Source-linked AI summary
Channel Estimation for FAS-assisted Multiuser mmWave Systems
Hao Xu, Gui Zhou, Kai-Kit Wong, Wee Kiat New, Chao Wang, Chan-Byoung Chae, Ross Murch, Shi Jin, Yangyang Zhang
TL;DR
FAS channel estimation requires CSI across many reconfigurable positions, exceeding the practicality of conventional fixed-antenna methods. The paper proposes L3SCR, which samples a few locations, extracts sparse angular and gain parameters, and reconstructs full CSI. Simulations report effective NMSE and sum-rate performance with minimal switching and pilot overhead.
Problem
FAS requires CSI over many possible antenna positions, while conventional fixed-antenna estimation is unsuitable and can incur high switching and pilot overhead.
Method
L3SCR measures channels at a few estimating locations, extracts sparse path angles and gains using mmWave sparsity, and reconstructs the complete channel matrix.
Results
The proposed scheme achieves effective NMSE and system sum-rate performance while requiring minimal hardware switching and pilot overhead.
Takeaways & Limitations
Sparse channel structure enables full-CSI reconstruction from reduced-dimensional measurements in FAS-assisted multiuser mmWave systems.
Abstract
from arXiv · showhide
This letter investigates the challenge of channel estimation in a multiuser millimeter-wave (mmWave) time-division duplexing (TDD) system. In this system, the base station (BS) employs a multi-antenna uniform linear array (ULA), while each mobile user is equipped with a fluid antenna system (FAS). Accurate channel state information (CSI) plays a crucial role in the precise placement of antennas in FAS. Traditional channel estimation methods designed for fixed-antenna systems are inadequate due to the high dimensionality of FAS. To address this issue, we propose a low-sample-size sparse channel reconstruction (L3SCR) method, capitalizing on the sparse propagation paths characteristic of mmWave channels. In this approach, each fluid antenna only needs to switch and measure the channel at a few specific locations. By observing this reduced-dimensional data, we can effectively extract angular and gain information related to the sparse channel, enabling us to reconstruct the full CSI. Simulation results demonstrate that our proposed method allows us to obtain precise CSI with minimal hardware switching and pilot overhead. As a result, the system sum-rate approaches the upper bound achievable with perfect CSI.
I. INTRODUCTION
FAS enables spatial diversity and interference reduction, but its continuously reconfigurable positions make full CSI acquisition difficult. The paper proposes sparse channel reconstruction for lower-overhead estimation in multiuser mmWave systems.
- FAS reconfigures antenna shape or position to access fading-channel variations and obtain additional communication gains.
- Full CSI across continuously variable FAS positions is difficult to acquire, making fixed-antenna channel-estimation schemes unsuitable.
- 256 antenna positions were required in an earlier planar-FAS estimator with 400 ports, creating high switching and pilot overhead.
- The proposed L3SCR method estimates reduced-dimensional channels at a few locations, extracts sparse path parameters, and reconstructs the complete channel matrix.
II. SYSTEM MODEL
The system is a narrow-band multiuser mmWave uplink with a fixed BS ULA and linear FAS users. Each user selects a port, and the target is the channel across all FAS ports.
- Each user has a linear FAS of size Wλ with N evenly distributed ports, while the BS uses a multi-antenna array.
- Users simultaneously transmit uplink data, and the BS detects each user with a linear receiver for its selected port.
- The sum-rate is optimized by choosing a favorable port nu from the N available ports.
- The channel-estimation objective is to recover Gu, the channel matrix spanning every port for each user.
III. LS ESTIMATION
The LS benchmark estimates each user-to-BS channel at every port using synchronized switching and orthogonal pilots. Its accuracy depends on pilot duration and transmit power, while its overhead grows with port count.
- All users synchronously switch across N ports and transmit mutually orthogonal pilot sequences for LS estimation.
- At each port, pilots are repeated over T subframes, each containing U time slots for multiuser channel estimation.
- The LS procedure estimates each port-specific channel and stacks the estimates into a matrix in CM×N.
- LS performance is determined by T and pu, but measuring all N ports requires extremely high switching and pilot overhead.
IV. L3SCR
L3SCR reduces measurements by exploiting sparse mmWave propagation: users sample only a few estimating locations, then the BS extracts path parameters and reconstructs full CSI.
- L3SCR has antennas switch and measure at only a few estimating locations, using sparse channel structure to reconstruct full CSI.
- The method can be generalized to a BS equipped with FAS, which would also switch among ports to collect pilot signals.
- Each user samples K ≪ N uniformly distributed estimating locations with spacing Δ satisfying Δ≤Wλ/(K−1).
- The geometric mmWave model represents each channel using propagation paths, complex gains, and receiver and transmitter steering vectors.
- Orthogonal pilots first produce an LS estimate of the reduced channel matrix Hu, after which sparse parameters are extracted and Gu is reconstructed.
A. Estimation of Number of Paths and AoAs
The method estimates the number of paths and AoAs from a coarse DFT-domain representation, then refines angular estimates through angular rotation to compensate for resolution limits and mismatch.
- A. Estimation of Number of Paths and AoAs: DFT-based processing produces a coarse estimate of the number of paths and AoAs from concentrated row power.The method searches for prominent row-power peaks, treating their count as the estimated number of paths.
- A. Estimation of Number of Paths and AoAs: The DFT representation is asymptotically row sparse, with power concentrated in a few rows and leakage into nearby rows.
