⚡ electrical engineering

Sparse Array Design for Near-Field MU-MIMO: Reconfigurable Array Thinning Approach

This paper proposes a reconfigurable array thinning approach for near-field MU-MIMO systems that dynamically activates antenna subsets to suppress grating lobes and maximize sum-rate, offering a practical alternative to fixed sparse or mechanically movable arrays without increasing hardware complexity.

Original authors: Ahmed Hussain, Asmaa Abdallah, Abdulkadir Celik, Emil Björnson, Ahmed M. Eltawil

Published 2026-05-22

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Ahmed Hussain, Asmaa Abdallah, Abdulkadir Celik, Emil Björnson, Ahmed M. Eltawil

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a future where your Wi-Fi router isn't just a little box with two antennas, but a massive wall covered in hundreds of tiny antennas. This "super-router" is designed to talk to many devices at once, even when they are standing at different distances from the wall, not just at different angles.

However, building a wall with hundreds of antennas is expensive, heavy, and uses a lot of electricity. The engineers in this paper asked: "Can we get the same super-powerful connection using far fewer antennas?"

Here is the simple breakdown of their solution, using some everyday analogies.

The Problem: The "Crowded Room" and the "Echo"

Think of the antenna wall like a choir.

The Goal: The choir wants to sing a specific note to a specific person in the crowd (a user's phone) without the sound leaking to others.
The Issue: If the choir members stand too far apart (which saves space and money), the sound waves bounce around and create "ghost echoes" called grating lobes. These echoes are like unwanted shout-outs to people standing in the wrong spots, causing confusion and noise.
The Old Solutions:
1. Fixed Sparse Arrays: You arrange the choir members in a specific pattern to avoid echoes. But if the crowd moves, the pattern is wrong, and the echoes come back.
2. Movable Antennas: You put the choir members on wheels so they can physically scoot around to the perfect spot for every new song. This works great, but it's slow, noisy, and the wheels break easily.

The New Idea: The "Smart Light Switch" Wall

The authors propose a clever middle ground. Imagine the wall is still covered in all 320 antennas (the full choir), but they are all connected to a smart switchboard.

Instead of moving the antennas, you simply turn them on or off.
You only activate the specific 32 antennas that are needed for the current situation, leaving the rest silent.
This is called Array Thinning. It's like having a full orchestra but only asking the specific musicians you need to play for a specific song, while the rest sit quietly.

The Two Strategies (The "How-To")

The team used a computer algorithm (called Particle Swarm Optimization, which acts like a swarm of birds searching for the best food source) to figure out exactly which antennas to turn on. They tried two different approaches:

1. The "Safety First" Approach (GTA)

The Goal: Stop the "ghost echoes" (grating lobes) from happening, no matter where the users are standing.
How it works: The computer designs a pattern that breaks up the regular spacing just enough to kill the echoes.
The Result: It works very well at stopping interference, similar to how a pre-arranged seating chart prevents people from bumping into each other. It's a "set it and forget it" solution.

2. The "Performance Max" Approach (STA)

The Goal: Get the absolute fastest internet speed (sum-rate) for the specific group of users right now.
How it works: The computer looks at exactly where the users are standing and picks the best 32 antennas to talk to them directly.
The Result: This is the winner. It performs almost as well as the "moving antenna" system (the one with the wheels) but without the mechanical headaches. It gets about 75% of the speed of the full 320-antenna wall while using only 10% of the hardware.

A Surprising Discovery: The "Distance" Myth

One of the paper's key findings is a bit like a magic trick.

In the old days, engineers knew that if you spaced antennas too far apart, you got "ghost echoes" in the direction (left/right).
They wondered: "What about distance (near/far)? Do we get echoes if someone is standing 10 meters away vs. 100 meters away?"
The Answer: No! The math shows that these "ghost echoes" only happen in the left/right direction. They do not happen in the distance direction. This means you can space antennas out to save money without worrying about "distance echoes" messing up the signal.

The Bottom Line

This paper suggests a practical way to build the super-fast wireless networks of the future. Instead of building expensive, heavy walls with thousands of antennas, or building complex, moving parts that break easily, we can just build a fixed wall and use a smart switch to turn on the right antennas at the right time.

It's like having a massive stadium of seats, but you only light up the specific seats needed for the current event. You get the same great view (performance) without the cost of filling the whole stadium or the hassle of moving the seats around.

Technical Summary: Sparse Array Design for Near-Field MU-MIMO via Reconfigurable Array Thinning

Problem Statement
Future wireless networks are expected to deploy massive antenna arrays operating in the radiative near-field (NF) regime. Unlike far-field systems where users are multiplexed solely in the angular domain, NF propagation allows for spatial multiplexing across both angle and range. However, realizing large-aperture arrays with half-wavelength spacing requires thousands of elements, leading to prohibitive hardware costs, power consumption, and computational complexity. While sparse arrays offer a solution by increasing inter-element spacing to reduce spatial correlation, fixed uniform sparse arrays suffer from grating lobes that cause strong interference. Existing alternatives, such as movable antenna (MA) architectures, rely on mechanical repositioning to adapt to user distributions, introducing latency and hardware complexity. Furthermore, conventional antenna selection methods typically rely on instantaneous channel state information (CSI) for performance metrics but do not explicitly formulate the problem as a structured sparse array design. There is a lack of solutions that provide dynamic adaptability without mechanical movement or excessive computational overhead, particularly regarding the behavior of grating lobes in the range domain.

