pr\u00e9sentent<\/p>\n\n\n\n
Autoris\u00e9 \u00e0 pr\u00e9senter ses travaux en vue de l\u2019obtention du Doctorat de l’Institut Polytechnique de Paris, pr\u00e9par\u00e9 \u00e0 T\u00e9l\u00e9com SudParis en :<\/p>\n\n\n\n
le JEUDI 27 NOVEMBRE 2025 \u00e0 14h00<\/p>\n\n\n\n
\u00e0<\/p>\n\n\n\n
Amphith\u00e9\u00e2tre Rose Dieng
19 Place Marguerite Perey 91120 Palaiseau<\/p>\n\n\n\n
Membres du jury :<\/strong><\/p>\n\n\n\n M. Tijani CHAHED<\/strong>, Professeur, T\u00e9l\u00e9com SudParis, FRANCE – Directeur de these Invit\u00e9 : <\/strong><\/p>\n\n\n\n M. Abdel LISSER<\/strong>, Professeur, L2S CentraleSupelec, Paris-Saclay University, FRANCE, co-encadrant<\/p>\n\n\n\n R\u00e9sum\u00e9 :<\/strong><\/p>\n\n\n\n Dans cette th\u00e8se, nous abordons le d\u00e9fi de l\u2019allocation efficace des ressources sous incertitude pour le transport de trafic critique en temps ultra fiable dans les r\u00e9seaux de nouvelle g\u00e9n\u00e9ration. Nous d\u00e9veloppons des m\u00e9thodes d\u2019optimisation et d\u2019apprentissage en ligne qui fournissent des garanties de performance rigoureuses pour des exigences probabilistes \u00e0 court terme et des contraintes cumulatives \u00e0 long terme. Nous commen\u00e7ons avec les communications ultra fiables et \u00e0 faible latence (URLLC). Les mod\u00e8les ant\u00e9rieurs pour le d\u00e9lai probabiliste imposent soit de fortes hypoth\u00e8ses sur les arriv\u00e9es, soit se concentrent principalement sur la stabilit\u00e9 des files d\u2019attente. Nous assouplissons ces hypoth\u00e8ses et formulons une minimisation de l\u2019utilisation des ressources sous contraintes de probabilit\u00e9, valable pour des processus d\u2019arriv\u00e9es g\u00e9n\u00e9raux. En exploitant les propri\u00e9t\u00e9s structurelles de l\u2019espace des politiques, nous concevons des algorithmes efficaces de type bandit pour les contextes hors ligne (statistiques connues) et en ligne (statistiques inconnues). Ces algorithmes convergent de mani\u00e8re d\u00e9montrable en un nombre fixe d\u2019it\u00e9rations tout en respectant des objectifs stricts de d\u00e9lai de 1 ms et de fiabilit\u00e9 de (10^{-5}), avec une consommation minimale de ressources. Nous poussons ensuite ces garanties vers l\u2019URLLC extr\u00eame visant une latence de (0.1,mathrm{ms}) et une fiabilit\u00e9 de l\u2019ordre de (10^{-7}), o\u00f9 la mise en file d\u2019attente est interdite et o\u00f9 les sch\u00e9mas d\u2019allocation de ressources doivent s\u2019appuyer sur une information limit\u00e9e des arriv\u00e9es (\u00e9chantillons historiques, moyenne, variance). Les m\u00e9thodes statiques ont tendance \u00e0 sur-allouer les ressources. Nous introduisons une politique de r\u00e9servation dynamique en ligne bas\u00e9e sur une approche par sc\u00e9nario \u00e0 fen\u00eatre glissante, qui est robuste et s\u00fbre : elle suit les r\u00e9servations minimales \u00e0 partir de donn\u00e9es empiriques et \u00e9vite la sur-provision conservatrice tout en pr\u00e9servant des contraintes de QoS strictes. Nous consid\u00e9rons ensuite les communications orient\u00e9es objectifs, en nous concentrant sur les applications haptiques tr\u00e8s sensibles aux rafales de pertes de paquets. Nous proposons un cadre th\u00e9orique de files d\u2019attente qui minimise les co\u00fbts en ressources en pr\u00e9sence de pertes dues \u00e0 la fois aux collisions avec d\u2019autres paquets haptiques et aux mauvaises conditions radio. Nous concevons une politique de contr\u00f4le conjointe combinant un renforcement adaptatif de la puissance de transmission avec la pr\u00e9emption de ressources initialement provisionn\u00e9es pour le haut d\u00e9bit mobile am\u00e9lior\u00e9 (eMBB), r\u00e9gie par des politiques \u00e0 seuil. Pour des utilisateurs h\u00e9t\u00e9rog\u00e8nes, l\u2019interd\u00e9pendance entre groupes d\u2019utilisateurs induit un espace d\u00e9cisionnel de haute dimension, rendant la recherche exhaustive irr\u00e9alisable. Pour traiter cette complexit\u00e9, nous utilisons un algorithme de recuit simul\u00e9 modifi\u00e9 avec gestion des contraintes par rejet direct des politiques non r\u00e9alisables ou par p\u00e9nalit\u00e9s bas\u00e9es sur les co\u00fbts. Enfin, nous \u00e9tudions la conformit\u00e9 \u00e0 long terme et introduisons l\u2019Optimisation convexe en ligne contrainte avec m\u00e9moire (COCO-M), o\u00f9 les pertes et les contraintes d\u00e9pendent des (m) derni\u00e8res d\u00e9cisions. Les travaux ant\u00e9rieurs consid\u00e9raient principalement une longueur de m\u00e9moire fixe. Nous g\u00e9n\u00e9ralisons \u00e0 