Search

International audience

Preprint submitted to EJOR; We consider a telecommunications company expanding its network capacity to face an increasing demand. The company can also invest in marketing to incentivize clients to shift to more recent technologies, hopefully leading to cheaper overall costs. To model the effect of marketing campaigns, previous works have relied on the Bass model. Since that model only provides a rough approximation of the actual shifting mechanism, the purpose of this work is to consider uncertainty in the shifting mechanism through the lens of robust optimization. We thus assume that the (discrete) shifting function can take any value in a given polytope and wish to optimize against the worst-case realization. The resulting robust optimization problem possesses integer recourse variables and non-linear dependencies on the uncertain parameters. We address these difficulties as follows. First, the integer recourse is tackled heuristically through a piece-wise constant policy dictated by a prior partition of the uncertainty polytope. Second, the non-linearities are handled by a careful analysis of the dominating scenarios. The scalability and economical relevance of our models are assessed through numerical experiments performed on realistic instances. In particular, we choose one of these instances to perform a case study with simulations illustrating the possible benefit of using robust optimization.

International audience; Full Duplex (FD) in cellular networks is expected to increase the cell spectral efficiency. However, while the downlink (DL) spectral efficiency (SE) increases with FD, the uplink (UL) SE decreases because of the Base Station to Base Station (BS) interference. In this paper, assuming a three-node model, we propose a method based on Successive Interference Cancellation (SIC) to reduce the BS-to-BS interference present in FD cellular networks. The approach consists in coordinating BSs to enable the decoding and the suppression of undesired signals that impair uplink transmissions. We analyze both distributed and Centralized Radio Access Networks (CRAN) architectures. Stochastic geometry is used to derive the coverage probability and mean data rate of the proposed scheme. In the distributed scenario, the FD UL average data rate is increased by 25% with our solution compared to a classical FD network, while our FD scheme still outperforms Half-Duplex (HD) on the DL. In the centralized scenario, our solution outperforms HD by 10% and classical FD by 78% on the UL, while preserving classical FD gains on the DL.

International audience; The use of batteries as a backup in case of power outages is a common practice adopted by telecommunications companies that need to keep their services always active. Besides, those batteries can also be used for other purposes such as participating in the energy market in order to reduce the electricity bill, as long as the safety usage rules are respected. In this context, batteries can be used when the energy costs more and recharged when the energy costs less, which is known as the demand-response mechanism. Our focus in this work is to optimize the use of batteries installed for backup to participate in the demand-response mechanism, in order to reduce the total energy cost for the company. We formally state the related optimization problem and propose two solving approaches to address it: a mixed-integer program and a heuristic to solve large instances. Simulations based on real data of a French telecommunications operator prove the relevance of using batteries to reduce the energy cost for the company by participating in the demandresponse mechanism. The proposed heuristic proves to be economically relevant and computationally efficient, being a good alternative to a mixed-integer program for large-scale problems.

International audience; In this paper, we consider the problem of learning a correlated equilibrium of a finite non-cooperative game and show a new adaptive heuristic, called Correlated Perturbed Regret Minimization (CPRM) for this purpose. CPRM combines regret minimization to approach the set of correlated equilibria and a simple device suggesting actions to the players to further stabilize the dynamic. Numerical experiments support the hypothesis of the pointwise convergence of the empirical distribution over action profiles to an approximate correlated equilibrium with all players following the devices' suggestions. Additional simulation results suggest that CPRM is adaptive to changes in the game such as departures or arrivals of players.

International audience; Licensed Shared Access (LSA) is a new concept proposed by the radio spectrum policy group in order to optimize spectrum usage: a Mobile Network Operator (MNO) can access temporarily to other incumbent's spectrum after obtaining a license. The licensing process is made via an auction mechanism. The mechanisms proposed in the literature for the LSA context are oneshot auction mechanisms which allocate all the available spectrum as a unique block. In this paper, we first show how to increase the performance of those auctions (in terms of revenue, efficiency and fairness of the allocation) while preserving truthful bidding, by splitting spectrum and converting single block auctions into multi-block auctions. Simulation results illustrate how appropriately choosing the number of blocks allows to increase the aforementioned metrics. Second, we show how to convert one-shot mechanisms to equivalent ascending mechanisms (in terms of allocations and payments) so that we add transparency and privacy to the auction.

