L2S | Multimedia and Networking (MULTINET) | CNRS - CentraleSupelec

The research activities of the Multimedia and Networks (MULTINET) team cover a broad spectrum of topics, ranging from networking aspects to multimedia applications. In the context of today’s major challenges (including the Internet of Things, mobile networks beyond 5G, smart cities, Industry 4.0, energy efficiency, immersive video, real-time video transport, and vehicular networks) our work is structured around two main research areas: the study of networking issues and the improvement of video services. In various contexts, such as wireless sensor networks and mobile networks beyond 5G, we model and optimise network-related problems and propose novel approaches in the form of algorithms and protocols to address challenges including routing, information collection and dissemination, mobility management, energy efficiency, resource allocation, and multi-domain orchestration. The improvement of video services involves not only enhancing the performance of standard video codecs, but also designing new emerging formats. To provide users with a better quality of experience, we develop novel subjective assessment protocols for visual quality, as well as new objective methods for predicting this quality.

Networks

Our research on networks aims to meet the growing requirements for agility and performance in future networks. It is fully aligned with the development of future generations of mobile networks as well as the evolution of Wi-Fi networks.

Wireless Network Optimization: We investigate the new uses and requirements of wireless networks in contexts such as the Internet of Things, smart cities, and Industry 4.0. We address a variety of challenges, including communications in white areas, dense network optimization, and the exploitation of new physical-layer functionalities and their impact on the control plane, particularly in the context of the Internet of Things.
Virtualization, Network Slicing, and SDN: We also investigate network slicing, with a particular focus on the allocation of shared hardware resources among different virtual networks with quality-of-service constraints and elastic traffic demands. The objective is to design demand models and control algorithms that optimize the use of physical resources by virtual network resources.
Energy Efficiency: We investigate promising energy-saving techniques for radio access networks, including symbol muting, channel shutdown, carrier deactivation, and advanced sleep modes that allow network components to enter low-power states during periods of inactivity. Integrating these approaches contributes to the design of more sustainable and energy-efficient wireless communication infrastructures for next-generation mobile networks.
QoS for Critical Applications: Communication networks are increasingly used to support critical applications, such as autonomous vehicles, Industry 4.0, and e-health. In this context, new protocols need to be designed for wireless ad hoc networks, including sensor networks and vehicular networks, to meet stringent quality-of-service requirements. Such guarantees are essential for ensuring the reliable operation of critical applications.

Video coding and transmission

With the proliferation of video services, users are becoming more and more demanding regarding the quality of experience. In this context, new breakthrough technologies are emerging in order to provide a much more realistic and immersive rendering.

Immersive video: New breakthrough technologies are emerging to provide a much more realistic and immersive rendering, in order to create a feeling of physical presence in the scene. The so-called “light field” representations record all the light rays entering a camera. Alternatively, volumetric approaches such as point clouds store the 3D coordinates of objects and surfaces as well as their associated color. These representations require effective compression, as well as the development of new quality indicators to assess their performance.
Video coding based on deep learning techniques: Another important trend is the emergence of deep learning methods for video compression. These approaches make it possible to reflect on video coding architectures in a complete departure with existing standard schemes. For this purpose, we intend to use auto-encoders or generative adversarial networks (GAN) to optimise certain coding tools, such as spatial and temporal prediction, transform or entropy coding. In the longer term, it is also interesting to study fully optimised end-to-end video coding models from a large data collection.
Low latency linear video coding: In the context of a very low latency constraint, relevant for applications such as vehicle remote control or augmented reality, an alternative to conventional approaches is the use of linear video coders where all the operations carried out are linear. Therefore, the quality of the decoded video is proportional to the quality of the communication channel.

Head

Sahar HOTEIT

Associate Professor – Université Paris-Saclay

Télécoms et réseaux – MULTINET

sahar.hoteit@l2s.centralesupelec.fr

Bât. Breguet .