des longueurs de m\u00e9moire arbitraires et int\u00e9grons des pr\u00e9dictions non fiables \u00e0 court horizon, en fournissant les premiers algorithmes avec des bornes de regret sous-lin\u00e9aire et de violation cumulative de contraintes sous-lin\u00e9aire dans ce cadre g\u00e9n\u00e9ral. Cela offre une bo\u00eete \u00e0 outils polyvalente pour l\u2019apprentissage en ligne et le contr\u00f4le pr\u00e9dictif de r\u00e9seau en conditions adverses. We address in this thesis the challenge of efficient resource allocation under uncertainty for the transport of time-critical ultra reliable traffic in next-generation networks. We develop optimization and online-learning methods that provide rigorous performance guarantees for short-horizon probabilistic requirements and long-term cumulative constraints. We begin with Ultra-Reliable Low-Latency Communications (URLLC). Prior models for probabilistic delay either impose strong assumptions on arrivals or focus primarily on queue stability. We relax these assumptions and formulate a chance-constrained minimization of resource usage that holds under general arrival processes. By exploiting structural properties of the policy space, we design efficient bandit-based algorithms for both offline (known statistics) and online (unknown statistics) settings. These algorithms provably converge in a fixed number of iterations while meeting stringent 1ms delay and (10^{-5}) reliability targets with minimal resource consumption. We then push these guarantees to extreme URLLC targeting (0.1,mathrm{ms}) latency and reliability on the order of (10^{-7}), where queuing is impermissible and the resource allocation schemes must rely on limited arrival information (historical samples, mean, variance). Static methods tend to over-allocate resources. We introduce an online, dynamic reservation policy based on a sliding-window scenario approach that is robust and safe: it tracks minimal reservations from empirical data and avoids conservative over-provisioning while preserving stringent QoS constraints. Next, we consider goal-oriented communications, focusing on haptic applications that are highly sensitive to bursts of packet losses. We propose a queuing-theoretic framework which minimizes resource costs in the presence of losses from both collisions with other haptic packets and poor radio conditions. We design a joint control policy that combines adaptive transmit-power boosting with preemption of resources initially provisioned for enhanced Mobile Broadband (eMBB), governed by threshold policies. For heterogeneous users, interdependence across user groups induces a high-dimensional decision space, ruling out exhaustive search. To address this complexity, we make use of a modified simulated-annealing algorithm with constraint handling through direct rejection of infeasible policies or cost-based penalties. Eventually, we study long-term compliance and introduce Constrained Online Convex Optimization with Memory (COCO-M), where losses and constraints depend on the last (m) decisions. Prior work considered mainly fixed memory length. We generalize to arbitrary memory lengths and incorporate untrusted short-horizon predictions, providing the first algorithms with provable sublinear regret and sublinear cumulative constraint violation in this general setting. This yields a versatile toolbox for online learning and predictive network control under adversarial conditions.<\/p>\n","protected":false},"excerpt":{"rendered":" L’Ecole doctorale : Ecole Doctorale de l’Institut Polytechnique de Paris et le Laboratoire de recherche SAMOVAR – Services r\u00e9partis, Architectures, Mod\u00e9lisation, Validation, Administration des R\u00e9seaux pr\u00e9sentent l\u2019AVIS DE SOUTENANCE de Monsieur Mohammed ABDULLAH Autoris\u00e9 \u00e0 pr\u00e9senter ses travaux en vue de l\u2019obtention du Doctorat de l’Institut Polytechnique de Paris, pr\u00e9par\u00e9 \u00e0 T\u00e9l\u00e9com SudParis en : […]<\/p>\n","protected":false},"author":4,"featured_media":0,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ocean_post_layout":"","ocean_both_sidebars_style":"","ocean_both_sidebars_content_width":0,"ocean_both_sidebars_sidebars_width":0,"ocean_sidebar":"","ocean_second_sidebar":"","ocean_disable_margins":"enable","ocean_add_body_class":"","ocean_shortcode_before_top_bar":"","ocean_shortcode_after_top_bar":"","ocean_shortcode_before_header":"","ocean_shortcode_after_header":"","ocean_has_shortcode":"","ocean_shortcode_after_title":"","ocean_shortcode_before_footer_widgets":"","ocean_shortcode_after_footer_widgets":"","ocean_shortcode_before_footer_bottom":"","ocean_shortcode_after_footer_bottom":"","ocean_display_top_bar":"default","ocean_display_header":"default","ocean_header_style":"","ocean_center_header_left_menu":"","ocean_custom_header_template":"","ocean_custom_logo":0,"ocean_custom_retina_logo":0,"ocean_custom_logo_max_width":0,"ocean_custom_logo_tablet_max_width":0,"ocean_custom_logo_mobile_max_width":0,"ocean_custom_logo_max_height":0,"ocean_custom_logo_tablet_max_height":0,"ocean_custom_logo_mobile_max_height":0,"ocean_header_custom_menu":"","ocean_menu_typo_font_family":"","ocean_menu_typo_font_subset":"","ocean_menu_typo_font_size":0,"ocean_menu_typo_font_size_tablet":0,"ocean_menu_typo_font_size_mobile":0,"ocean_menu_typo_font_size_unit":"px","ocean_menu_typo_font_weight":"","ocean_menu_typo_font_weight_tablet":"","ocean_menu_typo_font_weight_mobile":"","ocean_menu_typo_transform":"","ocean_menu_typo_transform_tablet":"","ocean_menu_typo_transform_mobile":"","ocean_menu_typo_line_height":0,"ocean_menu_typo_line_height_tablet":0,"ocean_menu_typo_line_height_mobile":0,"ocean_menu_typo_line_height_unit":"","ocean_menu_typo_spacing":0,"ocean_menu_typo_spacing_tablet":0,"ocean_menu_typo_spacing_mobile":0,"ocean_menu_typo_spacing_unit":"","ocean_menu_link_color":"","ocean_menu_link_color_hover":"","ocean_menu_link_color_active":"","ocean_menu_link_background":"","ocean_menu_link_hover_background":"","ocean_menu_link_active_background":"","ocean_menu_social_links_bg":"","ocean_menu_social_hover_links_bg":"","ocean_menu_social_links_color":"","ocean_menu_social_hover_links_color":"","ocean_disable_title":"default","ocean_disable_heading":"default","ocean_post_title":"","ocean_post_subheading":"","ocean_post_title_style":"","ocean_post_title_background_color":"","ocean_post_title_background":0,"ocean_post_title_bg_image_position":"","ocean_post_title_bg_image_attachment":"","ocean_post_title_bg_image_repeat":"","ocean_post_title_bg_image_size":"","ocean_post_title_height":0,"ocean_post_title_bg_overlay":0.5,"ocean_post_title_bg_overlay_color":"","ocean_disable_breadcrumbs":"default","ocean_breadcrumbs_color":"","ocean_breadcrumbs_separator_color":"","ocean_breadcrumbs_links_color":"","ocean_breadcrumbs_links_hover_color":"","ocean_display_footer_widgets":"default","ocean_display_footer_bottom":"default","ocean_custom_footer_template":"","ocean_post_oembed":"","ocean_post_self_hosted_media":"","ocean_post_video_embed":"","ocean_link_format":"","ocean_link_format_target":"self","ocean_quote_format":"","ocean_quote_format_link":"post","ocean_gallery_link_images":"on","ocean_gallery_id":[],"footnotes":""},"categories":[286,402],"tags":[],"class_list":["post-6982","post","type-post","status-publish","format-standard","hentry","category-fractualites-ennews-fr","category-seminaires-ness-2013-fr","entry"],"_links":{"self":[{"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/posts\/6982","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/users\/4"}],"replies":[{"embeddable":true,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/comments?post=6982"}],"version-history":[{"count":1,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/posts\/6982\/revisions"}],"predecessor-version":[{"id":6983,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/posts\/6982\/revisions\/6983"}],"wp:attachment":[{"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/media?parent=6982"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/categories?post=6982"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/tags?post=6982"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
M. Salah Eddine EL AYOUBI<\/strong>, Professeur, L2S CentraleSupelec, Paris-Saclay University , FRANCE – CoDirecteur de these
Mme Ana BU\u0161I\u0107<\/strong>, Charg\u00e9e de recherche, INRIA – Ecole Normale Sup\u00e9rieure, Paris, France, FRANCE – Examinateur
M. Michel KIEFFER<\/strong>, Professeur, L2S CentraleSupelec, Paris-Saclay University , FRANCE – Examinateur
M. Stefano SECCI<\/strong>, Professeur des universit\u00e9s, National Conservatory of Arts and Crafts (CNAM), Computer Science Department, FRANCE – Rapporteur
M. Georges KADDOUM<\/strong>, Professeur, \u00c9cole de technologie sup\u00e9rieure (\u00c9TS), Universit\u00e9 du Qu\u00e9bec, CANADA – Rapporteur<\/p>\n\n\n\n\u00ab Optimisation de l\u2019allocation des ressources dans les r\u00e9seaux sans fil du futur en pr\u00e9sence d\u2019incertitude \u00bb<\/h2>\n\n\n\n
pr\u00e9sent\u00e9 par Monsieur Mohammed ABDULLAH<\/h2>\n\n\n\n
Abstract :<\/strong><\/p>\n\n\n\n