⚡ electrical engineering

Taming the Bessel Landscape: Joint Antenna Position Optimization for Spatial Decorrelation in Fluid MIMO Systems

This paper addresses the joint transmitter and receiver antenna position optimization in fluid MIMO systems to maximize ergodic capacity by characterizing the non-convex "Bessel landscape" through analytical insights and proposing two alternating optimization algorithms—one based on particle swarm optimization and another on successive convex approximation—to effectively navigate this complex terrain.

Original authors: Tuo Wu, Kai-Kit Wong, Baiyang Liu, Kin-Fai Tong, Hyundong Shin

Published 2026-03-10

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Tuo Wu, Kai-Kit Wong, Baiyang Liu, Kin-Fai Tong, Hyundong Shin

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 you are trying to have a conversation in a crowded, noisy room.

In a traditional cell phone system (what we call MIMO), you have a fixed set of microphones (antennas) on the receiver and fixed speakers on the transmitter. They are bolted to the wall at specific distances. If the room is full of echoes (which happens in cities with lots of buildings), the sound bounces around in a way that makes the microphones hear the same echo at the same time. This is called spatial correlation. It's like having two microphones so close together that they just record the exact same noise, wasting half your recording power.

The Problem:
To get a clearer signal, engineers usually just add more microphones. But this is expensive, heavy, and uses a lot of battery power. Plus, if you pack them too tight, they still just hear the same echo.

The Big Idea: Fluid MIMO
This paper introduces a concept called Fluid MIMO. Imagine instead of bolted microphones, you have liquid microphones that can flow around. You can physically move the microphones to any spot within a certain range (like a flexible antenna array).

The goal is to move these "liquid" microphones to the exact spots where the echoes cancel each other out, making every microphone hear a unique, clear voice.

The Challenge: The "Bessel Landscape"
Here is where it gets tricky. The way sound waves (or radio waves) interfere with each other isn't a smooth hill you can just walk up. It's a rugged, bumpy landscape full of tiny hills and valleys.

The authors call this the "Bessel Landscape" (named after a math function called the Bessel function that describes how waves behave).

If you move your antenna a tiny bit to the left, the signal might get better.
If you move it a tiny bit to the right, it might get worse.
If you move it a bit further, it might get great again, then terrible again.

It's like trying to find the highest peak in a foggy mountain range where the ground goes up and down every few inches. Standard math tools usually get stuck in a small valley and think they've found the top.

The Solution: Two Smart Strategies
The authors developed two ways to navigate this bumpy terrain to find the best antenna positions:

The "Swarm of Birds" Approach (AO-PSO):
Imagine releasing a flock of birds into the foggy mountain range. Each bird flies randomly, looking for the highest point. They share information: "Hey, I found a high spot over here!" and "I found an even higher one!"
- Pros: They are very good at finding the absolute highest peak, even if the terrain is crazy.
- Cons: It takes a long time because the birds have to fly around a lot to check everything.
The "Super-Hiker" Approach (AO-SCA):
This is a smarter, faster method. Instead of flying randomly, this hiker has a special map that tells them exactly which way is "up" at their current location. They take a giant step in that direction, check the map again, and take another step.
- Pros: It is 100,000 times faster than the bird swarm. It zooms to the top almost instantly.
- Cons: It relies on the map being accurate. However, the authors proved that for this specific problem, the map is so good that the hiker finds the exact same spot as the birds, just much faster.

The Key Discoveries

The Magic Distance: For just two antennas, the authors found a "magic distance" (about 38% of a radio wavelength) where the antennas hear completely different signals. It's like finding the perfect spot to stand in a room so you don't hear the echo.
The "High SNR" Rule: When the signal is strong (like being close to the cell tower), the main goal is simply to make the antennas as "independent" as possible. The math shows that maximizing the "determinant" (a fancy math word for how spread out the signals are) is the key to speed.

The Results
When they tested this:

Old Way (Fixed Antennas): Struggled in the noisy room, getting about 40-50 bits of data per second.
New Way (Fluid Antennas): By moving the antennas to the perfect spots, they got over 7 bits per second more (which is a huge jump in data speed) compared to the old way.
Speed: The "Super-Hiker" algorithm found the solution in a fraction of a second, while the "Bird Swarm" took minutes.

In Summary
This paper is about teaching cell towers and phones how to "dance" with radio waves. Instead of being stuck in one spot, the antennas can flow to the perfect positions to avoid interference. The authors figured out the math to navigate the tricky "bumpy" terrain of wave interference and created two tools to do it: one that is thorough but slow, and one that is lightning-fast and just as accurate. This means faster internet and better connections without needing to build more expensive hardware.

1. Problem Statement

The paper addresses the optimization of Fluid MIMO systems, where both the transmitter (TX) and receiver (RX) utilize Fluid Antenna Systems (FAS). Unlike traditional Fixed-Position Antennas (FPAs), FAS allows antennas to move continuously within a physical aperture.

