Adaptive Schemes for Spatio-Temporally Correlated Multiuser MIMO Channel Estimation

Abstract

The class of least square and least mean square adaptive algorithm are respectively known for excellent convergence rates and simplicity. In this work, we use these algorithms to estimate spatio-temporally correlated MU-MIMO downlink wireless channels. We present FBLMS and FBRLS algorithm which achieve good performance at a low complexity. Frequency division duplex (FDD) MU-MIMO systems with a large number of antennas have the tendency to utilizes the downlink resources very well but also increase the overhead required to estimate the channel. In order to reduce this feedback burden, we quantize the feedback into 2-bits and utilize SOI-Kalman and signed adaptive algorithms to estimate the channel and compare its performance with analog feedback case. We propose a joint data and pilot transmission scheme to reduce the loss of throughput in training period which methodically transmits data to a group of users in training period and data to the rest. This scheme becomes interference limited hence sumrate results are compared with conventional approach in both analog and quantized feedback cases considering the spatial correlation and zero-forcing beamforming. The temporal correlation is exploited to reduce the amount of feedback required and tracking the variations in channels. This Channel side information(CSIT) acquire using these schemes is imperfect hence it becomes very important to assess their performance. We derive sumrate, mean square error and outage probability expressions and compare the results of all the schemes for both conventional and proposed scheme using quantized and analog feedback.