Wideband Massive MIMO Channel Estimation via Sequential Atomic Norm Minimization

Abstract: The recently introduced atomic norm minimization (ANM) framework for parameter estimation is a promising candidate towards low overhead channel estimation in wireless communications. However, previous works on ANM-based channel estimation evaluated performance on channels with artificially imposed channel path separability, which cannot be guaranteed in practice. In addition, direct application of the ANM framework for massive MIMO channel estimation is computationally infeasible due to the large dimensions. In this paper, a low-complexity ANM-based channel estimator for wideband massive MIMO is proposed, consisting of two sequential steps, the first estimating the channel over the spatial and the second over the frequency dimension. Its mean squared error performance is analytically characterized in terms of tight lower bounds. It is shown that the proposed algorithm achieves excellent performance that is close to the best that can be achieved by any unbiased channel estimator in the regime of low to moderate number of channel paths, without any restrictions on their separability.