Covariance-Based Signaling and Feedback Data Parameterization for the Time-Varying MIMO Broadcast Channel

Linear precoding (beamforming) techniques that maximize the sum rate in the multi-antenna broadcast channel often suffer severe performance degradation when the channel state information at the transmitter is erroneous or outdated. This paper presents a linear precoding algorithm that uses the channel statistics rather than channel state to provide more stable performance in time-varying environments. This stable beamforming method, however, requires that the transmitter know the full spatial correlation matrix, significantly increasing the amount of data communicated over the feedback channel. Parameterization of the spatial correlation using popular channel models is therefore considered to reduce the size of the feedback data stream.