⚡ electrical engineering

Mobility Aware Power Control for VCSEL Based Indoor OWC

This paper proposes a mobility-aware power control framework for VCSEL-based indoor optical wireless communication that utilizes a hybrid Gauss-Markov and learning-based approach to predict channel variations caused by user movement and orientation, thereby improving energy efficiency compared to conventional methods.

Original authors: Walter Zibusiso Ncube, Ahmad Adnan Qidan, Taisir El-Gorashi, Jaafar M. H. Elmirghani

Published 2026-04-27

📖 3 min read☕ Coffee break read

CC BY 4.0

Original authors: Walter Zibusiso Ncube, Ahmad Adnan Qidan, Taisir El-Gorashi, Jaafar M. H. Elmirghani

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 at a crowded, high-tech indoor music festival. Instead of using Wi-Fi (which is like a giant, messy floodlight that shines everywhere), the venue uses "Laser Spotlights" (VCSELs) to send data to people’s phones.

These spotlights are incredibly fast and can send massive amounts of information, but there is a catch: the beams are extremely narrow. If you move just a few inches to the left, or if you tilt your phone slightly to take a selfie, you might step right out of the beam, and your connection will drop instantly.

The Problem: The "Guessing Game"

Currently, most systems handle this in one of three inefficient ways:

The "Too Much" Approach (Conservative): To make sure you don't lose connection, the system blasts the spotlight at maximum brightness all the time. It’s like leaving a massive stadium floodlight on 24/7 just because you might walk into the room. It works, but it wastes a huge amount of electricity.
The "Too Late" Approach (Reactive): The system waits until your signal actually dies before it tries to turn up the power. This is like waiting until you’re stumbling in the dark before turning on a flashlight.
The "Blind" Approach (Conventional): The system only looks at where you are right this second. It has no idea that you are currently walking toward a wall or turning a corner.

The Solution: The "Smart Spotlight" (MAPC)

The researchers in this paper created a system called MAPC (Mobility-Aware Power Control). Think of this as a "Predictive Spotlight System" that uses a "Digital Crystal Ball."

Here is how it works using two clever layers of "brains":

Layer 1: The "Physics Brain" (Gauss-Markov Model)
This part of the brain understands the basic rules of movement. It knows that humans don't teleport; if you are walking north at 3 mph, you will likely still be moving north in the next second. It predicts your general path based on momentum.

Layer 2: The "Intuition Brain" (LSTM/Machine Learning)
Physics alone isn't enough because humans are unpredictable. We stop suddenly to check a text, or we tilt our phones to look at a map. This second layer is like a seasoned observer who has watched thousands of people move. It recognizes patterns—like the way a person’s body tilts right before they turn a corner—and "corrects" the physics brain's guess.

The Result: Efficiency Meets Reliability

By combining these two brains, the system can predict exactly where you will be and how your phone will be tilted a fraction of a second into the future.

Instead of blasting light everywhere, the system says: "I predict they are walking toward the corner and tilting their phone down. I will precisely aim a medium-strength beam at that exact spot."

The "Win-Win" Outcome:

Energy Savings: Because the system isn't "over-lighting" the room, it uses much less power (improving "Energy Efficiency").
Smooth Connection: Because it anticipates your moves, you don't experience those annoying "signal drops" when you move.

In short: The paper moves us from a world of "blasting light and hoping for the best" to a world of "smart, predictive light that follows you perfectly."

Technical Summary: Mobility-Aware Power Control for VCSEL-Based Indoor OWC

1. Problem Statement

The paper addresses the challenges of maintaining energy efficiency and link reliability in indoor Optical Wireless Communication (OWC) systems that utilize Vertical-Cavity Surface-Emitting Lasers (VCSELs).

While VCSELs enable high-speed, multi-gigabit data rates through highly directional, narrow Gaussian beams, this directionality makes the system extremely sensitive to user mobility and device orientation. Small changes in a user's position or the tilt of their device can cause significant beam misalignment, leading to rapid fluctuations in optical channel gain. Existing power control strategies typically ignore these mobility dynamics, leading to two suboptimal outcomes:

Inefficiency: Provisioning excessive power to "buffer" against potential signal drops (wasting energy).
Unreliability: Failing to anticipate signal drops, leading to Quality-of-Service (QoS) violations.

2. Methodology

The authors propose a Mobility-Aware Power Control (MAPC) framework that shifts from reactive to proactive resource management. The methodology is divided into three core stages:

A. Hybrid Mobility Modeling (GM-ANN):
To capture the complexity of human movement, the authors combine two approaches:

Gauss-Markov (GM) Model: A stochastic model used to capture the continuous, temporally correlated evolution of velocity, direction, and orientation (elevation and azimuth angles).
Long Short-Term Memory (LSTM) Network: A deep learning component that learns from sequences of past states to predict non-linear, "behavior-driven" dynamics—such as abrupt turns, sudden pauses, or device tilting—which the standard GM model cannot capture.
Result: A refined prediction of the user's future state $\hat{s}_u(t+1)$ .

B. Channel and Rate Prediction:
Using the predicted mobility states, the system estimates future channel gains ( $\hat{H}$ ) by calculating predicted distances and radial displacements. This allows the system to forecast the achievable data rate ( $\hat{R}_u$ ) for the next time slot, accounting for both signal strength and inter-AP interference.

C. Deep Learning-Based Power Allocation:
The optimization goal is to maximize Energy Efficiency (EE)—the ratio of total achievable data rate to total transmit power—subject to QoS (minimum data rate) and hardware constraints.

Since the optimization problem is non-convex and computationally heavy for real-time use, the authors use a Convolutional Neural Network (CNN).
The CNN is trained offline using optimal solutions derived from Mixed-Integer Linear Programming (MILP). During runtime, the CNN provides near-instantaneous power allocation decisions based on predicted channel states.

3. Key Contributions

Integrated Hybrid Model: The first work to jointly integrate a hybrid Gauss-Markov-Artificial Neural Network (GM-ANN) mobility model with power control specifically for VCSEL-based OWC.
Proactive Resource Management: Moving beyond "instantaneous" or "reactive" control by using predicted mobility to guide power decisions.
Real-Time Optimization via CNN: Replacing complex mathematical solvers with a trained CNN to enable low-latency, real-time power allocation in dynamic environments.

4. Results

The framework was evaluated in a simulated $5 \times 5 \times 3 \text{ m}^3$ indoor environment against three benchmarks: Conventional Power Control (CPC), Conservative Power Control (ConsPC), and Reactive Power Control (RPC).

Prediction Accuracy: The LSTM-refined model showed that while error increases with the prediction horizon, short-term predictions remain highly accurate for channel estimation.
Energy Efficiency (EE) Gains: MAPC consistently outperformed all benchmarks. It achieved average EE gains of 30–35% over CPC and ConsPC, and a massive 45–65% improvement over the reactive RPC scheme.
Robustness: As user speed increased (from 0.2 to 1.5 m/s), the EE of all systems dropped due to misalignment; however, the MAPC framework exhibited significantly higher stability and a much slower rate of degradation compared to the benchmarks.

5. Significance

This research is significant for the development of next-generation indoor wireless networks (such as 6G and high-capacity IoT). It demonstrates that for highly directional technologies like VCSEL-based OWC, mobility intelligence is not just an add-on but a requirement for achieving the dual goals of high data rates and sustainable, energy-efficient operation. The work provides a blueprint for how machine learning can bridge the gap between stochastic physical movement and deterministic network resource management.