7th International Conference on Machine Learning for Networking (MLN'2024)
Keynote: Gamal Elghazaly, Research Fellow at Université du Luxembourg
Title: From Capturing Reality to Regenerating It: Transforming Autonomous Driving Simulation and Testing
Abstract
In the shift toward end-to-end AI models in autonomous driving, scalable, domain-gap-free simulation environments are becoming critical. Traditional simulations struggle with achieving realistic dynamics and scale, limiting their effectiveness for autonomous vehicle training and evaluation. This keynote explores how advancements in Neural Radiance Fields (NeRFs) and Gaussian Splatting are addressing these challenges, creating high-fidelity, data-driven simulations that capture and regenerate real-world scenes with unprecedented detail. By leveraging dynamic scene graph representations, these techniques not only enhance the realism of urban environments—including dynamic actors like vehicles, pedestrians and cyclists—but also support versatile scene editing and customization, enabling lifelike, adaptable training grounds for scalable autonomous systems. This presentation outlines frameworks for generating customizable, photorealistic scenes real-world data, setting new standards in autonomous driving simulation accuracy and scalability. The presentation will also shed some light on the potential of these methods to bridge the gap between virtual and real-world testing, offering new avenues for model training and validation in diverse, ever-evolving urban landscapes.
Biography
Dr. Gamal Elghazaly is a Research Fellow at the Interdisciplinary Centre for Security, Reliability and Trust (SnT) of the University of Luxembourg, specializing in connected and automated mobility. His research focuses mainly on enhancing the capabilities of automated vehicles to safely navigate and interact with complex environments through robotics, computer vision, and AI. With over seven years of industry experience, Dr. Elghazaly held pivotal roles at Milla Group, a leading French autonomous shuttle manufacturer, where he served as both Senior R&D Engineer and Chief Scientific Officer. In these positions, he led software development, R&D initiatives, and defined strategic research directions, contributing to pioneering autonomous driving projects with industry giants like Renault and Valeo. Dr. Elghazaly currently co-heads the 360Lab at SnT, advancing automated driving research, and manages projects in scalable autonomous driving. He received his Ph.D. in Robotics from the University of Montpellier, France in 2017, specializing in motion planning and control, and completed a Master’s in Robotics in 2013 from Ecole Centrale de Nantes, France and the University of Genova, Italy. Additionally, he actively contributes to Luxembourg’s smart mobility initiatives and reviews for top conferences and journals in robotics, computer vision, AI and intelligent transportation.