International audience; The integration of RIS elements in upcoming 6G networks appears a major breakthrough to extend network coverage and capacity. This paper introduces a new link-layer scheme for RIS-enabled communications building on existing models for the physical layer of RIS technologies. The scheme is able to integrate the selection of precoders/beams in a codebook and scheduling of UEs at once. Furthermore, elements for the integration of the proposed scheme in current 3GPP-5G specifications are addressed. The scheduler combines a slow mechanism operating at the downlink OFDMA frame scale with a standard proportional fair scheduler operating at the OFDMA slot scale and accommodating both LOS and non-LOS UEs. We introduce an optimization framework for the proposed scheduler whose performance is hence simulated in a reference scenario. The spectral efficiency figures in our tests confirm a large gain of the scheme against a baseline direct path scheme. Finally, the scheduler optimization permits to achieve a further improvement of 15-20% for non line-of-sight users.

International audience; This article analyzes the impacts of objective (duration, conditions, demographics) and subjective (preferences, attitudes, etc.) factors on well-being in transportation. We rely on an exploratory survey carried out in 2017 and 2018 on the daily trips of 126 students (608 trips) toward their place of studies at Paris-Saclay. Using a questionnaire accessible from a smartphone, participants were asked to describe their daily trip (duration, conditions, activities, etc.), and give a rate after each trip. The results confirm the negative effect of travel time for trips longer than 80 minutes, and a time loss aversion effect. Difficult transport conditions (delays, space, noise, etc.) have also negative impacts on travelers’ well-being. Finally, some activities realised on smartphones (music, videos, etc.) affect travelers’ satisfaction, because they are interesting or necessary (music), or performed to pass the time in transports (games, videos). If usual travelers seem to be resigned by the trip duration, they seem to be more sensitive to unfavorable transport conditions as well as certain activities carried out while traveling.; Cet article propose une analyse de l’impact d’éléments objectifs (temps, conditions, caractéristiques socio-démographiques) et subjectifs (préférences, attitudes, etc.) sur le bien-être ressenti par les utilisateurs de transports en commun. Pour cela, une enquête exploratoire a été réalisée en 2017 et 2018 sur les déplacements quotidiens de 126 étudiants (608 trajets) vers leur lieu d’étude sur le plateau de Saclay. Grâce à un questionnaire accessible depuis leur smartphone, les participants sont interrogés sur leurs trajets quotidiens (temps, conditions, activités, etc.) et attribuent une note à chaque trajet effectué. Les résultats issus de nos modèles confirment un effet négatif de la durée du transport au-delà de 80 minutes, ainsi qu’une aversion à la perte de temps par rapport à la durée habituelle du trajet. Les conditions de transport difficiles (retard, espace, bruit, etc.) ont également un impact négatif sur le bien-être des passagers. Enfin, certaines activités réalisées en période de mobilité (musique, vidéos, etc.), font varier la satisfaction apportée par les déplacements, car elles sont intéressantes ou nécessaires (musique) ou réalisées pour faire passer le temps dans les transports (jeux, visionnage de vidéos). Si les voyageurs habitués à leur trajet semblent être résignés par la durée transport (effet de résignation), ils semblent plus sensibles aux conditions de transport défavorables ainsi qu’à certaines activités pratiquées au cours des déplacements.

The evolution of the Internet of Things (IoT) is driving an extraordinary growth of traffic and processing demands, persuading 5G players to change their infrastructures. In this context, Fog computing emerges as a potential solution, providing nearby resources to run IoT applications. However, the Fog raises several challenges which hinders its adoption. In this article, we consider the reconfiguration problem , i.e., how to dynamically adapt the placement of IoT applications running on the Fog, depending on application needs and evolution of resource usage. We propose and evaluate a series of reconfiguration algorithms, based on both online scheduling and online learning approaches. Through an extensive set of experiments in a realistic testbed, we demonstrate that the performance strongly depends on the quality and availability of information from both Fog infrastructure and IoT applications. This information mainly concerns the application’s resource usage (estimated by the user during the design of the application) and the availability of resources in the infrastructure (collected by commercial off-the-shelf monitoring tools). Finally, we show that a reactive and greedy strategy, which relies on this additional information, can overcome the performance of state-of-the-art online learning algorithms, even in a scenario with inaccurate information.

