The main themes of the Réseaux UBiquitaires adaptatIf Hauts débits (RUBIH) team concern the optimization of resource allocation in large-scale networks such as the Internet of Things (IoT) and 5G (and beyond), as well as the transmission of massive data of different types and different Quality of Service (QoS). Indeed, the coexistence of large-scale sensor networks with traditional high-speed cellular networks inherited from the past poses new challenges in terms of managing overall network QoS in the context of smart cities, industry 4.0 or Intelligent Transportation System (ITS). Node mobility issues, which occur at several network levels, are also the subject of special developments. The RUBIH team is heavily involved in resource allocation issues, which require the use of complex optimization tools ranging from automatic control, signal processing and game theory to Artificial Intelligence (AI) techniques.

Research topics

• Resource allocation in 5G networks and beyond and QoS management in the IoT - The RUBIH team works on resource allocation strategies in wireless telecommunications networks. The aim is to intelligently allocate resources, in frequency and power for example, to the various nodes of the 5G network and beyond, while guaranteeing QoS to the various users. The RUBIH team proposes to exploit the spatial, frequency and temporal diversities of the radio channel to implement link adaptation strategies (UEP and UPA techniques, for example) and scheduling strategies enabling the most optimal and dynamic interaction possible between the PHY, MAC and Application layers for image or video transmission. In this respect, associations between the MIMO concept and orthogonal (OFDMA) or non-orthogonal (NOMA) access techniques such as Power Division Multiple Access (PDMA) or Sparse Code Multiple Access (SCMA) are relevant solutions in this cross-layer optimization strategy responding to multi-user QoS. Likewise, with the aim of improving spectrum management and energy efficiency in the context of the massive deployment of IoT networks comprising tens or even hundreds of thousands of sensors, for example, the second stage consists of implementing new centralized/decentralized approaches to virtual network slicing, separating the control plane from the data plane with Software Defined Networking (SDN), and decentralized deployment of Multi-access Edge Computing (MEC) using multi-agent systems to guarantee heterogeneous QoS requirements for multiple users.

• AI and Compressive Sensing techniques applied to heterogeneous networks - The RUBIH team is also working on more demanding heterogeneous wireless networks with strong interference and QoS constraints, for example in industrial environments (Industry 4.0) or for ITS, based on machine learning techniques such as Deep Learning and Deep Reinforcement Learning (DRL). In this context, it focuses on Virtual Network Functions (VNF) and mobility management in 5G cellular and vehicular networks and beyond. For multimedia wireless sensor networks, the RUBIH team is looking to propose innovative network deployment solutions for reliable, low-complexity uplink image or video transfer. To achieve this, link adaptation strategies integrate Compressive Sensing schemes with low complexity learning techniques to reduce the acquisition rate of multimedia sensors, e.g. for video surveillance or transport. Indeed, since the data captured is very often correlated, it is possible to take advantage of this to reduce the number of transmissions. In addition to Compressive Sensing techniques, Matrix Completion techniques are used to obtain the missing data. These include techniques such as clustering and Principal Component Analysis (PCA).

• Massive MIMO and improved spectral efficiency - The RUBIH team is also studying the use of a very large number of Massive MIMO antennas in reception, transmission and relay-reflection, to increase transmission channel capacity, particularly useful in densely populated areas, for example in smart cities. Hybrid beamformer techniques have been developed to reduce radio complexity. One area of active research concerns Reconfigurable Intelligent Surfaces (RIS) to improve cellular network coverage. High mobility is another challenge, generating channels that are doubly selective in time and frequency. The RUBIH team is working on OTFS modulation and the problems associated with its channel identification. It uses machine learning techniques to reduce implementation complexity.

Partnerships

• Airbus - Amirkabir University of Technology, Iran - Audensiel – Caplogy – CEA – CISTEME - IMT Atlantique - LaBRI, Université de Bordeaux - LAMIH, Université Polytechnique Hauts-de-France - LEAT, Université Côte d’Azur - LIMOS, Université Clermont Auvergne - LS2N, École Centrale de Nantes - L3i, Université de La Rochelle - Orange Lab - Polytechnique Montréal, Canada - Sup'Com, Tunisie - Transpod - Université Gustave Eiffel - University of Nebraska–Lincoln, USA

National and international research programs

• European Programs : H2020-ICT-2016-2017 SmartMet (2017-2020), ERA-NET CHIST-ERA SAMBAS (2020-2025)
• National Programs : ANR-20-CE25-0004 MOMENT (2020-2024), ANR-21-CE25-0005 SAFE (2022-2026), FITNESS PEPR Réseaux du Futur (2023-2028), PERSEUS PEPR Réseaux du Futur (2023-2028)