Spatial Non-Stationarity in Extremely Large Antenna Arrays

20 concepts11 papers

In traditional wireless communications, channel modeling relies on a foundational assumption: spatial stationarity. When a base station receives a signal, we assume that every antenna element in its array shares an identical statistical view of the surrounding environment. Under this assumption, a signal emitted by a user arrives at the array as a uniform plane wave, experiencing the same scattering clusters, obstructions, and path loss across the entire receiver.

This mathematical convenience holds true when antenna arrays are physically small. However, as wireless networks transition toward 6G, systems are scaling up to Extremely Large-Scale MIMO (XL-MIMO) and Holographic MIMO (H-MIMO), where arrays feature thousands of elements and stretch across entire walls or building facades. At this physical scale, the assumption of a uniform channel environment completely breaks down, giving rise to a phenomenon known as spatial non-stationarity (SnS).

The Physics of Aperture-Level Variation

Why does spatial stationarity fail in extremely large arrays? The answer lies in the sheer physical dimensions of the antenna aperture relative to its environment.

When an antenna array spans several meters, different segments of the same physical aperture "see" entirely different worlds. As Shuhang Yang et al. illustrate in their work, one end of a massive array might enjoy a perfectly clear line-of-sight to a mobile user, while the opposite end is entirely obstructed by an indoor pillar or an outdoor concrete structure.

[User] 
  │  \
  │   \  (Clear Path)
  ▼    \
[Blocker] \
  │        ▼
  ▼      [==================================]
(Obstructed End)      Extremely Large Array      (Clear End)

This aperture-level variation is further compounded by two distinct physical realities:

  1. Near-Field Spherical Waves: As detailed by Ouyang et al. in their comprehensive survey, the massive physical size of these arrays pushes the boundaries of the "near-field" region much further out. Signals no longer arrive as flat, parallel plane waves; instead, they arrive as spherical wavefronts. This means that the distance, phase shift, and angle of arrival vary substantially from one antenna element to the next across the array surface.
  2. Localized Scattering: Early empirical evidence of this spatial variation in high-frequency bands was captured by Lin et al. in their millimeter-wave (mmWave) propagation measurements. They demonstrated that channel properties fluctuate significantly depending on the specific spatial location of the antenna elements, meaning that different parts of an array interact with entirely distinct environmental scatterers.

Mapping the Environment via Visibility Regions

To mathematically capture and exploit this non-stationary behavior, researchers use the concept of a Visibility Region (VR). A visibility region defines the specific subset of antenna elements on an aperture that can actively "see" a particular user or a specific scattering cluster.

As Gong et al. emphasize in their near-field channel modeling framework, we can no longer model the channel as a single monolithic matrix. Instead, the channel must be modeled locally across the physical surface of the antenna.

Crucially, these visibility regions are not always simple, contiguous blocks of antennas. Ren et al. demonstrate in their research that fragmented obstructions—such as a picket fence or a row of columns—create non-contiguous visibility regions. This causes the signal energy to flicker on and off across the physical extent of the array, introducing a "picket-fence effect" that standard stationary models fail to predict.

Why Traditional Signal Processing Fails

If a system design assumes spatial stationarity when the channel is actually non-stationary, its performance degrades severely. Traditional channel estimation algorithms attempt to compute a single, global channel profile or angular direction for the entire array. In an XL-MIMO system, this forces the receiver to average out localized physical variations.

As noted by Zhe Wang et al. in their analytical framework for Effective Degrees of Freedom (EDoF), ignoring these variations leads to massive inaccuracies when predicting communication capacity and signal performance. If the receiver processes noise from obstructed antenna elements alongside the true signal from visible elements, the overall signal-to-noise ratio (SNR) plummets.

Aperture-Aware Solutions

To overcome the challenges of spatial non-stationarity, next-generation signal processing must be explicitly visibility-region-aware. Rather than treating the array as a single entity, modern frameworks dynamically adapt to the localized conditions of the aperture.

Joint Detection and Estimation

Instead of treating visibility region detection and channel estimation as separate problems, Tang et al. proposed a two-stage joint framework in their paper. By first isolating which specific antennas are actually receiving meaningful energy from a user (the VR) and ignoring the rest, the system avoids processing noise from obstructed elements. This approach drastically improves channel estimation accuracy, particularly in low-SNR conditions.

Adaptive Subarray Segmentation

To bridge the gap between complex non-stationary realities and low-complexity algorithms, Shuhang Yang et al. introduced an adaptive subarray segmentation paradigm in their work. This technique dynamically divides the massive array into smaller sub-apertures that are small enough to be considered locally stationary. This allows engineers to deploy classic, well-understood estimation tools within each sub-aperture while still capturing the global non-stationarity across the entire array.

Hardware-Constrained Optimization

Scaling up to thousands of antennas introduces severe hardware and power constraints. Maly et al. address this in their paper by evaluating XL-MIMO systems that utilize power-efficient 1-bit analog-to-digital converters (ADCs). By combining Bussgang linearization, random dithering, and angular-domain optimization, the authors demonstrate that robust channel estimation is entirely achievable even when the radio environment is spatially non-stationary and the hardware is heavily quantized.

New Frontiers: Beam Focusing and Sensing

While spatial non-stationarity complicates channel estimation, it also unlocks entirely new capabilities for 6G networks. In a tutorial on 6G XL-MIMO, Zhe Wang et al. explain in their overview that identifying localized visibility regions allows systems to optimize energy efficiency by dynamically turning off or ignoring inactive antenna segments.

Furthermore, the transition to near-field, non-stationary modeling enables a paradigm shift from beamforming to beam focusing. As discussed by Dezhi Wang et al. in their theoretical exploration, modeling the aperture-dependent channel allows a base station to focus electromagnetic energy into a specific three-dimensional spot (a voxel) rather than merely radiating energy along a two-dimensional angular direction.

To make this practical without incurring massive training overhead, Liu et al. demonstrate in their paper that low-cost power sensors can be paired with a "time inversion" algorithm to first sense and localize scatterers. By aggregating data across multiple subcarriers, this sensing-assisted approach resolves the inconsistent visibility across the array, slashing baseband sampling overhead by up to 66% and making aperture-level processing viable for real-world 6G deployment.

Key Foundational Work

Millimeter-Wave Propagation Modeling and Measurements for 5G Mobile Networks (2019) — contributed 3 concept elements to this direction.
A Tutorial on Extremely Large-Scale MIMO for 6G: Fundamentals, Signal Processing, and Applications (2023) — contributed 2 concept elements to this direction.
Plug-in Channel Estimation with Dithered Quantized Signals in Spatially Non-Stationary Massive MIMO Systems (2023) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_spatial_non_stationarity_aperture_level__9ca9b9f8,
  title = {spatial non-stationarity (aperture-level spatial non-stationarity)},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/spatial-non-stationarity-aperture-level-spatial-non-stationarity/rdr_9ca9b9f8207e4450b89c7b49f7ccb654},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-21},
}

This exact version (version 1, frozen text):

@misc{lacuna_spatial_non_stationarity_aperture_level__9ca9b9f8_v1,
  title = {spatial non-stationarity (aperture-level spatial non-stationarity)},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/spatial-non-stationarity-aperture-level-spatial-non-stationarity/rdr_9ca9b9f8207e4450b89c7b49f7ccb654/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-21},
}

See something wrong? Report a mistake