International audience; Distributed Denial of Service (DDoS) cyberattacks represent a major security risk for network operators and internet service providers. They thus need to invest in security solutions to protect their network against DDoS attacks. The present work focuses on deploying a network function virtualization based architecture to secure a network against an ongoing DDoS attack. We assume that the target, sources and volume of the attack have been identified. However, due to 5G network slicing, the exact routing of the illegitimate flow in the network is not known by the internet service provider. We seek to determine the optimal number and locations of virtual network functions in order to remove all the illegitimate traffic while minimizing the total cost of the activated virtual network functions. We propose a robust optimization framework to solve this problem. The uncertain input parameters correspond to the amount of illegitimate flow on each path connecting an attack source to the target and can take values within a predefined uncertainty set. In order to solve this robust optimization problem, we develop an adversarial approach in which the adversarial sub-problem is solved by a Branch & Price algorithm. The results of our computational experiments, carried out on medium-size randomly generated instances, show that the

Network Slicing is a major stake in 5G networks and beyond, which is mainly enabled by Network Function Virtualization (NFV) and Software Defined Networks (SDN). These two paradigms enable telcos to offer virtual networks meeting specific needs of the vertical markets on the top on the same shared physical infrastructure. An important challenge in the implementation at scale of Network Slicing is Network Slice Placement and the optimal allocation of virtualized resources. This can be formulated as a multi-objective Integer Linear Programming (ILP) problem. However, ILP suffers from combinatorial explosion when solving NP-hard problem, especially for large-scale scenarios. Heuristics and Machine Learning (ML) algorithms have been investigated as efficient means of tackling this kind of problem.This PhD thesis investigates how to optimize Network Slice Placement in distributed large-scale infrastructures focusing on online heuristic and Deep Reinforcement Learning (DRL) based approaches. First, we rely on ILP to propose a data model for enabling on-Edge and on-Network Slice Placement. In contrary to most studies related to placement in the NFV context, the proposed ILP model considers complex Network Slice topologies and pays special attention to the geographic location of Network Slice Users and its impact on the End-to-End (E2E) latency. Extensive numerical experiments show the relevance of taking into account the user location constraints. Then, we rely on an approach called the “Power of Two Choices"(P2C) to propose an online heuristic algorithm for the problem which is adapted to support placement on large-scale distributed infrastructures while integrating Edge-specific constraints. The evaluation results show the good performance of the heuristic that solves the problem in few seconds under a large-scale scenario. The heuristic also improves the acceptance ratioof Network Slice Placement Requests when compared against a deterministic online ILP-based solution.Finally, we investigate the use of ML methods, more specifically DRL, for increasing scalability and automation of Network Slice Placement considering a multi-objective optimization approach to the problem. We first propose a DRL algorithm for Network Slice Placement which relies on the Advantage Actor Critic algorithm for fast learning, and Graph Convolutional Networks for feature extraction automation. Then, we propose an approach we call Heuristically Assisted Deep Reinforcement Learning (HA-DRL), which uses heuristics to control the learning and execution of the DRL agent. We evaluate this solution trough simulations under stationary, cycle-stationary and non-stationary network load conditions. The evaluation results show that heuristic control is an efficient way of speeding up the learning process of DRL, achieving a substantial gain in resource utilization, reducing performance degradation, and is more reliable under unpredictable changes in network load than non-controlled DRL algorithms.; Le découpage du réseau est un enjeu majeur des réseaux 5G et au-delà, ce qui est principalement permis par la virtualisation des fonctions réseau (NFV) et les réseaux définis par logiciel (SDN). Ces deux paradigmes permettent aux opérateurs de télécommunications de proposer des réseaux virtuels répondant aux besoins spécifiques des marchés verticaux en utilisant une même infrastructure physique partagée. Un défi important dans la mise en oeuvre à l’échelle du découpage du réseau est le placement des tranches de réseau et l’allocation optimale des ressources virtualisées. Ce problème peut être formulé comme un problème d’optimisation multi-objectifs via la programmation linéaire en nombres entiers (ILP). Cependant, l’ILP souffre d’une explosion combinatoire lors de la résolution de problèmes NP-difficiles, en particulier pour les scénarios à grande échelle. Les heuristiques et les algorithmes d’apprentissage automatique (ML) sont envisagés comme des moyens efficaces pour résoudrece type de problème. Cette thèse examine comment optimiser le placement de tranches (slices) de réseau dans les infrastructures distribuées à grande échelle en se concentrant sur des approches heuristiques en ligne et basées sur l’apprentissage par renforcement profond (DRL). Tout d’abord, nous nous appuyons sur l’ILP pour proposer un modèle de données permettant le placement de tranches de réseau sur le bord et le coeur du réseau. Contrairement à la plupart des études relatives au placement de fonctions réseau virtualisées, le modèle ILP proposé prend en compte les topologies complexes des tranches de réseau et accorde une attention particulière à l’emplacement géographique des utilisateurs des tranches réseau et à son impact sur le calcul de la latence de bout en bout. Des expérimentations numériques nous ont permis de montrer la pertinence de la prise en compte des contraintes de localisation des utilisateurs. Ensuite, nous nous appuyons sur une approche appelée "Power of Two Choices" pour proposer un algorithme heuristique en ligne qui est adapté à supporter le placement sur des infrastructures distribuées à grande échelle tout en intégrant des contraintes spécifiques au bord du réseau. Les résultats de l’évaluation montrent la bonne performance de l’heuristique qui résout le problème en quelques secondes dans un scénario à grande échelle. L’heuristique améliore également le taux d’acceptation des demandes de placement de tranches de réseau parrapport à une solution déterministe en ligne en utilisant l’ILP. Enfin, nous étudions l’utilisation de méthodes de ML, et plus particulièrement de DRL, pour améliorer l’extensibilité et l’automatisation du placement de tranches réseau en considérant une version multi-objectif du problème. Nous proposons d’abord un algorithme DRL pour le placement de tranches réseau qui s’appuie sur l’algorithme "Advantage Actor Critic" pour un apprentissage rapide, et sur les réseaux convolutionels degraphes pour l’extraction de propriétés. Ensuite, nous proposons une approche que nous appelons "Heuristically Assisted DRL" (HA-DRL), qui utilise des heuristiques pour contrôler l’apprentissage et l’exécution de l’agent DRL. Nous évaluons cette solution par des simulations dans des conditions de charge de réseau stationnaire, ensuite cyclique et enfin non-stationnaire. Les résultats de l’évaluation montrent que le contrôle par heuristique est un moyen efficace d’accélérer le processus d’apprentissage du DRL, et permet d’obtenir un gain substantiel dans l’utilisation des ressources, de réduire la dégradation des performances et d’être plus fiable en cas de changements imprévisibles de la charge du réseau que les algorithmes DRL non contrôlés.

