🔢 mathematics

Information-Energy Capacity Region for SLIPT Systems over Lognormal Fading Channels: A Theoretical and Learning-Based Analysis

This paper provides a theoretical and learning-based analysis of the information-energy capacity region for SLIPT systems over lognormal fading channels, proving that the optimal input distribution is discrete and proposing a GAN-based framework for efficient capacity estimation and optimization.

Original authors: Nizar Khalfet, Kapila W. S. Palitharathna, Symeon Chatzinotas, Ioannis Krikidis

Published 2026-04-27

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Nizar Khalfet, Kapila W. S. Palitharathna, Symeon Chatzinotas, Ioannis Krikidis

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 run a high-tech, solar-powered flashlight in the middle of a stormy ocean. You want this flashlight to do two things at once: send a secret coded message (Information) and recharge its own battery (Energy) using the light it emits.

This paper, written by researchers at the University of Cyprus and the University of Luxembourg, is essentially a "master manual" for how to do this perfectly when the environment is messy and unpredictable.

Here is the breakdown of their work using everyday analogies:

1. The Problem: The "Wobbly Mirror" Effect (Lognormal Fading)

In a perfect world, if you shine a light, it travels in a straight, steady beam. But in the real world—especially underwater or in the atmosphere—the water is murky, and the air is turbulent.

Think of it like trying to shine a laser pointer at a friend through a piece of wavy, moving glass. The beam doesn't just get dimmer; it flickers and dances unpredictably. This is what scientists call "Lognormal Fading." Most previous studies assumed the light was steady, but this paper says, "Wait, the world is wobbly! We need to account for the wobbles if we want our gadgets to actually work."

2. The Dilemma: The "Budgeter’s Headache" (The Trade-off)

The researchers are dealing with a very difficult balancing act. Imagine you have a limited amount of "Light Juice" (Power).

If you use all the light to send a very complex, high-speed message, you won't have enough left to recharge the battery.
If you use all the light to charge the battery, your message will be too slow or too simple to understand.

Furthermore, the battery (the "Energy Harvester") isn't perfect. It’s like a sponge that gets harder to soak up water the more saturated it becomes (this is the Nonlinear Energy Harvesting part). You can't just throw more light at it and expect a linear increase in power.

3. The Discovery: "Don't Be Smooth, Be Choppy" (Optimal Input Distribution)

If you were trying to send a message with light, you might think the best way is to vary the brightness smoothly, like a dimmer switch.

However, the researchers proved mathematically that smoothness is a waste of energy. They found that the most efficient way to communicate while charging is to use "Mass Points"—which is a fancy way of saying you should use specific, distinct "steps" of brightness (like a staircase rather than a ramp). They proved that instead of a continuous stream of different brightness levels, you only need a few specific "clicks" of brightness to get the best results.

4. The Solution: The "AI Tutor" (CIECL Framework)

Because the math for "wobbly light + nonlinear batteries + power limits" is incredibly hard for a standard computer to solve, the researchers brought in an AI (a Generative Adversarial Network, or GAN).

Think of this AI as two students competing in a game:

Student A (The Generator): Tries to create the perfect "light pattern" to send the message and charge the battery.
Student B (The Discriminator): Acts like a strict teacher. It looks at the patterns and tries to guess if they are "real, efficient patterns" or "fake, inefficient ones."

As they play this game millions of times, Student A becomes a master at finding the exact "staircase" of brightness levels that maximizes the message speed without letting the battery die.

Summary: Why does this matter?

As we move toward 6G technology and the Internet of Things (IoT), we will have billions of tiny sensors underwater or in the air. These sensors can't be plugged into a wall; they need to "eat" light to survive.

This paper provides the mathematical blueprint and the AI tools to ensure these tiny devices can talk to us and power themselves at the same time, even when the environment is trying to shake the signal apart.

Technical Summary: Information-Energy Capacity Region for SLIPT Systems over Lognormal Fading Channels

This paper presents a rigorous information-theoretic and machine-learning-based analysis of Simultaneous Lightwave Information and Power Transfer (SLIPT) systems. It specifically addresses the challenges posed by lognormal fading, which is a dominant phenomenon in optical wireless communication (OWC) environments such as underwater and atmospheric channels.

