Flexible XL-MIMO via Array Configuration Codebook: Codebook Design and Array Configuration Training

Haiquan Lu, Hongqi Min, Yong Zeng, Shaodan Ma

Abstract

Extremely Large-Scale MIMO (XL-MIMO) is a key technology for achieving the high data rates and spatial resolution required for 6G. However, the hardware complexity and energy consumption of fully digital XL-MIMO systems are prohibitive. This paper proposes the Array Configuration Codebook (ACC) framework, which utilizes a limited number of radio-frequency (RF) chains to dynamically activate a subset of "antenna pixels" in structured geometric patterns. By constraining the antenna selection problem to a predefined codebook of electromagnetic geometries, the proposed method significantly reduces computational complexity and training overhead while maintaining high spectral efficiency and localization accuracy.

Introduction

As wireless communication moves toward the 6G era, the demand for terabit-per-second data rates has led to the development of Extremely Large-Scale MIMO (XL-MIMO). By increasing the number of antenna elements to hundreds or thousands, XL-MIMO offers unprecedented spatial resolution. However, a fully digital architecture, where each antenna element is paired with a dedicated RF chain, is impractical due to extreme costs and power requirements.

Antenna Selection (AS) has emerged as a cost-effective alternative, connecting a small number of RF chains to a large pool of antenna pixels. The primary challenge in AS is the combinatorial explosion of possible antenna subsets; for an array with 256 pixels and 32 RF chains, the search space is astronomically large. This paper introduces the Array Configuration Codebook (ACC) to transform this intractable search into a structured selection process based on geometric principles.

System Architecture and the ACC Framework

The proposed system model connects $N$ RF chains to $M$ antenna pixels ($M \gg N$) through an ACC-based selection network. This architecture allows the base station to adapt its physical array configuration to the specific needs of the environment, whether for multi-user communication or high-precision sensing.

Figure 1: System architecture of the ACC-based XL-MIMO system, showing the mapping from RF chains to antenna pixels.

Instead of searching for arbitrary antenna combinations, the ACC restricts configurations to "codewords" derived from established electromagnetic geometries. These include:

  • Compact Array (CA): Standard half-wavelength spacing for robust performance in dense environments.
  • Uniform Sparse Array (USA): Increased element spacing to enhance spatial resolution.
  • Modular Array (MoA): Sub-arrays grouped for practical deployment constraints.
  • Nested Array (NA) & Co-prime Array (CPA): Non-uniform structures designed to create large virtual apertures, particularly beneficial for sensing and localization.
Figure 2: Detailed parameters and geometric structures for the various array configurations included in the ACC.

Array Configuration Training

To efficiently identify the optimal codeword, the authors developed a Two-Stage Scanning Scheme. The first stage, Array-Level Scanning, performs a coarse search across different geometric architectures to identify the best general structure for the current channel conditions. The second stage, Pixel-Level Scanning, performs a fine-grained search by shifting the reference point of the selected architecture pixel-by-pixel across the large-scale array. This hierarchical approach reduces the training overhead from thousands of exhaustive scans to approximately 1,120, making real-time adaptation feasible.

Key Insights and Performance Results

The study reveals that there is no universal "best" array configuration; rather, the optimal geometry depends on the deployment scenario:

  • Communication Efficiency: In scenarios with fewer users, the Uniform Sparse Array (USA) significantly outperforms the Compact Array (CA) by providing higher spatial resolution. For example, in a 10-user simulation, the USA achieved a sum rate of ~115 bps/Hz compared to ~50 bps/Hz for the CA. However, as user density increases, the grating lobes of sparse arrays can cause interference, making the CA more robust.
  • Localization Accuracy: For sensing tasks, Nested Arrays (NA) are superior. By exploiting the "difference co-array" concept, NAs create a continuous virtual aperture that allows for high-precision Angle of Arrival (AoA) estimation, even when the number of users exceeds the number of physical RF chains.

Conclusion

The Array Configuration Codebook (ACC) provides a flexible and computationally efficient framework for XL-MIMO. By leveraging geometric shortcuts, the system can dynamically reconfigure its antenna array to optimize for either spectral efficiency or localization accuracy. Future work will focus on extending this framework to stochastic channel models, downlink beamforming, and 2D planar array structures.

Summary updated

BibTeX
@article{lu2025flexible,
  author = {Haiquan Lu and Hongqi Min and Yong Zeng and Shaodan Ma},
  title = {{Flexible XL-MIMO via Array Configuration Codebook: Codebook Design and Array Configuration Training}},
  journal = {IEEE Trans. Communications},
  year = {2025},
  doi = {10.1109/tcomm.2025.3622989},
}

Loading related content...

Figure