Core Challenge: In rich scattering environments, the spatial correlation between antennas is governed by the zeroth-order Bessel function ( $J_0(\cdot)$ ) of the inter-element spacing. This creates a highly non-convex optimization landscape (termed the "Bessel landscape") with numerous local optima due to the oscillatory nature of $J_0$ .
Objective: To maximize the ergodic capacity of the MIMO system by jointly optimizing the continuous positions of $N$ TX antennas and $M$ RX antennas.
Constraints: The optimization must respect physical aperture limits and a minimum inter-element spacing ( $d_{min}$ ) to avoid mutual coupling.
Coupling: The problem is difficult because the TX and RX positions are coupled through the Kronecker channel model ( $R = R_R \otimes R_T$ ), and the objective function does not admit a trivial decomposition.

2. Methodology

The authors propose a two-pronged approach: first, deriving analytical insights to understand the problem structure, and second, developing two algorithms to solve the optimization.

A. Analytical Insights

Before proposing algorithms, the paper establishes three key theoretical results:

High-SNR Capacity Approximation: At high Signal-to-Noise Ratio (SNR), the ergodic capacity decomposes into separable terms dependent on the log-determinants of the TX and RX correlation matrices:
$C(t, r) \approx N \log_2 \gamma + \log_2 \det(R_T) + \log_2 \det(R_R) + \kappa_N$
This reveals that maximizing capacity is asymptotically equivalent to maximizing $\det(R_T)\det(R_R)$ , i.e., minimizing spatial correlation.
Capacity Loss Bound: A closed-form bound is derived linking the performance gap relative to an ideal i.i.d. channel directly to $\det(R_T)\det(R_R)$ .
Optimal Spacing for $N=2$ : For a two-antenna system, the globally optimal spacing is derived in closed form as the first positive zero of $J_0$ , which is approximately $0.383\lambda$ . This is notably smaller than the conventional $0.5\lambda$ spacing.

B. Proposed Algorithms

The authors employ an Alternating Optimization (AO) framework, iteratively optimizing TX positions while fixing RX, and vice versa. Two solvers are proposed for the subproblems:

AO-PSO (Particle Swarm Optimization):
- A derivative-free global optimization method.
- Uses a swarm of particles to explore the non-convex Bessel landscape without gradient information.
- Includes a projection mechanism to enforce aperture and minimum spacing constraints.
- Pros: Robust against local optima; explores the global landscape.
- Cons: Computationally expensive due to Monte-Carlo capacity estimation.
AO-SCA (Successive Convex Approximation):
- A gradient-based local optimization method leveraging the high-SNR approximation.
- Derives a closed-form gradient of the objective function using the first-order Bessel function ( $J_1(\cdot)$ ), since $J_0'(x) = -J_1(x)$ .
- Constructs convex surrogates (quadratic approximations) at each iteration.
- Pros: Orders of magnitude faster than PSO; no Monte-Carlo sampling required.
- Cons: A local optimizer (though simulations show it finds the global optimum in this specific landscape).

3. Key Contributions

Problem Formulation: First formulation of joint TX/RX position optimization for Fluid MIMO under the exact Jakes correlation model with Kronecker structure.
Theoretical Breakthroughs:
- Proved that high-SNR capacity maximization is equivalent to maximizing the determinant of correlation matrices.
- Established a closed-form capacity loss bound.
- Derived the optimal $N=2$ spacing ( $0.383\lambda$ ).
Algorithmic Innovation:
- Developed AO-PSO for global exploration of the Bessel landscape.
- Developed AO-SCA, which utilizes the analytical gradient of the Bessel function to achieve rapid convergence.
Convergence Guarantees: Both algorithms are proven to be monotonically convergent.

4. Simulation Results

Extensive Monte-Carlo simulations validate the theoretical findings:

Performance Gain: The proposed schemes significantly outperform Fixed-Position Antennas (FPA). At high SNR (30 dB), the gain exceeds 7 bps/Hz compared to FPA.
Near-Optimal Performance: Both AO-PSO and AO-SCA achieve ergodic capacities very close to the i.i.d. MIMO upper bound (within ~0.5 bps/Hz), effectively decorrelating the channels.
Algorithm Comparison:
- AO-SCA achieves identical capacity to AO-PSO but is $10^5$ times faster (sub-second runtime vs. minutes).
- AO-SCA converges in 2–3 outer iterations, while AO-PSO requires 5–8.
Scalability: As the number of antennas ( $N$ ) increases, the capacity loss for FPA grows rapidly, whereas the proposed FAS optimization maintains near-i.i.d. performance.
Low-SNR Behavior: At low SNR, position optimization yields negligible gains (capacity is power-limited), confirming that the benefits of FAS are most pronounced in the moderate-to-high SNR regime.

5. Significance

This paper bridges a critical gap in Fluid Antenna System research by moving beyond single-antenna port selection to joint multi-antenna position optimization.

Practical Impact: It demonstrates that Fluid MIMO can achieve the theoretical capacity of ideal i.i.d. channels without increasing the number of antennas, simply by optimizing their physical placement.
Design Rule: The finding that optimal spacing can be smaller than $0.5\lambda$ (specifically $0.383\lambda$ for $N=2$ ) challenges conventional array design heuristics.
Computational Efficiency: The development of AO-SCA provides a practical, real-time capable solution for deploying Fluid MIMO in 6G networks, overcoming the computational barriers of non-convex optimization in oscillatory landscapes.

In summary, the paper successfully "tames" the complex Bessel landscape, proving that joint position optimization in Fluid MIMO is a viable and highly effective strategy for maximizing spectral efficiency in future wireless systems.