Beamspace Aware Adaptive Channel Estimation for Single-Carrier Time-varying Massive MIMO Channels

摘要

In this paper, the problem of sequential beam construction and adaptivechannel estimation based on reduced rank (RR) Kalman filtering forfrequency-selective massive multiple-input multiple-output (MIMO) systemsemploying single-carrier (SC) in time division duplex (TDD) mode areconsidered. In two-stage beamforming, a new algorithm for statisticalpre-beamformer design is proposed for spatially correlated time-varyingwideband MIMO channels under the assumption that the channel is a stationaryGauss-Markov random process. The proposed algorithm yields a nearly optimalpre-beamformer whose beam pattern is designed sequentially with low complexityby taking the user-grouping into account, and exploiting the properties ofKalman filtering and associated prediction error covariance matrices. Theresulting design, based on the second order statistical properties of thechannel, generates beamspace on which the RR Kalman estimator can be realizedas accurately as possible. It is observed that the adaptive channel estimationtechnique together with the proposed sequential beamspace construction showsremarkable robustness to the pilot interference. This comes with significantreduction in both pilot overhead and dimension of the pre-beamformer loweringboth hardware complexity and power consumption.