Downlink Massive MIMO Channel Estimation via Deep Unrolling: Sparsity Exploitations in Angular Domain

In frequency division duplex (FDD) massive multiple input multiple output (MIMO) systems, reliable downlink channel estimation is essential but requires huge pilot overhead due to hundreds of antennas at base station (BS). In order to reduce pilot overhead without compromising the channel estimation, compressive sensing (CS) has been widely applied for channel estimation by exploiting the inherent sparse structure of massive MIMO channel in angular domain. However, it still suffers from high complexity during the optimization process and the requirement of prior information on the number of spatial paths (PINP). To overcome these challenges, this paper develops a novel hybrid channel estimation scheme by integrating model-driven CS and data-driven deep unrolling techniques. The proposed scheme is composed of a coarse estimation part and a fine correction part, which is implemented in a two-stage manner by exploiting both inter- and intra-frame sparsities of channels in angular domain. Additionally, a threshold function is proposed to eliminate the requirement of the number of spatial paths. Theoretical results are provided to indicate the convergence of both fine correction and coarse estimation. Numerical results demonstrate that our scheme can achieve high accuracy with less pilot overhead and low complexity.