Research On Channel Prediction Methods In Fast-varying Massive MIMO Environments

Different from the conventional multiple-input multiple-output(MIMO)system with normal-size antennas,as the increase of the number of antennas especially at the base station terminal,massive MIMO channel has two particular channel characteristics.Specifically,the massive MIMO channel is characterized by fast-variation and non-stationarity.However,outdated channel state information(CSI)will be acquired because of the fast-variation,thus seriously degrade the system performance.In this circumstance,reliable channel prediction according to its several past known channel observations is very necessary,which can mitigate the effect of outdated CSI effectively.In this paper,we have a research on the channel prediction methods in fast-varying massive MIMO environments.First,considering the channel properties,a first-order Taylor expansion(FIT)channel modeling has been proposed.Then,based on the FIT model,a new channel prediction approach is proposed,which consists of the channel estimation stage and channel prediction stage.And the interval of effective prediction(IEP)during which period this predictor can be adopted is derived.The performance of the proposed channel prediction method is testified by Monte Carlo numerical simulations.It has been shown that,within the IEP,a reliable channel prediction can be obtained with low computational complexity,which then improves the communication system's performance effectively.To improve the problem that the predicted channel coefficient may suffer from flutter interference and vibrates with fluctuations,we propose a more sophisticated symbol detection aided channel prediction transmitting sequence structure in this paper.This framework which is aimed to reduce the overhead of the pilots utilizes the detected data signals,and regards them as pilots to aid the channel prediction,which are consequently named as virtual pilots.Besides,to further investigate the performance of the proposed channel prediction method,we derive a Cramer-Rao bound(CRB)of the channel prediction error theoretically and analyze the system's bit error rate(BER)indicator adopting this channel predictor when the zero-forcing receiver is implemented at the receiver.At last,by comparing the theoretically and analytical results with the simulation values,the simulations show that our analytical expressions have a close match of the real simulated performance under various simulation parameter settings,such as different number of antenna settings,different signal to noise ratio value and so on.Moreover,the simulation results indicate that when adopting this symbol detection aided channel prediction scheme,the frequency of sending the pilot signals can be significantly reduced.Also,the application of this proposed scheme can be further expanded when combining it with channel coding such as low density parity check(LDPC)coding,thereby greatly improving the spectrum efficiency of the system.