Source-linked AI summary
Channel Estimation for Movable-Antenna MIMO Systems Via Tensor Decomposition
Ruoyu Zhang, Lei Cheng, Wei Zhang, Xinrong Guan, Yueming Cai, Wen Wu, Rui Zhang
TL;DR
The paper addresses high-overhead channel estimation for MIMO systems with movable antennas at both transmitter and receiver. It introduces a two-stage movement and CP-tensor method to estimate multipath parameters and reconstruct arbitrary-position channels, achieving lower pilot overhead with comparable or better estimation accuracy than benchmark methods.
Problem
Fixed-grid measurements can require prohibitive pilot overhead, while channel estimation must recover channels across arbitrary Tx/Rx movable-antenna positions.
Method
A two-stage Tx-Rx successive movement pattern forms third-order tensors whose CP factor matrices yield multipath angles and complex gains for channel reconstruction.
Results
The proposed method achieves lower NMSE than two CS-based algorithms and comparable NMSE to benchmarks with 80% pilot-overhead reduction.
Takeaways & Limitations
Finite-position measurements can provide the information required for complete channel reconstruction over the Tx and Rx movement regions.
Abstract
from arXiv · showhide
In this letter, we investigate the channel estimation problem for MIMO wireless communication systems with movable antennas (MAs) at both the transmitter (Tx) and receiver (Rx). To achieve high channel estimation accuracy with low pilot training overhead, we propose a tensor decomposition-based method for estimating the parameters of multi-path channel components, including their azimuth and elevation angles, as well as complex gain coefficients, thereby reconstructing the wireless channel between any pair of Tx and Rx MA positions in the Tx and Rx regions. First, we introduce a two-stage Tx-Rx successive antenna movement pattern for pilot training, such that the received pilot signals in both stages can be expressed as a third-order tensor. Then, we obtain the factor matrices of the tensor via the canonical polyadic decomposition, and thereby estimate the angle/gain parameters for enabling the channel reconstruction between arbitrary Tx/Rx MA positions. In addition, we analyze the uniqueness condition of the tensor decomposition, which ensures the complete channel reconstruction between the whole Tx and Rx regions based on the channel measurements at only a finite number of Tx/Rx MA positions. Finally, simulation results are presented to evaluate the proposed tensor decomposition-based method as compared to existing methods, in terms of channel estimation accuracy and pilot overhead.
I. INTRODUCTION
Conventional fixed-geometry MIMO cannot fully exploit spatial channel variation, motivating movable antennas and new estimation methods. The proposed tensor-based approach estimates multipath parameters and reconstructs channels across arbitrary Tx/Rx positions with reduced pilot overhead.
- Fixed antenna geometry limits MIMO’s ability to exploit wireless-channel spatial variation and spatial degrees of freedom.
- Movable antennas enable flexible local movement at the transmitter and receiver to seek better channel conditions.
- Existing compressed-sensing estimation methods estimate angular and gain parameters but rely on discrete angle quantization.
- The proposed method uses two-stage successive Tx-Rx movement, third-order tensors, and CP decomposition to estimate multipath parameters and reconstruct arbitrary-position channels.
- Tensor-decomposition uniqueness analysis supports complete channel reconstruction from measurements at only finitely many Tx/Rx positions.
II. SYSTEM AND CHANNEL MODEL
The system models movable Tx and Rx antennas over two-dimensional regions and seeks channel estimates for arbitrary position pairs. Directly measuring every grid pair is costly, while shared finite multipath parameters motivate tensor-based estimation.
- Tx and Rx movable antennas adjust positions within rectangular local regions, with channel matrices depending on their coordinate collections.
- The field-response channel model represents the MIMO channel through Tx and Rx field-response matrices and a path response matrix.
- Multipath components are parameterized by propagation-distance differences and azimuth and elevation departure or arrival angles.
- The target is channel estimation at any Tx/Rx position pair, while pilot measurements restrict antennas to centers of discretized regional grids.
- Measuring all Tx/Rx grid pairs requires prohibitive pilot overhead as grid counts grow, whereas shared finite path parameters motivate tensor representations.
III. TENSOR DECOMPOSITION-BASED CHANNEL ESTIMATION AND RECONSTRUCTION
The proposed estimation framework uses successive antenna movement and tensor decomposition to recover multipath angles and gains, enabling channel reconstruction throughout the Tx and Rx movement regions.
- A two-stage Tx-Rx successive movement pattern makes received pilot signals representable as a third-order tensor.
- Tensor factor matrices provide the basis for estimating multipath AoDs, AoAs, and complex gains.
- The estimated parameters reconstruct channels between arbitrary transmitter and receiver movable-antenna positions.
- The method also analyzes the uniqueness condition of the tensor decomposition.
A. Tensor Representation
Pilot training moves antenna groups successively over selected grid positions, producing received-signal matrices that can be reshaped into third-order tensors. Their factor matrices retain channel information and support gridless angle estimation and reconstruction.
