Squared channel matrix prediction for CSI reporting in Rayleigh fading channels

The need for accurate and relevant CSI feedback is one of the pillars for the efficiency of cellular systems. However, one problem facing CSI reporting is the effect of CSI aging such that the reported CSI information is obsolete when data is scheduled based on it. CSI may instead be predicted in order to mitigate some of the aging loss. In doing so two things need to be considered. First, the CSI prediction needs to have low bias also for cases where prediction is hard, such as short channel coherence time or infrequent CSI reporting. Second, it is desirable not to predict the interference part of the channel quality, as the interference is in the control of the network and hence there is no need for the UE to predict it. This paper addresses the CSI prediction problem taking these two aspects into consideration and proposes a method based on squared channel matrix prediction.