Pilot decontamination through compressive wideband channel estimation

We consider a multi-cell massive MIMO system in which every base station is equipped with an array with M ≫ 1 antennas and serves a collection of users inside its cell. We study a scenario in which the base station and the users use OFDM signaling, and the channel state information (CSI) is estimated under time division duplexing (TDD) via orthogonal pilots transmitted from the users in the uplink and is used under reciprocity assumption to precode the data to the users in the downlink. It has been observed that in the limit of very large number of antennas M at each base station, the only factor limiting the achievable throughput of the system is pilot contamination. Pilot contamination occurs because the same set of orthogonal pilots are reused in adjacent cells and as a result the estimate of the channel state obtained at the base station in a specific cell becomes polluted by the simultaneous pilot transmission of the users from the other cells. In this paper, we propose a low-complexity algorithm that uses the received wideband pilot snapshots of a user inside an observation window containing several coherence blocks to obtain an estimate of the angle-delay power spread function (PSF) of its contaminated channel vectors. Since the copilot users can generally be distinguished in the delay domain due to their longer delays, the estimated PSF can be decomposed into two parts: 1) the signal part corresponding to the PSF of the channel vectors of the user inside the cell and 2) the interference part corresponding to the that of copilot users. We use this decomposition to obtain an estimate of the covariance matrix of the user channel vector, which we exploit to decontaminate the received pilot signal in the next coherence blocks. We use numerical simulations to assess the performance of our proposed PSF estimation and pilot-decontamination algorithm.