- Rx movement stage: During the first stage, Rx antennas move over J/N positions while Tx antennas remain fixed at an initial position.
- Tx movement stage: During the second stage, Tx antennas move over I positions, with one active Tx antenna transmitting a unit-power pilot at each position.
- Tensor construction: Pilot processing forms matrices whose columns collect received signals across moved antenna positions and whose factors encode Tx-side steering responses.
- Tensor construction: The tensor factor matrices incorporate the channel information needed for subsequent channel estimation and reconstruction.
1) Estimation of Factor Matrices:
The method estimates tensor factor matrices from received-signal tensors using CP decomposition, with ALS providing an iterative solution procedure.
- CP decomposition is applied to the received signal tensor to estimate its factor matrices.
- ALS alternately updates one factor matrix while fixing the others until convergence.
- The procedure estimates {Ây, Âx, Ḋt} for one tensor and {B̂y, B̂x, Ḋr} for the other after convergence.
2) Estimation of AoDs and AoAs:
The estimated factor matrices are used to recover path angles despite CP ambiguities, while their Vandermonde structure enables gridless AoD estimation and the algorithm outputs reconstructed channel parameters.
- CP-derived factor matrices contain scaling and permutation ambiguities, but these preserve pairing between corresponding estimated and true factors.
- The successive movement pattern gives Ay and Ax an inherent Vandermonde structure that facilitates gridless AoD estimation.
- The algorithm estimates factor matrices, angles, complex gains, and a reconstructed channel matrix from received signals and antenna movement regions.
- The same estimation procedure is applied to obtain AoAs from the receiver-side factor matrices.
3) Channel Reconstruction:
After estimating path angles, the method estimates the path response matrix and uses the resulting field-response matrices to reconstruct channels at arbitrary transmitter and receiver MA positions.
- The estimated AoDs and AoAs are used to construct estimated field-response matrices for channel reconstruction.
- The received signals from both movement stages are combined in the reconstruction procedure through vectorization and a joint observation model.
- The channel is reconstructed for any selected set of transmitter and receiver MA positions using the estimated channel parameters.
C. Uniqueness Condition of Tensor Decomposition
The paper analyzes CP-decomposition uniqueness to ensure that factor matrices recover the unknown multipath parameters, finding that finite antenna movements can suffice for complete channel reconstruction.
- CP uniqueness is essential because the decomposed factor matrices must recover all unknown multipath channel parameters.
- A sufficient uniqueness condition for a third-order tensor is min(I1, L) + min(I2, L) + min(I3, L) ≥ 2L + 2.
- For the transmitter tensor, uniqueness requires min(Ix, Lt) + min(Iy, Lt) + min(N, Lt) ≥ 2Lt + 2.
- The required transmitter and receiver movement counts depend on the number of channel paths, not the total grid positions, reducing pilot training overhead.
IV. SIMULATION RESULTS
Simulations evaluate channel reconstruction across SNR and pilot-training-region coverage. The proposed tensor method outperforms the benchmark CS methods and achieves comparable NMSE with substantially less pilot overhead.
- Simulation setup: The simulation uses M = N = 4 antennas, 8λ × 8λ Tx/Rx regions, and Lt = Lr = 3 channel paths.Adjacent grid centers are separated by Δ = λ/5, and βt and βr denote the visited fractions of the Tx and Rx regions.
- Simulation setup: The evaluation compares the proposed method with the Cramér-Rao bound and successive transmitter-receiver CS benchmarks using OMP and SOMP.NMSE is the channel-reconstruction metric, and SNR is defined as P/σ2.
- NMSE versus SNR: Across SNR values, the proposed method achieves lower NMSE than the two CS-based algorithms for βt = βr = 25% or 56.25%.Its NMSE continues decreasing with SNR, whereas the benchmark methods saturate; the proposed method follows a trend similar to the CRB.
- NMSE versus pilot coverage: Increasing pilot-training coverage improves the proposed method’s channel-estimation performance and maintains an advantage over the benchmark algorithms.The comparison varies βt at SNR = 15 dB with βr set to 25% or 56.25%.
- NMSE versus pilot coverage: 80% reduction in pilot overhead is achieved when the proposed method uses βt = 20% while matching the benchmarks’ NMSE at βt = 100%.This comparison is reported for the pilot-training-region coverage experiment.
V. CONCLUSION
The paper proposes tensor decomposition-based channel estimation and reconstruction for MA-enabled MIMO systems. Simulations and tensor-uniqueness analysis show significant pilot-overhead reduction while maintaining benchmark channel-estimation accuracy.
- Conclusion: The method uses a two-stage Tx-Rx successive antenna movement pattern and estimates multi-path parameters to reconstruct channels between arbitrary Tx/Rx MA positions.The estimated parameters include path angles and complex gains.
- Conclusion: Tensor-decomposition uniqueness analysis and simulations demonstrate significant pilot-overhead reduction with the same channel-estimation accuracy as benchmark schemes.