Efficient Multi-Beam Training For Terahertz Wireless communications

Songjie Yang, Zhongpei Zhang, Zhenzhen Hu, Nuan Song, Hao Liu

TL;DR

Researchers have developed a two-stage "auxiliary-assisted" beam training scheme for Terahertz (THz) communications that slashes the time needed to establish a wireless link. By using a small, low-cost antenna array to "scout" the best signal directions before fine-tuning with a large-scale data array, the system achieves near-optimal spectral efficiency with a computational complexity that scales linearly with the number of signal paths.

What Problem Does This Paper Address?

As we look toward 6G, Terahertz (THz) frequencies (0.1–10 THz) offer massive bandwidth for ultra-high data rates. However, THz waves suffer from extreme path loss and atmospheric molecular absorption. To overcome this, transmitters and receivers must use massive antenna arrays to create highly directional, pencil-thin beams.

The "catch-22" is beam training. To find each other, the transmitter and receiver usually perform an exhaustive search, sweeping through every possible beam combination. With the massive arrays required for THz, this process becomes incredibly slow and resource-heavy, eating up the very bandwidth the technology is supposed to provide. This paper aims to find a way to establish these links quickly and efficiently without sacrificing signal quality.

Methods

The authors propose a hybrid beamforming architecture that uses two sets of arrays: a small-scale auxiliary array and a large-scale data transmission array. The process happens in two distinct stages:

  1. Stage 1: Coarse Estimation (The Scout): The small auxiliary array performs a rough scan. Instead of a simple sweep, it uses Singular Value Decomposition (SVD) on the received pilot signals. This allows the system to roughly estimate the Angles of Arrival (AoA) and Angles of Departure (AoD) for the primary signal paths.
  2. Stage 2: Accurate Estimation (The Sniper): Once the "neighborhood" of the best angles is known, the large-scale data array takes over. The authors frame the fine-tuning as a Compressed Sensing problem. They use the Compressive Sampling Matching Pursuit (CoSaMP) algorithm to exploit the inherent "sparsity" of the THz channel (meaning the signal only travels along a few distinct paths).

By using the same frequency for both arrays, the system ensures high "channel reciprocity," meaning the scout's findings are highly relevant to the main array's task.

Datasets

This study does not use a traditional "dataset" in the machine learning sense. Instead, it utilizes standard mathematical channel models for the THz band:

  • LOS (Line of Sight) Model: Incorporates spreading loss and molecular absorption loss.
  • NLOS (Non-Line of Sight) Model: Accounts for rough surface reflection loss.
  • Simulation Parameters: The authors simulated a 200GHz system with 64-antenna arrays for data and 16-antenna arrays for auxiliary tasks, assuming three propagation paths ($L=3$).

Key Insights and Ideas

  • Linear Complexity: The most significant breakthrough is that the complexity of this scheme is linear with respect to the number of signal paths ($L$). Traditional methods often scale with the number of antennas or beam combinations, which is much higher.
  • The Power of Sparsity: THz channels are naturally sparse; signals don't bounce around as much as lower frequencies. By treating beam training as a sparse recovery problem (Compressed Sensing), the authors can reconstruct the channel using far fewer measurements.
  • Hardware Efficiency: By using a small-scale hybrid architecture for the auxiliary stage, the system reduces the number of expensive RF chains needed, making the hardware more cost-effective.
  • Reciprocity Advantage: Unlike methods that use lower-frequency "out-of-band" signals to help find THz beams, this method uses the same frequency for both stages, avoiding the mismatch issues that plague cross-band estimation.

Key Results

  • Spectral Efficiency: Simulation results show that the proposed scheme's spectral efficiency is nearly identical to an exhaustive search (the gold standard for performance) but achieved at a fraction of the time/computational cost.
  • Outage Probability: The scheme significantly outperformed the "Digital Beamforming Architecture Assisted Hybrid Beam Training" (DAH-BT) method, especially in low Signal-to-Noise Ratio (SNR) environments.
  • Robustness: Even with strong path fading—a common headache in the THz band—the two-stage approach maintained a stable link where omni-directional approaches would typically fail.

Limitations and Future Work

  • Static Scenarios: The paper primarily focuses on a static or slow-moving environment. In highly mobile scenarios (like a fast-moving vehicle), the time-varying nature of the channel might require even faster adaptation.
  • Hardware Impairments: The study assumes relatively ideal hardware. Real-world THz components often suffer from phase noise and non-linearities that could impact the accuracy of the SVD and CoSaMP stages.
  • Multi-User Scaling: While the paper mentions the auxiliary array could estimate the next user's path while the main array transmits data, a full multi-user interference analysis was not the primary focus.

Summary updated

Abstract

Although Terahertz communication systems can provide high data rates, it needs high directional beamforming at transmitters and receivers to achieve such rates over a long distance. Therefore, an efficient beam training method is vital to accelerate the link establishment. In this study, we propose a low-complexity beam training scheme of terahertz communication system which uses a low-cost small-scale hybrid architecture to assist a large-scale array for data transmission. The proposed scheme includes two key stages: (1) coarse AoAs/AoDs estimation for beam subset optimization in auxiliary array stage, and (2) accurate AoAs/AoDs estimation by exploiting channel sparsity in data transmission array stage. The analysis shows that the complexity of the scheme is linear with the number of main paths, and thus greatly reduces the complexity of beam training. Simulation results have verified the better performance in spectral efficiency of the proposed scheme than that of the related work.
BibTeX
@misc{yang2021efficient,
  author = {Songjie Yang and Zhongpei Zhang and Zhenzhen Hu and Nuan Song and Hao Liu},
  title = {{Efficient Multi-Beam Training For Terahertz Wireless communications}},
  year = {2021},
  eprint = {2112.15346},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2112.15346},
}
Figure