Enabling Covariance-Based Feedback in Massive MIMO: A User Classification Approach

Abstract

In this paper, we propose a novel channel feedback scheme for frequency division duplexing massive multi-input multi-output systems. The concept uses the notion of user statistical separability which was hinted in several prior works in the massive antenna regime but not fully exploited so far. We here propose a hybrid statistical-instantaneous feedback scheme based on a user classification mechanism where the classification metric derives from a rate bound analysis. According to classification results, a user either operates on a statistical feedback mode or instantaneous mode. Our results illustrate the sum rate advantages of our scheme under a global feedback overhead constraint.

A Covariance-Based Hybrid Channel Feedback in FDD Massive MIMO Systems

“…Moreover, due to the absence of instantaneous channel gain, channel feedback is decided by channel statistics. 31) AND (32) For the ease of analysis for system sum rate, we first rewrite the quantized channel in the form of DFT matrix V. Following the similar derivation given in (56), there exists a column vector in V which is identical or closest to the predicted channel feedback h B I,i . The index of the DFT vector satisfies η I,i ( u * I,i ) − 2 m I,i…”

“…l,l,i for the class-I user i in the l-th cell. Then, by exploiting the predicted channels and beam domain covariance matrices, approximate precoding vectors of the users in the l-th cell are obtained with the similar procedure given in Equation (31) and (32), and presented as w l,l,k = V (:, m l,l,k ) , k ∈ K l . Due to the limited space, we omit the details to obtain m l,l,k and directly present the result for a class-I user as m l,l,k = x I,i,min +…”

“…. , a M ) denotes 1 Note that the idea of hybrid feedback has been presented in our previous conference paper [32]. However, there are differences.…”