on what have been said above,the spatial correlation characteristics between the antennas have been added to the channel prediction model based multi-task,and the pairs of transceiver antennas are more closely combined using the framework of multi-task learning.After simulation,it is found that the algorithm has lower prediction error than single task scheme when it has higher spatial correlation characteristics.