Feedback Reduction for MIMO Broadcast Channel with Heterogeneous Fading

Abstract: This paper considers feedback load reduction for multiuser multiple input multiple output (MIMO) broadcast channel where the users' channel distributions are not homogeneous. A cluster-based feedback scheme is proposed such that the range of possible signal-to-noise ratio (SNR) of the users are divided into several clusters according to the order statistics of the users' SNRs. Each cluster has a corresponding threshold, and the users compare their measured instantaneous SNRs with the thresholds to determine whether and how many bits they should use to feed back their instantaneous SNRs. If a user's instantaneous SNR is lower than a certain threshold, the user does not feed back. Feedback load reduction is thus achieved. For a given number of clusters, the sum rate loss using the cluster-based feedback scheme is investigated. Then the minimum number of clusters given a maximum tolerable sum rate loss is derived. Through simulations, it is shown that, when the number of users is large, full multiuser diversity can be achieved by the proposed feedback scheme, which is more efficient than the conventional schemes.