1. The Problem

Traditional studies of SLIPT often assume Additive White Gaussian Noise (AWGN) channels with deterministic path loss. However, real-world optical channels suffer from turbulence-induced fading. The authors identify a gap in existing literature regarding:

Channel Modeling: The impact of lognormal fading on the information-energy trade-off.
Nonlinear Constraints: The complexity introduced by practical, nonlinear Energy Harvesting (EH) models (specifically photovoltaic cell behavior) combined with Peak-Power (PP) and Average-Power (AP) constraints.
Non-Coherent Setting: The lack of Channel State Information (CSI) at both the transmitter and receiver, which is a realistic constraint for low-complexity IoT and sensor nodes.

The core objective is to characterize the information-energy capacity region—the set of all achievable rate-power pairs—under these complex, non-convex constraints.

2. Methodology

The paper employs a dual-pronged approach combining mathematical optimization and deep learning:

A. Theoretical Framework (Information Theory):

System Model: A point-to-point SLIPT system using Intensity Modulation and Direct Detection (IM/DD). The transmitter uses an LED/LD, while the receiver uses a Photodiode (PD) for information and a Photovoltaic (PV) cell for energy.
Mathematical Optimization: The problem is formulated as maximizing ergodic mutual information subject to:
1. $0 \leq x \leq A$ (Peak Power)
2. $\mathbb{E}\{X\} \leq \epsilon$ (Average Power)
3. $\mathbb{E}\{bX \ln(1 + cX)\} \geq E_{th}$ (Nonlinear EH requirement)
Smith’s Framework: The authors extend Smith’s framework to prove that the optimal input distribution is discrete with a finite number of mass points, even under lognormal fading and nonlinear EH constraints.

B. Learning-Based Framework (CIECL):

CIECL (Cooperative Information-Energy Capacity Learning): The authors propose a novel framework based on Generative Adversarial Networks (GANs).
Architecture: A Generator (G) learns to produce the capacity-achieving input distribution, while a Discriminator (D) distinguishes between paired (true) and unpaired (shuffled) channel input-output samples to estimate mutual information.
Constraint Enforcement: Unlike standard GANs, the CIECL modifies the loss function with penalty terms (Lagrange multipliers) to explicitly enforce PP, AP, and nonlinear EH constraints.

3. Key Contributions

Derivation of Transition Probability: A novel explicit expression for the transition probability distribution of the lognormal fading channel using Hermite polynomial bases.
Discreteness Proof: A mathematical proof that the optimal input distribution for this specific SLIPT setting is discrete.
High-SNR Characterization: An analytical expression for an achievable input distribution in the high-SNR regime.
Transition Point Analysis: Identification of the critical threshold where a simple binary input (like OOK) ceases to be optimal and multi-level modulation becomes necessary.
CIECL Framework: A data-driven method that outperforms traditional iterative algorithms (like Blahut-Arimoto) in handling non-convex, nonlinear constraints.

4. Results and Discussion

Optimal Distribution Behavior:
- At low Peak-Power (PP) levels, the optimal distribution is binary.
- As the EH requirement ( $E_{th}$ ) increases, the mass points shift toward higher amplitudes to satisfy energy needs, creating a trade-off that reduces the information rate.
Capacity Region: The results show a clear rate-power trade-off. When the EH constraint is inactive, the energy rate can increase without affecting the information rate. Once the constraint becomes active, any further increase in energy harvesting necessitates a reduction in information rate.
Channel Impact: Lognormal fading significantly shrinks the capacity region compared to the idealized AWGN channel due to signal attenuation and fluctuations.
Algorithm Performance: The GAN-based CIECL framework demonstrates superior performance in maximizing mutual information and higher efficiency in exploring the constrained input space compared to the Blahut-Arimoto (BA) algorithm.

5. Significance

This work provides a fundamental theoretical foundation for designing future 6G and IoT optical wireless systems. By proving that finite-alphabet modulation is optimal and providing a learning-based tool to optimize it, the paper offers practical design guidelines for engineers to balance high-speed communication with reliable energy delivery in challenging, turbulent environments.