A Hybrid Dynamic Subarray Architecture for Efficient DOA Estimation in THz Ultra-Massive Hybrid MIMO Systems

Ye Tian, Jiaying Ren, Te‐Kao Wu, Wei Liu, Chau Yuen, Mérouane Debbah, Naofal Al‐Dhahir, Matthew C. Valenti, Hing Cheung So, Yonina C. EldararXiv 2025 · DOI · arXiv

TL;DR

This paper introduces a Hybrid Dynamic Subarray (HDS) architecture and two specialized algorithms designed to make high-precision, 2D Direction-of-Arrival (DOA) estimation practical for 6G Terahertz (THz) systems. By combining a flexible hardware design with bias-correction math and dimensionality reduction, the authors achieve the accuracy of "gold-standard" methods at a fraction of the computational cost.

What Problem Does This Paper Address?

As we move toward 6G, Terahertz (THz) frequencies offer massive bandwidth but suffer from extreme signal loss. To compensate, researchers use Ultra-Massive MIMO (UM-MIMO)—arrays with thousands of tiny antennas—to create highly directional beams.

However, two major hurdles remain:

  1. Hardware Complexity: A "fully digital" system (one RF chain per antenna) is too expensive and power-hungry for thousands of elements. "Hybrid" architectures reduce RF chains but make it harder to capture the spatial information needed to find where a signal is coming from (DOA estimation).
  2. The "Massive" Math Problem: In ultra-massive arrays, standard statistical tools (like eigenvalue decomposition) become biased. Furthermore, searching for a signal in 2D space (azimuth and elevation) is computationally "prohibitive" for real-time 6G applications.

Methods

The authors propose a three-pronged solution:

  • HDS Architecture: They designed a flexible "Hybrid Dynamic Subarray" structure. Unlike rigid designs, this uses a switch network that allows RF chains to connect to different subarrays dynamically. This flexibility is key to simplifying 2D problems into 1D ones.
  • RD-MUSIC (Reduced-Dimension MUSIC): This algorithm transforms a heavy 2D spectral search into two much faster 1D searches. Crucially, it incorporates General Asymptotic Theory (GAT) to correct the mathematical bias that occurs when the number of antennas is very large relative to the number of signal samples.
  • IMRD-MUSIC (Improved RD-MUSIC): This version exploits the inherent sparsity of THz channels (where signals usually travel along just a few paths). By using a specific two-RF-chain configuration, it generates a "closed-form" (direct mathematical) initial guess for the elevation angle, nearly eliminating the need for a wide-range search.

Datasets

The paper relies on extensive Monte-Carlo simulations rather than a static external dataset. This is standard for physical-layer communications research.

  • Channel Model: They used a THz-specific model consisting of one Line-of-Sight (LOS) path and two Non-Line-of-Sight (NLOS) paths.
  • Setup: Simulations featured a $32 \times 32$ Uniform Planar Array (1,024 antennas) operating at 1.0 THz.
  • Benchmarks: The proposed methods were compared against traditional 2D-MUSIC and the Cramer-Rao Lower Bound (CRLB), which represents the theoretical limit of estimation accuracy.

Key Insights and Ideas

  • Bias Correction is Mandatory: The authors highlight that in UM-MIMO, the "sample" covariance matrix deviates significantly from the "true" one. Without the GAT-based correction used in RD-MUSIC, estimation accuracy hits a floor regardless of how much power you use.
  • Sparsity as an Advantage: While THz propagation is difficult, its sparsity is a gift for signal processing. IMRD-MUSIC uses this to skip the "brute force" search entirely for the initial estimate.
  • Strategic Hardware Trade-offs: The HDS architecture proves that you don't need a "fully connected" hybrid system (where every RF chain hits every antenna) to get top-tier performance. A well-designed subarray system can match the performance of much more complex hardware.

Key Results

  • Accuracy: Both RD-MUSIC and IMRD-MUSIC achieve Root Mean Square Error (RMSE) levels that closely track the CRLB, matching the performance of the computationally expensive 2D-MUSIC.
  • Efficiency: IMRD-MUSIC significantly reduces the "Spectral Search" complexity. While 2D-MUSIC scales with the product of azimuth and elevation grid points ($n_\theta \times n_\phi$), IMRD-MUSIC scales only with a small refined search area and the number of signal paths ($L$).
  • Robustness: The algorithms remain effective even with a limited number of pilot signals ($T=8$ or $12$), which is vital for reducing 6G signaling overhead.

Limitations and Future Work

  • Narrowband Assumption: The current work assumes a narrowband signal. In ultra-wideband THz systems, "beam squint" (where the beam direction changes with frequency) becomes a major issue that this paper does not yet address.
  • Hardware Imperfections: The study assumes perfect phase shifters and switches. In reality, THz hardware suffers from phase noise and insertion losses that could degrade performance.
  • Future Directions: The authors suggest extending this framework to wideband systems and exploring how these DOA estimates can be used for integrated sensing and communication (ISAC), another pillar of 6G.

Summary updated

Abstract

Terahertz (THz) communication combined with ultra-massive multiple-input multiple-output (UM-MIMO) technology is promising for 6G wireless systems, where fast and precise direction-of-arrival (DOA) estimation is crucial for effective beamforming. However, finding DOAs in THz UM-MIMO systems faces significant challenges: while reducing hardware complexity, the hybrid analog-digital (HAD) architecture introduces inherent difficulties in spatial information acquisition the large-scale antenna array causes significant deviations in eigenvalue decomposition results; and conventional two-dimensional DOA estimation methods incur prohibitively high computational overhead, hindering fast and accurate realization. To address these challenges, we propose a hybrid dynamic subarray (HDS) architecture that strategically divides antenna elements into subarrays, ensuring phase differences between subarrays correlate exclusively with single-dimensional DOAs. Leveraging this architectural innovation, we develop two efficient algorithms for DOA estimation: a reduced-dimension MUSIC (RD-MUSIC) algorithm that enables fast processing by correcting large-scale array estimation bias, and an improved version that further accelerates estimation by exploiting THz channel sparsity to obtain initial closed-form solutions through specialized two-RF-chain configuration. Furthermore, we develop a theoretical framework through Cramér-Rao lower bound analysis, providing fundamental insights for different HDS configurations. Extensive simulations demonstrate that our solution achieves both superior estimation accuracy and computational efficiency, making it particularly suitable for practical THz UM-MIMO systems.
BibTeX
@misc{tian2025hybrid,
  author = {Ye Tian and Jiaying Ren and Te‐Kao Wu and Wei Liu and Chau Yuen and Mérouane Debbah and Naofal Al‐Dhahir and Matthew C. Valenti and Hing Cheung So and Yonina C. Eldar},
  title = {{A Hybrid Dynamic Subarray Architecture for Efficient DOA Estimation in THz Ultra-Massive Hybrid MIMO Systems}},
  year = {2025},
  eprint = {2501.18378},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2501.18378},
  howpublished = {arXiv},
}
Figure