- A. Estimation of Number of Paths and AoAs: Angular rotation refines the coarse AoA estimates by compensating for mismatch caused by finite DFT resolution.The rotation shifts the power beam continuously around its DFT index, enabling a search for the best compensation parameter.
B. Estimation of AoDs and Channel Gains
After estimating path number and AoAs, the method projects the channel onto an AoA subspace and estimates each path’s AoD and gain as a 1-sparse reconstruction problem.
- B. Estimation of AoDs and Channel Gains: Accurate path-number and AoA estimates allow projection onto an AoA steering-matrix subspace before estimating remaining path parameters.
- B. Estimation of AoDs and Channel Gains: Each projected column contains the AoD and channel-gain information of one propagation path, yielding a 1-sparse reconstruction problem.
- B. Estimation of AoDs and Channel Gains: Low-complexity matched filters with a dictionary matrix are applied to estimate the AoDs and channel gains.The dictionary size is denoted by C, and increasing C provides a finer angular search.
- B. Estimation of AoDs and Channel Gains: L3SCR requires at least two distinct estimation locations because K = 1 prevents AoD estimation.
- B. Estimation of AoDs and Channel Gains: Errors in the estimated number of paths or AoAs can compromise the performance of the proposed scheme.
- B. Estimation of AoDs and Channel Gains: The switching step is restricted to less than the wavelength during estimation to avoid angular mismatch.
C. Channel Reconstruction
The estimated path count, AoAs, AoDs, and channel gains are inserted into the planar-wave geometric channel model to reconstruct the full channel matrix.
- C. Channel Reconstruction: Once the sparse path parameters are estimated, the full channel matrix Gu is reconstructed using the planar-wave geometric model.
- C. Channel Reconstruction: The reconstructed channel uses estimated AoDs to form the BS-side steering terms across the antenna array.
D. Analysis of Pilot Overhead and Computational Complexity
L3SCR reduces antenna switching and pilot overhead by sampling K locations rather than all N ports, while its computational cost is characterized separately for angle and gain estimation.
- D. Analysis of Pilot Overhead and Computational Complexity: With K = 6 and T = 1, L3SCR obtains accurate full CSI using low hardware switching and pilot overhead.
- D. Analysis of Pilot Overhead and Computational Complexity: L3SCR uses KTU pilot symbols, whereas LS requires NTU pilot symbols and N hardware switches.
- D. Analysis of Pilot Overhead and Computational Complexity: The complexity of estimating path number and AoAs is dominated by angular rotation, while AoD and gain estimation depends on dictionary multiplications.
- D. Analysis of Pilot Overhead and Computational Complexity: The overall L3SCR complexity is expressed as a combination of the angle-estimation and AoD/gain-estimation costs.
V. SIMULATION RESULTS
Simulations compare L3SCR with LS and OMP using NMSE, sum rate, and computational time. L3SCR reduces overhead while its estimation and rate performance depend on antenna-array size, sampled locations, SNR, and pilot repetitions.
- Evaluation setup: The simulations compare L3SCR, LS, and OMP using NMSE and average system rate over 1000 channel realizations.The BS uses maximum ratio combining, and exhaustive searching selects user ports to maximize the average rate.
- Effect of M and K: As M increases, L3SCR and OMP achieve lower NMSE and higher sum rate.OMP performs better when M is small, whereas L3SCR performs better when M is large.
- Rate and NMSE trade-off: LS often achieves lower NMSE, but its NTU pilot overhead produces lower sum rate than L3SCR and OMP.The comparison reflects the trade-off between estimation accuracy and pilot overhead.
- Effect of SNR and T: L3SCR and OMP NMSE decrease initially with SNR and then saturate, while LS NMSE decreases linearly with ρ.L3SCR and OMP performance also depends on M and K.
- Computational time: OMP requires much higher computational complexity than L3SCR, and both methods take less computational time as ρ increases.At low ρ, more iterations are needed for convergence; OMP additionally computes matrix inversions.
VI. CONCLUSIONS
The letter proposes sparse channel estimation for multiuser mmWave communication with FAS-equipped mobile users. Each fluid antenna samples a limited set of ELs to reconstruct full CSI with low switching and pilot overhead, achieving effective NMSE and sum-rate performance.
- Conclusion: The scheme uses mmWave channel sparsity to estimate full CSI from measurements at a limited set of ELs.Each fluid antenna selectively switches and measures the channel across those locations.
- Conclusion: The proposed reconstruction requires minimal hardware switching and pilot overhead while supporting NMSE and system sum-rate performance.The conclusion identifies both estimation error and sum rate as evaluation criteria.
APPENDIX A PROOF OF LEMMA 1
The appendix proves Lemma 1 by treating two angular cases separately. For sufficiently large M, it identifies an integer antenna index that yields the required geometric-sum relationships and completes the result for both cases.
- Case analysis: The proof separates the analysis into ϕu,l ∈ [0, π/2] and ϕu,l ∈ (π/2, π] cases.The two angular ranges are handled separately before applying the corresponding algebraic identities.
- Geometric-sum step: For indices m ≠ ml, expressing m as ml + f and applying the geometric-progression sum yields the stated zero relation.The argument explicitly uses a nonzero natural-number offset f.
- Second case: The second angular case uses e−j2π(m′−1) = 1 to rewrite the relevant expression and obtain the analogous result.The appendix then states that the proof proceeds similarly and completes Lemma 1.
- Case analysis: When M is sufficiently large, an integer ml ∈ {1, ..., M} exists for the relevant angular case.The proof uses this index to establish the required relations involving the array response.