Optical researcher seeks not just a speedy network, but a reliable one
In the realm of optical networks, speed has long been the driving force.
Our global communications infrastructure, powered by high-capacity fiber optic cables, has thrived on the principle that faster is better. Nokia Bell Labs has been at the forefront of this pursuit, pushing the boundaries of optical speed and achieving world-record milestones.
But the landscape has shifted. In the post-COVID era, where seamless connectivity fuels nearly every aspect of modern life — information, education, healthcare, and beyond — speed is no longer the sole priority.
Reliability is now just as essential. In a world that demands uninterrupted connections, we can no longer afford costly disruptions. The new frontier is not just about making networks faster, but about ensuring they are resilient, dependable and prepared for the unexpected.
The best way to avoid breakdowns is to predict them before they happen. By foreseeing potential failures, the networks that power our hyper-connected world can remain strong, stable, and ready to meet the growing demands of tomorrow.
That’s where optical network researcher Khouloud Abdelli steps in. Her mission is to detect potential disruptions before they can impact service.
Abdelli’s approach goes beyond the fiber optics themselves — advanced cables capable of sensing their environment while transmitting data. Her expertise is using cutting-edge artificial intelligence (AI) tools to decode this vast stream of data to predict the threats before they become problems.
“My work focuses on developing advanced machine learning models to tackle complex problems by extracting insights from network monitoring data,” she explained. “The heart of my research is applying these insights to predictive strategies that significantly enhance network reliability. The goal is to foresee failures before they happen and prevent disruptions.”
Abdelli will be presenting the findings of her research next week (Sept. 22-26) in Frankfurt, Germany at ECOC 2024, the 50th European Conference on Optical Communications, which is one of the most prestigious gatherings of the optical research community.
Predicting to future to protect the network
In today’s information-driven world, data is often hailed as the new oil — an invaluable resource, but only if it is effectively extracted and refined. This analogy resonates deeply with Abdelli, especially in her world of optical networks, where data serves as the lifeblood for cutting-edge machine learning solutions. These innovations drive critical advancements in anomaly detection, system optimization, and, ultimately, the overall reliability of the network.
“Think of it like a child who learns about the world by interacting with its surroundings. Similarly, a machine learning model learns and improves by analyzing the data it is given,” she explained. “Just as a child’s understanding depends on the richness and diversity of their experiences, the effectiveness of a machine learning model depends on the quality and relevance of the data it processes.”
Once her colleagues harness fiber optics as sensors to gather critical network data, Abdelli steps in to train machine learning models that extract the patterns of system behavior. These finely tuned models then detect and pinpoint anomalies by comparing the system’s predicted performance with its real-time state. Any deviations from the norm are immediately flagged, enabling rapid corrective action.
“The impact is profound. By boosting the reliability of optical networks, we elevate global connectivity, leading to a vastly improved customer experience,” she said. “When you can accurately pinpoint potential issues, you not only minimize downtime but also significantly reduce maintenance costs, ensuring smoother, more efficient operations.”
At ECOC 2024, she and her colleagues will be presenting papers that highlight her work across three research areas.
The first, titled “Unsupervised Anomaly Detection and Localization with Generative Adversarial Networks,” introduces an innovative method to overcome the difficulty of detecting rare and subtle anomalies in optical networks, where data is often limited and imbalanced.
The second, titled “Multi-Step Forecasting of State of Polarization Changes in Aerial Fibers Using Wavelet Neural Networks,” proposes a novel weather-adaptive strategy for forecasting variations in aerial fiber links.
The third, titled “Threat Classification on Deployed Optical Networks Using MIMO Digital Fiber Sensing, Wavelets, and Machine Learning,” is based on a field test in Saudi Arabia that successfully classified mechanical threats to optical networks, specifically focusing on disturbances caused by jackhammers and excavators. It paves the way for improving the stability of future optical networks to enable proactive rerouting and break avoidance.
A Life of Parallel Pursuits
Growing up in Tunisia, Abdelli always dreamed of becoming a researcher. When the time came to choose a career, she opted for engineering because of that passion for technological exploration.
“Engineers shape the future by solving problems for society,” she said. “For me, it is about inventing, solving complex problems and sparking new ideas.”
After earning a diploma in telecommunications from the Higher School of Communications of Tunis (SUP’COM), University of Carthage, she moved to Munich, Germany, to become a telecom consultant for Telefonica. But she quickly realized that her true passion lay in research. So, she enrolled in a PhD in Electrical Engineering and Information Technology at Kiel University and began working as a researcher at ADVA, an optical networking vendor now part of ADTRAN, where she conducted her PhD research and specialized in predictive maintenance, optical fiber fault detection and the lifetime prediction of network components.
Her doctorate dissertation in machine learning-based predictive maintenance for optical networks, led to her current role as optical network researcher at Nokia Bell Labs in Stuttgart.
Outside her professional work, Abdelli’s life is devoted almost exclusively to the pursuit of Islamic knowledge. She dedicates most of her free time to reading books, listening to scholarly lectures and memorizing the Quran.
Abdelli described her spiritual journey as one that “nourishes the soul” and gives her life meaning.
“Just as the body requires food and drink to thrive, so too does the soul need the remembrance of the Creator,” she explained. “This remembrance is as vital to the heart as water is to fish. Consider what would happen if a fish were deprived of water.”
Abdelli acknowledged the challenges her Muslim faith can present in a professional environment but said she generally managed to align her work practices with her beliefs by working remotely. In line with her values, Abdelli chose not to provide photos for herself for this profile, citing a desire for modesty.
For the most part, though, she said her devout religious lifestyle lived comfortably alongside, rather than in opposition to, her intellectual pursuits.
“This dual commitment creates a unique synergy in my life. The discipline and analytical thinking required in machine learning research compliment the reflective and contemplative nature of religious study,” she said. “Just as I strive to uncover patterns and insights within data, I also seek to extract and comprehend the profound messages within the Quran. It leads to a holistic understanding that bridges the world of science with the realm of faith.”