Source-linked AI summary
Channel Estimation for Millimeter-Wave Massive MIMO with Hybrid Precoding over Frequency-Selective Fading Channels
Zhen Gao, Linglong Dai, Chen Hu, Zhaocheng Wang
TL;DR
The paper addresses channel estimation for multi-user mmWave massive MIMO with hybrid precoding over frequency-selective fading, where few RF chains serve many antennas. It proposes DCS-based pilot and estimation procedures with DGMP and adaptive measurements, and reports accurate estimation with lower pilot overhead than an existing scheme.
Problem
Channel estimation is challenging because hybrid architectures have far fewer RF chains than antennas, while existing schemes mainly address narrow-band channels.
Method
The scheme exploits angle-domain structured sparsity through distributed compressive sensing, using designed pilots and DGMP with an adaptive measurement matrix.
Results
Simulation results show that the proposed scheme approaches ideal-AoA/AoD spectral-efficiency and BER performance while requiring reduced training overhead.
Takeaways & Limitations
The method provides multi-user uplink channel estimation for mmWave massive MIMO frequency-selective fading channels with lower pilot overhead than the existing scheme.
Abstract
from arXiv · showhide
Channel estimation for millimeter-wave (mmWave) massive MIMO with hybrid precoding is challenging, since the number of radio frequency (RF) chains is usually much smaller than that of antennas. To date, several channel estimation schemes have been proposed for mmWave massive MIMO over narrow-band channels, while practical mmWave channels exhibit the frequency-selective fading (FSF). To this end, this letter proposes a multi-user uplink channel estimation scheme for mmWave massive MIMO over FSF channels. Specifically, by exploiting the angle-domain structured sparsity of mmWave FSF channels, a distributed compressive sensing (DCS)-based channel estimation scheme is proposed. Moreover, by using the grid matching pursuit strategy with adaptive measurement matrix, the proposed algorithm can solve the power leakage problem caused by the continuous angles of arrival or departure (AoA/AoD). Simulation results verify that the good performance of the proposed solution.
I. INTRODUCTION
MmWave massive MIMO channel estimation is difficult with hybrid architectures because RF chains are far fewer than antennas, while prior schemes largely target narrow-band channels. The paper addresses multi-user uplink estimation over frequency-selective fading by exploiting structured sparsity and continuous-angle effects.
- Hybrid precoding reduces hardware cost and power consumption but makes channel estimation challenging with limited RF chains and hundreds of antennas.
- Existing schemes estimate AoA/AoD under discrete-angle assumptions, address only single-user channels, or introduce noise through repeated amplify-and-forward operations.
- The proposed scheme targets multi-user uplink channel estimation for broad-band frequency-selective fading channels.
- The approach exploits angle-domain structured sparsity that remains nearly unchanged across the system bandwidth.
- Distributed compressive sensing jointly designs transmit pilots and receive estimation, while grid matching pursuit with adaptive measurements addresses power leakage from continuous AoA/AoD.
II. SYSTEM MODEL
The system is a multi-user mmWave massive MIMO-OFDM link with hybrid precoding and frequency-selective multipath channels. Each user’s channel is modeled in the delay domain using path gains, delays, and steering vectors parameterized by AoA/AoD.
- The base station uses many antennas but fewer RF chains, while each user has one RF chain; the BS supports K user equipments.
- Hybrid analog-digital precoding enables spatial multiplexing of multiple data streams with low hardware cost and energy consumption.
- The uplink frequency-selective channel is represented in the delay domain as a sum of multipath components.
- Each path is characterized by a delay and complex gain, with azimuth AoA/AoD parameters for a uniform linear array.
- The path-gain model uses Rician fading with one LOS path and L_k−1 NLOS paths whose gains follow mutually independent complex distributions.
- The array steering vectors depend on wavelength and antenna spacing.
III. DCS-BASED CHANNEL ESTIMATION SCHEME
The paper proposes a distributed compressive sensing scheme to jointly estimate the frequency-selective fading channels.
- The proposed scheme uses distributed compressive sensing to jointly estimate the frequency-selective fading channels.
A. Uplink Pilot Training
OFDM pilot training converts the frequency-selective channel into subcarrier-specific observations while retaining common RF precoding and combining across subcarriers. The resulting angle-domain channel representation is sparse and structured across frequency, enabling joint pilot-based estimation.
- Pilot training: OFDM uses a cyclic prefix and DFT to combat frequency-selective fading and obtain received signals on pilot subcarriers.The cyclic-prefix length exceeds the maximum delay, and the DFT length exceeds the cyclic-prefix length.
- Pilot training: The RF precoding and combining matrices are shared across all subcarriers because the phase-shifter network provides nearly constant phase response over frequency.
