In this paper, we propose efficient algorithms for estimating the signal subspace of mobile users in a wireless communication environment with a multi-antenna base-station with M antennas. When M is large, because of the high angular resolution of the receiver, any realization of the random channel vector of any given user is approximately contained in a userspecific subspace of dimension p M . Efficient multiuser MIMO schemes can be obtained from such subspace information, which is stable in time and can be accurately estimated even in the presence of fast fading (e.g., for mm-Wave channels). We are interested in the massive MIMO regime of M 1. In order to reduce the RF front-end complexity and overall A/D conversion rate, the M -antenna base-station transmitter/receiver is split into the product of a baseband linear projection (digital) and an RF reconfigurable beamforming network (analog) with only m M RF chains. Hence, only m-dimensional analog observations can be obtained for subspace estimation. We develop efficient algorithms that estimate the dominant signal subspace of the users from sampling only m = O(2 √ M) specific array elements according to a coprime scheme. For a given target dimension of the signal subspace p ≤ M , our algorithms return a p-dimensional beamformer with a performance comparable with the best pdim beamformer designed by knowing the exact covariance matrix of the received signal. We assess the performance of our proposed estimators both analytically and empirically via numerical simulations, and compare it with that of the other state-of-the-art methods in the literature.
Saeid Haghighatshoar, G. Caire
Journal name not available for this finding