Channel Estimation and Hybrid Precoding for Frequency Selective Multiuser mmWave MIMO Systems

TL;DR

Addresses the challenge of configuring hybrid analog–digital precoding for frequency-selective MU-mmWave MIMO. It proposes uplink compressive channel estimation to configure downlink beams, exploiting reciprocity and sparsity across $K$ subcarriers, and develops two hybrid design strategies: projected gradient-based factorization and alternating minimization, complemented by a CRLB analysis. The method uses SW-OMP for uplink channel recovery and analyzes performance under both perfect and imperfect CSI, with extensive simulations. The results demonstrate that UL CSI acquisition can significantly outperform DL-based estimation under fixed overhead, while revealing tradeoffs between overhead, complexity, and robustness to grid/off-grid effects and model mismatch.

Abstract

Configuring the hybrid precoders and combiners in a millimeter wave (mmWave) multiuser (MU) multiple-input multiple-output (MIMO) system is challenging in frequency selective channels. In this paper, we develop a system that uses compressive estimation on the uplink to configure precoders and combiners for the downlink (DL). In the first step, the base station (BS) simultaneously estimates the channels from all the mobile stations (MSs) on each subcarrier. To reduce the number of measurements required, compressed sensing techniques are developed that exploit common support on the different subcarriers. In the second step, exploiting reciprocity and the channel estimates, the base station designs hybrid precoders and combiners. Two algorithms are developed for this purpose, with different performance and complexity tradeoffs: 1) a factorization of the purely digital solution, and 2) an iterative hybrid design. Extensive numerical experiments evaluate the proposed solutions comparing to state-of-the-art strategies, and illustrating design tradeoffs in overhead, complexity, and performance.

Figures (4)

• Figure 1: Estimation performance (NMSE) for both UL and DL multiuser channel estimation considering both on-grid and off-grid channel realizations: (a) NMSE versus SNR. (b) NMSE versus number of training frames $M$ for SNR ranging $\left\lbrace-10,-5,0\right\rbrace$ dB.
• Figure 2: (a) Achievable Sum-Rate vs. SNR for error-free channel estimation. Digital MMSE, Digital MRC, and Digital CB represent the digital benchmarks, HD-LSR and HD-PG are the factorizations using RuMeGoHe16 and Alg. \ref{['alg:ProjGrad']}. Finally, AM MMSE, AM CB, and AM MRC are the methods from Sec. \ref{['sec:AM']}. (b) Achievable Sum-Rate vs. N# of RF Chains with ${L_\text{BS}}=2,4,8,12,16$. We consider ${N_\text{BS}}=64$, ${L_{\text{MS},u}}=4$, and a fixed SNR of $0$dB.
• Figure 3: (a) Achievable Sum-Rate vs. SNR for perfect and imperfect CSI using digital solutions with off-grid channel model. We include precoders and combiners jointly obtained with the algorithm PaWoNg04 as benchmark. (b) Achievable Sum-Rate vs. SNR with hybrid precoders and combiners. EY-UB CB represents the Eckart-Younglow rank approximation of CB. HD-PG and HD-LSR are the hybrid designs using Alg. \ref{['alg:ProjGrad']} and RuMeGoHe16. MMSE-EY, CB-EY, MRC-EY, Nu-SVD-EY correspond to Eckart-Young low rank approximations for the methods MMSE, CB, maximum ratio combiner MRC, and Nu-SVD. AM are the alternating minimization methods.
• Figure 4: Achievable Sum-Rate vs. SNR with hybrid precoders and combiners. Compare with Fig. \ref{['fig:HybridICSI']}.