- Channel sparsity: Because NLOS paths suffer much greater path loss than LOS paths, mmWave channels exhibit angular sparsity with few dominant paths.
- Channel sparsity: Quantizing the virtual angular domain with DFT matrices transforms the frequency-domain channel into a sparse angle-domain representation.
- Channel sparsity: The support set identifies dominant angular components, whose sparsity level can equal the number of paths when quantized AoA/AoD resolutions match the array dictionaries.
B. DCS-Based Channel Estimation
The proposed estimator uses distributed compressive sensing to jointly recover multi-user frequency-selective channels from structured sparse measurements. DGMP adds adaptive measurement matrices and grid matching to address leakage from continuous AoA/AoD.
- Motivation: Conventional MMSE estimation requires training overhead tied to the high dimension of the angle-domain channel, potentially exceeding channel coherence time.
- DCS formulation: Compressed sensing reduces pilot overhead by exploiting angular sparsity, while frequency-domain subchannels share structured sparsity across the system bandwidth.
- Power leakage: Continuous AoA/AoD and finite angular resolution cause power leakage that can impair the sparsity representation and channel estimation performance.
- DGMP algorithm: DGMP alternates an outer loop for selecting users and updating LOS path estimates with an inner loop that refines AoA/AoD using local overcomplete grids.
- DGMP algorithm: Joint processing of measurements and residuals across subcarriers exploits structured sparsity, while adaptive measurement matrices and grid matching provide higher-resolution AoA/AoD estimation.
C. Pilot Design According to DCS Theory
The pilot design targets reliable distributed sparse recovery by diversifying measurement matrices across subcarriers while respecting hybrid-precoding hardware constraints. The associated DGMP procedure estimates LOS steering vectors and path gains from received signals and sensing matrices.
- Measurement-matrix design: Reliable channel estimation depends on carefully designed measurement matrices, whose total observations must support robust sparse recovery.
- DGMP procedure: DGMP takes received signals, sensing matrices, an AoA/AoD resolution factor, and an error threshold as inputs.
- DGMP procedure: The algorithm outputs steering-vector estimates for each user’s LOS path and estimates of the corresponding path gains.
- Measurement-matrix design: The design uses random pilot phases so sensing-matrix elements become zero-mean i.i.d. complex Gaussian and differ across subcarriers.
- Evaluation: Fig. 2 compares spectral efficiency against training overhead G and SNR for adaptive CS, DGMP, and ideal-AoA/AoD performance.
- Hardware constraints: Pilot precoding and combining entries satisfy constant-modulus constraints, and RF matrices remain common across subcarriers.
IV. SIMULATION RESULTS
Simulations evaluate the proposed channel estimation scheme under specified mmWave massive MIMO-OFDM settings against ideal AoA/AoD knowledge and adaptive CS. The proposed DGMP-based approach achieves near-bound performance with lower training overhead and improved BER.
- Simulation setup: The simulations use a 30 GHz carrier, 0.25 GHz sampling frequency, 100 ns maximum delay spread, and 32 pilot symbols.The setup also specifies L_CP = 25 and compares against ideal AoA/AoD knowledge and adaptive CS.
- Spectral efficiency: The proposed DGMP algorithm approaches the ideal-AoA/AoD downlink spectral-efficiency bound when G ≥20.Adaptive CS requires G > 90 at SNR = 0 dB to approach the same bound.
- Training overhead: The proposed scheme substantially reduces the training overhead required for frequency-selective-fading channel estimation compared with adaptive CS.This advantage is attributed to exploiting angle-domain structured sparsity across the system bandwidth.
- BER performance: With 16-QAM, DGMP uses G = 30 versus G = 40 for adaptive CS and achieves BER close to the ideal-AoA/AoD performance bound.The proposed channel estimation scheme outperforms adaptive CS despite its reduced training overhead.
V. CONCLUSIONS
The paper proposes a DCS-based multi-user uplink estimator for mmWave massive MIMO over FSF channels. Its pilot design and DGMP algorithm exploit structured sparsity and adaptive measurements to reduce overhead and address power leakage.
- Contribution: The proposed DCS-based uplink scheme estimates multi-user mmWave massive MIMO channels under frequency-selective fading.The scheme is designed to effectively combat FSF channels.
- Method: An efficient pilot scheme and reliable DGMP algorithm are developed within the DCS framework to exploit angle-domain structured sparsity.This exploitation targets reduced training overhead.
- Algorithmic design: Grid matching pursuit with an adaptive measurement matrix addresses power leakage caused by continuous AoA/AoD values.The conclusion identifies this as a principal capability of the proposed algorithm.
- Conclusion: Simulations confirm accurate FSF channel estimation with much lower pilot overhead than the existing scheme.The conclusion reports this outcome without specifying a numerical overhead reduction.