L’ED SISMI propose le sujet de thèse suivant :

Intitulé du sujet : Deep reinforcement learning for dynamic deployment and resource management of drones-based 5g network in disaster areas

Ce projet serait sous la direction de Abbas BRADAI de XLIM à l’Université de Poitiers

Co-directeurs renseignés : Emmanuel MOULAY /

Les financements sont : Joint PHD program – university of Chicago

Le début de la thèse est prévu pour : 10/2022

Mots clés du sujet : resources allocation, QoS, Drones, Deep Reinforcement Learning, control

Présentation du sujet : In this thesis, we aim to propose new solutions to optimally and dynamically deploy a fleet of drones that will serve as a 5G mobile communication network. Each drone will carry communications resources, telecommunications antenna as well as computing and storage resources, and will serve as a mini mobile base station. In this context, in this project we will jointly optimize the drones’ positions, the communications between them as well as the allocation of the scarce resources (computing and communication) of such network. Moreover, we consider different QoS classes, and we differentiate them by allocating dedicated networking and computing resources for each class (slice).

Objectifs : In this thesis, we aim to propose new solutions to optimally and dynamically deploy a fleet of drones that will serve as a 5G mobile communication network. Each drone will carry communications resources, telecommunications antenna as well as computing and storage resources, and will serve as a mini mobile base station. In this context, in this project we will jointly optimize the drones’ positions, the communications between them as well as the allocation of the scarce resources (computing and communication) of such network. Moreover, we consider different QoS classes, and we differentiate them by allocating dedicated networking and computing resources for each class (slice).

Description du sujet : In this project, we aim to propose new solutions to optimally and dynamically deploy a fleet of drones that will serve as a 5G mobile communication network. Each drone will carry communications resources, telecommunications antenna as well as computing and storage resources, and will serve as a mini mobile base station. In this context, in this project we will jointly optimize the drones’ positions, the communications between them as well as the allocation of the scarce resources (computing and communication) of such network. Moreover, we consider different QoS classes, and we differentiate them by allocating dedicated networking and computing resources for each class (slice). Hence, inter-slices and intra-slices resource allocation will be considered in this project. The aim of this project is therefore to propose new strategies for the dynamic deployment of drones, network slicing and resources allocation in disaster areas with a new generation of drones-based 5G telecommunications networks. In such complex and dynamic system, we propose solutions based on Deep Reinforcement Learning (DRL) techniques.

Compétences acquises à l’issue de la thèse : – DRL for networking and telecommunication
– Control & communication for drone based networks

Présentation de l’équipe d’accueil : The PhD. student will spend half of thesis duration at university of Poitiers, RUBIH team and the other half at University of Chicago, Prof. Jiang’s research group. The RUBIH team is part of the SRI axis (Intelligent Systems and Networks). Its main research area is the study, control and optimization of wireless telecommunications networks, through a “system approach”, to meet the required QoS as well as the energy efficiency of the network. Dr. Abbas BRADAI, head of the RUBIH team, is expert on 5G networks and DRL applied to networking. Dr. Emmanuel MOULAY is expert on control theory and multi-agent systems.
Prof Jiang’s research group is part of Department of Computer Science. It is identified with a series of work in the intersection between machine learning and networked systems, including the first realistic simulation of DRL (Sim2real), the first practical data-driven control plane for Internet-scale video streaming and the first content-adaptive video analytics system at scale.

Compétences souhaitées pour les candidats : – Wireless networks
– 5G
– Network simulator
– Deep learning notions

Pour plus d’informations et pour candidater, merci de contacter :

Date de dépôt : 28/02/2022 à 11 h 22 min