Methodology
The authors propose a reconfigurable array thinning framework. In this approach, a full dense uniform linear array (FULA) remains physically fixed, but a subset of antennas is selectively activated via an RF switch network to form a flexible sparse array. The activation pattern is determined by a binary thinning vector $\mathbf{b}$ .

The paper first establishes a theoretical foundation regarding grating lobes in the near-field:

Angle Domain: Grating lobes occur when the phase shift between adjacent elements equals an integer multiple of $2\pi$ , dependent on the steering angle.
Range Domain: Through analysis of the NF array response vector, the authors demonstrate that grating lobes do not appear along the range dimension. The phase variation is quadratic rather than linear, preventing coherent re-alignment of element phases at specific distances. Instead, only small ripples are observed.

Based on this analysis, the authors develop two Particle Swarm Optimization (PSO)-based strategies to determine the optimal thinning vector $\mathbf{b}$ :

Grating-Lobe-Based Thinned Array (GTA): A pre-optimized design that minimizes the Peak Sidelobe Level (PSLL) in the angular domain to suppress grating lobes. This approach uses a continuous relaxation PSO to optimize the thinning pattern offline based on statistical constraints, ensuring the suppression of lobes across a coverage interval.
Sum-Rate-Based Thinned Array (STA): A dynamic design that directly maximizes the multi-user sum-rate. This method optimizes the thinning vector online (or for specific channel realizations) using the achievable sum-rate as the objective function, assuming knowledge of user locations (angle and range).

Both methods utilize a PSO framework where particles represent continuous priority vectors mapped to binary activation decisions, enforcing a fixed number of active antennas ( $N_T$ ) and preserving edge elements to maintain aperture size.

Key Contributions

Theoretical Insight: The paper provides the first analysis confirming that while grating lobes exist in the angular domain for sparse arrays in the near-field, they are absent in the range domain. This finding simplifies the design constraints for NF sparse arrays.
Novel Framework: It introduces the first application of array thinning specifically for near-field communication, distinguishing it from conventional antenna selection by treating it as a structured sparse array design problem.
Dual Optimization Strategies: The proposal of two distinct PSO-based strategies:
- GTA: Offers a practical, pre-optimized solution that suppresses grating lobes without requiring real-time user tracking.
- STA: Offers a high-performance solution that dynamically maximizes sum-rate, bridging the gap between fixed sparse arrays and complex movable antenna systems.
Efficiency: The approach avoids the mechanical latency and complexity of movable antennas while offering better adaptability than static sparse arrays.

Simulation Results
The proposed designs were evaluated against benchmarks including a full ULA (FULA), a movable ULA (MULA), a pre-optimized thinned array (PTA), and a uniform sparse ULA (SULA). Key findings include:

Grating Lobe Suppression: GTA effectively suppresses grating lobes in the angular domain across various ranges, outperforming conventional SULA which exhibits strong lobes.
Sum-Rate Performance:
- The STA achieves performance comparable to the MULA (which requires physical repositioning) and attains approximately 75% of the FULA sum-rate while using only 10% of the active elements (32 out of 320 antennas).
- The GTA performs similarly to the PTA, demonstrating that grating-lobe suppression effectively reduces interference.
- While GTA and PTA have slightly lower sum-rates (approx. 5% lower) than STA due to elevated sidelobes from irregular spacing, they offer the advantage of offline optimization.
Range Domain Behavior: Simulations confirm that when users are aligned in angle but separated in range, the SULA performs nearly identically to the FULA, validating the theoretical absence of range-domain grating lobes.
Complexity: The PSO-based STA is computationally more efficient than MULA optimization because it operates on a discrete selection space rather than a continuous position space, requiring fewer iterations to converge.

Significance and Claims
The paper claims that reconfigurable array thinning offers a practical and efficient deployment strategy for near-field MU-MIMO systems. By eliminating the need for mechanical movement (unlike MULA) and overcoming the suboptimality of fixed sparse arrays under dynamic user distributions, the proposed framework balances hardware efficiency with high spectral efficiency. The authors modestly note that while STA requires frequent updates based on channel geometry, GTA and PTA provide robust pre-optimized alternatives. The work suggests that array thinning enables graceful performance degradation and hardware efficiency, as faulty elements can be deactivated and the configuration re-optimized. Future work is identified as exploring deep learning-based solutions to further mitigate the computational complexity of the current PSO approach.

The Problem: The "Crowded Room" and the "Echo"

The New Idea: The "Smart Light Switch" Wall

The Two Strategies (The "How-To")

A Surprising Discovery: The "Distance" Myth

The Bottom Line

More like this