⚡ electrical engineering

Constellation Design for Robust Interference Mitigation

This paper proposes a low-complexity "ML-G" detector and an interference-aware constellation design optimization to improve symbol detection performance in single-carrier communication systems subject to non-Gaussian Nakagami- $m$ interference.

Original authors: Athanasios T. Papadopoulos, Thrassos K. Oikonomou, Dimitrios Tyrovolas, Sotiris A. Tegos, Panagiotis D. Diamantoulakis, Panagiotis Sarigiannidis, George K. Karagiannidis

Published 2026-02-12

📖 3 min read☕ Coffee break read

CC BY 4.0

Original authors: Athanasios T. Papadopoulos, Thrassos K. Oikonomou, Dimitrios Tyrovolas, Sotiris A. Tegos, Panagiotis D. Diamantoulakis, Panagiotis Sarigiannidis, George K. Karagiannidis

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 with a friend in a crowded, noisy coffee shop. This paper is essentially a scientific blueprint for how to "tune your ears" and "change how you speak" so that you can understand each other, even when the background noise is incredibly loud and unpredictable.

Here is the breakdown of the paper using everyday analogies:

1. The Problem: The "Unfair" Noise

In most communication systems (like your cell phone), engineers assume the background noise is like a steady, gentle hum of a fan—predictable and "even" in all directions. This is what scientists call Gaussian Noise.

However, this paper looks at a much nastier kind of noise: Structured Interference.

The Analogy: Imagine instead of a steady fan, the noise is a group of people nearby having a heated, rhythmic argument. Sometimes they shout (high amplitude), sometimes they whisper, and their voices come in waves (phase). Because this noise has a "pattern" and "direction," the standard way your phone tries to filter it out fails. It’s like trying to use a simple earplug to block out a rhythmic drumbeat; it just doesn't work.

2. The Solution Part A: The "Smart Ear" (The ML-G Detector)

Since the noise isn't a steady hum, the researchers realized we need a smarter way to listen. They developed something called the ML-G Detector.

The Analogy: Imagine you are at that coffee shop. Instead of just trying to hear "loud vs. quiet" sounds, you start to realize, "Okay, every time the person at the next table slams their cup, there is a specific rhythmic pattern to the noise."
The ML-G detector is like a brain that has learned the "rhythm" and "shape" of the interference. Instead of just looking for the clearest sound, it calculates the probability of what your friend said by accounting for the specific "shape" of the noise. It doesn't just ask, "Which word is loudest?" It asks, "Given the way that specific noise is currently pulsing, which word is most likely to be the real one?"

3. The Solution Part B: The "Smart Speech" (Constellation Design)

The researchers didn't stop at just listening better; they also suggested changing how we speak. In digital communication, we send information using "constellations"—sets of specific signal patterns (like different musical notes).

The Analogy: If you know the coffee shop is noisy in a specific way—say, people are mostly shouting at a certain pitch—you wouldn't try to communicate using very subtle, similar-sounding notes. Instead, you would pick notes that are very far apart or notes that sit in the "quiet gaps" of the noise.
The paper uses math to "rearrange" the symbols (the notes). When the noise gets intense, the "constellation" (the map of your notes) stops looking like a perfect, pretty circle and starts looking like a weird, asymmetric shape. It might look "squashed" or "stretched," but it is mathematically optimized to stay readable even when the interference is trying to drown it out.

The "Too Long; Didn't Read" Summary

Most tech tries to fight noise by simply turning up the volume. This paper says: "Don't just turn up the volume; understand the rhythm of the noise to listen better, and change the pattern of your message to dance around the noise."

By doing both—using a smarter "ear" (the detector) and a smarter "voice" (the constellation)—they proved that communication can remain clear even in environments that would normally cause a total blackout.

Technical Summary: Constellation Design for Robust Interference Mitigation

This paper addresses the challenge of reliable symbol detection and constellation design in wireless communication systems operating in environments with strong, structured, and non-Gaussian interference.

1. The Problem: Detection Mismatch

Conventional receiver designs typically assume Additive White Gaussian Noise (AWGN). However, in modern spectrum-sharing scenarios (e.g., coexistence between communication and sensing/radar systems), receivers encounter interference characterized by:

Non-Gaussian Statistics: The interference amplitude follows a Nakagami- $m$ distribution rather than a Gaussian one.
Non-Uniform Phase: The interference phase is not uniformly distributed but exhibits structured behavior due to multipath scattering and waveform characteristics.
High Power: The interference power (INR) can be comparable to or even exceed the signal power (SNR).

When these statistical properties are ignored, standard Euclidean distance-based detectors suffer from "detection mismatch," leading to suboptimal decision regions and significantly higher Symbol Error Probabilities (SEP).

2. Methodology

The authors propose a two-pronged approach: a new detection metric and an interference-aware constellation optimization framework.

A. The ML-G Detector (Detection Framework):
To achieve near-optimal performance without the computational burden of a full Maximum Likelihood (ML) search, the authors develop the Maximum-Likelihood Gaussian-phase approximate (ML-G) detector.

Statistical Modeling: The interference is modeled using a Nakagami- $m$ envelope-phase distribution.
Gaussian Moment Matching: To make the likelihood function tractable, the non-uniform phase distribution is approximated by a Gaussian distribution via moment matching.
Series Representation: The resulting detection metric is expressed as an infinite series involving confluent hypergeometric functions. The authors mathematically prove that this series converges rapidly, allowing for efficient implementation via a small number of terms (truncation).

B. Constellation Design (Optimization Framework):
Since the ML-G detector creates asymmetric and non-standard decision regions, traditional symmetric constellations (like standard QAM or PSK) are no longer optimal.

Optimization Problem: The authors formulate an optimization problem to minimize the average SEP subject to an average energy constraint.
Global/Local Search: They employ a differential-evolution algorithm for global exploration of the symbol placements, followed by local refinement to ensure energy constraints and inter-symbol separation are met.

3. Key Contributions

Derivation of the ML-G Metric: A computationally efficient, closed-form approximate ML detector that explicitly incorporates joint amplitude and phase statistics.
Geometric Analysis: A characterization of how interference "warps" decision boundaries in the complex plane, moving them away from standard Voronoi cells.
Interference-Adaptive Constellations: The development of a method to design signal alphabets that "shape" themselves to the specific geometry of the interference-induced decision regions.
Convergence Guarantees: Mathematical proofs (Propositions 2 and 3) ensuring that the proposed series-based detector can be truncated with a guaranteed error tolerance ( $\epsilon$ ).

4. Results and Performance

Detection Gains: Simulation results show that the ML-G detector significantly outperforms both the conventional Euclidean distance detector and the Constant Amplitude Interference (CAI) model, particularly in interference-limited regimes ( $\gamma_I \gg \gamma_S$ ).
Adaptive Geometry: Optimized constellations exhibit highly non-standard, asymmetric structures. For example, at high interference levels, a 16-ary constellation may evolve into two parallel layers to maximize separability along the "cleaner" axes of the decision regions.
Robustness: The proposed approach provides a continuous transition from standard Euclidean behavior (when interference is weak) to highly specialized, robust geometries (when interference is strong).

5. Significance

This work shifts the paradigm of interference mitigation from external suppression (e.g., spatial filtering or temporal avoidance) to internalized robustness. By embedding the statistical "fingerprint" of the interference directly into the receiver's decision logic and the transmitter's signal structure, the paper provides a mathematically rigorous framework for achieving reliable communication in increasingly crowded and complex spectral environments.