Acquiring accurate channel state information (CSI) is essential for millimeter-wave (mmWave) multi-input multi-output (MIMO) systems. However, it remains challenging due to high-dimensional channels, limited pilot observations, and measurement noise. To address these challenges, we propose flow-guided angular channel estimation (FACE), a novel channel estimation framework based on flow matching, an emerging generative modeling paradigm that learns continuous transport dynamics between probability distributions. FACE exploits the inherent angular-domain sparsity of mmWave channels and incorporates the learned flow matching model as a generative channel prior. Numerical results show that FACE consistently achieves higher channel estimation accuracy than both conventional estimators and diffusion-based generative estimator across diverse channel and system configurations. Moreover, FACE maintains high estimation accuracy with fewer pilot observations and achieves an average achievable rate close to the perfect-CSI benchmark. Compared with diffusion-based estimation, FACE requires substantially fewer sampling steps while achieving superior or comparable estimation accuracy, reducing inference time and computational cost. Overall, the proposed framework demonstrates the potential of flow matching for accurate, pilot-efficient, and computationally efficient mmWave MIMO channel estimation.
Figures & tables
Fig. 1 : Normalized instantaneous channel power gain of the spatial- and angular-domain channels for L=20 : (a) spatial-domain channel H and (b) angular-domain channel Ha .
Fig. 2 : Proposed FACE framework for mmWave channel estimation.
mmWave channel
FACE framework
Parameter
Value
Parameter
Value
fc
60 GHz
Architecture
U-Net
Nt
{32,64,128}
Optimizer
Adam
Nr
{16,32}
Residual blocks
2
NtRF
16
Batch size
32
L
{3,10,50}
Learning rate
1×10−4
TABLE I : Simulation parameters
Fig. 3 : Variation of NMSE with SNR for different ULA geometry.
Fig. 4 : Variation of NMSE with SNR for different UPA geometry.
Fig. 5 : Variation of NMSE with SNR for different multi-paths for Nt=64 , Nr=16 ULA geometry.
Fig. 6 : Variation of NMSE with SNR for different multi-paths for Nt=64 , Nr=16 UPA geometry.
Fig. 7 : Comparison of the empirical cumulative distribution of the NMSE and the average achievable rates of the estimated channels: (a) CDF of ULA and UPA geometry, and (b) achievable rate of ULA and UPA geometry.
Fig. 8 : Channel estimation performance of the considered estimation schemes using real channel measurements: (a) Comparison of NMSE with SNR and (b) average achievable rate.
Fig. 9 : Channel estimation performance of FACE and the diffusion-based estimator: (a) Spatial and angular domains, and (b) Pilot length impact.
Fig. 10 : The channel estimation evolution and difference map for multi-paths L=20 : (a) FACE evolution over time, (b) diffusion evolution over steps.
Fig. 11 : Impact of the number of sampling steps on FACE and the diffusion-based estimator: (a) Channel estimation performance and (b) inference time.
Approximate NMSE
FACE complexity
Diffusion complexity
−20 dB
10.25
205.00
−25 dB
20.50
205.00
TABLE II : Computational cost comparison of FACE and the diffusion-based estimator in terms of GFLOPs for achieving a target NMSE
College of Electronic Information and Optical Engineering, Nankai University · School of Computing Science, University of Glasgow · School of Natural and Computing Science, University of Aberdeen
School of Electrical and Computer Engineering, University of Tehran, Tehran, Iran · Department of Computer Engineering, Sharif University of Technology, Tehran, Iran