Sense-Then-Train: An Active-Sensing-Based Beam Training Design for Near-Field MIMO Systems

TL;DR

This work tackles beam training for near-field ELAA MIMO, where conventional codebooks are inefficient due to spherical-wave propagation and extra DoFs. It introduces sense-then-train (STT), a codebook-free, active-sensing framework that first extracts low-dimensional LoS subspaces in the wavenumber domain using a sensing phase, then trains beamformers online via neural networks within this truncated space; for multi-beam transmission, Gram-Schmidt enforces inter-beam orthogonality. The method yields fast, low-complexity beam training with near-optimal spectral efficiency compared to CSI-based benchmarks, and it supports both single- and multi-beam cases with online adaptation. The approach leverages ping-pong pilots and locally constructible truncated WTMs to reduce overhead while exploiting the near-field DoFs, making it practical for large-scale arrays and high-frequency bands. Overall, STT offers a scalable, adaptive beam-training paradigm for near-field MIMO that preserves performance while dramatically reducing CSI and training overheads.

Abstract

An active-sensing-based sense-then-train (STT) scheme is proposed for beam training in near-field multiple-input multiple-output (MIMO) systems. Compared to conventional codebook-based schemes, the proposed STT scheme is capable of not only addressing the complex spherical-wave propagation but also effectively exploiting the additional degrees-of-freedoms (DoFs). The STT scheme is tailored for both single-beam and multi-beam cases. 1) For the single-beam case, the STT scheme first utilizes a sensing phase to estimate a low-dimensional representation of the near-field MIMO channel in the truncated wavenumber domain. Then, in the subsequent training phase, the neural network modules at transceivers are updated online to align beams, utilizing sequentially received ping-pong pilots. This approach can efficiently obtain the aligned beam pair without relying on predefined codebooks or training datasets. 2) For the multi-beam case, based on the single-beam STT, a Gram-Schmidt method is further utilized to guarantee the orthogonality between beams in the training phase. Numerical results unveil that 1) the proposed STT scheme can significantly enhance the beam training performance in the near field compared to the conventional far-field codebook-based schemes, and 2) the proposed STT scheme can perform fast and low-complexity beam training, while achieving a near-optimal performance without full channel state information in both cases.

Figures (16)

• Figure 1: An illustration of a near-field MIMO system.
• Figure 2: An illustration of channel representation in the wavenumber domain, i.e., $|\tilde{\mathbf{H}}_{\mathrm{a}}|$, under $M=N=255$, $d_{\rm BU}=15~m$, and $f=28~{\rm GHz}$.
• Figure 3: An illustration of the single-beam STT scheme.
• Figure 4: An overview the of proposed STT method for the single-beam case. Data/gradient flows are denoted by the black/red line.
• Figure 5: An illustration of the multi-beam STT scheme.