Parametric Near-Field Channel Estimation for Extremely Large Aperture Arrays

The evolution of wireless communication has seen a shift towards massive MIMO and ultra-large antenna arrays to meet increasing data demands. Early works like Björnson et al. introduced the concept of ELAA, emphasizing the importance of near-field effects at high frequencies and large array scales. Traditional far-field models, assuming plane waves, become invalid as array dimensions surpass the Fresnel distance, leading to phase distortions and non-orthogonal channel representations. Recent advances include sparse polar domain representations and compressed sensing techniques, but these face limitations in practical scenarios due to correlation among grid points and computational costs. The need for precise user localization and channel estimation in near-field conditions has driven research towards parametric models and high-resolution spectral methods like MUSIC, which can exploit the structure of the signal subspace for accurate parameter extraction.

The core challenge lies in estimating multiple users' locations and channels in the near-field region of extremely large arrays. High-dimensional parameter spaces, coupled with limited snapshots and low SNR, cause rank deficiency and spectral peak conflation, impairing estimation accuracy. Existing methods either rely on exhaustive grid searches with prohibitive complexity or lack robustness in multi-user scenarios. Accurately modeling spherical wavefronts and efficiently estimating their parameters remains an open problem, crucial for enabling precise beamforming, localization, and interference management in future wireless networks.

This work introduces a two-step MUSIC-based framework tailored for near-field multi-user environments:

• �� Polar coordinate transformation captures spherical wavefront phase variations, reducing the parameter space.
• �� Sequential estimation of angles and distances via 2D and 1D spectral searches significantly lowers computational load.
• �� Spatial smoothing enhances robustness against snapshot scarcity, enabling multi-user separation.
• �� LS correction refines parameter estimates, compensating residual errors.

These innovations collectively address the computational and robustness limitations of prior approaches, offering a scalable solution for large-scale near-field channel estimation.

• �� Construct the sample covariance matrix from received signals, extract noise subspace via eigen-decomposition.
• �� Convert user positions into polar coordinates, approximating phase variations with Fresnel approximation.
• �� Perform 2D MUSIC spectrum search over azimuth and elevation angles to identify peaks.
• �� Fix estimated angles, conduct 1D MUSIC spectrum search for distance estimation.
• �� Convert estimated polar parameters back to Cartesian coordinates.
• �� Use these to build the parametric channel model.
• �� Apply spatial smoothing by dividing the array into overlapping subarrays, reconstructing the covariance matrix.
• �� Use LS-based correction to refine the estimated parameters, ensuring high accuracy.

Simulations involve a 100-antenna array at 3 GHz, with 4 users randomly placed within the near-field zone. The performance of the proposed method is compared against LS and R-LS estimators across SNR levels from 10dB to 30dB. Metrics include NMSE and beamforming gain, evaluated over 10,000 Monte Carlo runs. The experiments analyze robustness to snapshot limitations, user density, and array calibration errors, validating the efficiency and accuracy of the proposed approach in realistic scenarios.

Results show a 30% reduction in NMSE and 20% improvement in beamforming gain over classical methods, especially at low SNR. Spatial smoothing effectively mitigates rank deficiency, enabling precise multi-user localization. The two-step MUSIC reduces computational complexity by nearly 99%, making real-time processing feasible. The method maintains stable performance across various environmental conditions, demonstrating its robustness and scalability.

This approach can be directly applied to next-generation wireless systems, supporting high-density user environments, precise localization, and beamforming in 6G networks. It is suitable for scenarios like drone communication, autonomous vehicles, and smart factories, where accurate near-field channel knowledge is essential. Deployment requires array calibration and parameter initialization, but the method's robustness makes it practical for real-world use.

The method's performance diminishes under extremely low SNR (< 5dB) or in highly cluttered environments with multipath effects. Computational costs, while reduced, remain significant for ultra-large arrays. Sensitivity to array imperfections and environmental dynamics warrants further research. Future work should focus on integrating machine learning for adaptive estimation and extending the framework to dynamic, real-time tracking scenarios.