Beamspace Channel Estimation via Atomic Norm Minimization

Beamspace Massive multiple-input multiple-output (Massive MIMO) based on lens antenna arrays (LAAs) has shown significant potential to provide high data rate as it can achieve sufficient beamforming gain with a limited number of radio frequency (RF) chains. However, channel estimation is extremely challenging due to the transceiver architecture with few RF chains and a large number of antennas. Most existing beamspace channel estimation schemes are grid-based sparse signal reconstruction. Although compressed sensing (CS) can guarantee the reconstruction performance for signals located on the grid, it cannot perfectly map all true signals onto the grid due to its continuous nature, which leads to the basis mismatch problem, and thus reduces the estimation performance. To this end, in this paper we investigate a continuous dictionary for beamspace channel estimation. Specifically, the beamspace channel estimation is formulated as an atomic norm minimization (ANM) problem. On this basis, a continuous dictionary is employed for channel angle estimation to avoid the effects of base mismatch caused by grid-based CS methods. The effectiveness of the proposed ANM estimation method is verified by the simulation results, as the proposed scheme outperforms the existing CS-based grid estimation methods.