As countries in Europe implement strategies to control the COVID-19 pandemic, different options are chosen regarding schools. Through a stochastic age-structured transmission model calibrated to the observed epidemic in Île-de-France in the first wave, we explored scenarios of partial, progressive, or full school reopening. Given the uncertainty on children’s role, we found that reopening schools after lockdown may increase COVID-19 cases, yet protocols exist to keep the epidemic controlled. Under a scenario with stable epidemic activity if schools were closed, reopening pre-schools and primary schools would lead to up to 76% [67, 84]% occupation of ICU beds if no other school level reopened, or if middle and high schools reopened later. Immediately reopening all school levels may overwhelm the ICU system. Priority should be given to pre- and primary schools allowing younger children to resume learning and development, whereas full attendance in middle and high schools is not recommended for stable or increasing epidemic activity. Large-scale test and trace is required to keep the epidemic under control. Ex-post assessment shows that progressive reopening of schools, limited attendance, and strong adoption of preventive measures contributed to a decreasing epidemic after lifting the first lockdown., The role of children in the spread of COVID-19 is not fully understood, and the circumstances under which schools should be opened are therefore debated. Here, the authors demonstrate protocols by which schools in France can be safely opened without overwhelming the healthcare system.

6th European Cyber Week (6ème semaine européenne de la cybersécurité et de la cyberdéfense)

International audience; The fifth generation (5G) of mobile telecommunication network is designed with an ambition to be a network faster, stronger, better and smarter than its predecessor. With the digital transformation, all industry sectors will develop new applications with new requirements regarding telecommunication networks that 5G should be able to meet. To meet the requirement of future 5G use cases and applications, it is crucial to study the complexity of such network system by distinguishing different parts, layers, components as well as their interdependencies. This paper describes the 5G networks from an End-to-End perspective (device, radio network, core network, data network) and from a multi-layer perspective (orchestration, virtualisation/containerization and infrastructure) to show how this system (or system of systems) is complex, especially when we address resilience challenges. Resilience requirements and challenges are further explained by proposing relevant scenarios and use cases. In this paper, we mainly intend to highlight 5G network complexity and open a discussion on methodologies to model such complex network for its resilience study with the hope that this paper could inspire the future study of researchers in the related field.

International audience; The 5G system should be flexible to cope with diversity of functions and infrastructure, in an efficient way while supporting a wide range of services associated with a set of requirements through Service Function Chaining (SFC). Fault management plays a paramount role to fulfill the SFC requirements towards reliability of telecommunication networks. Model-based (MB) approaches are knowledge-based systems that reason about a system from an explicit representation of its structure and functional behavior through a dependency model. MB techniques solve novel diagnosis problems and provide explanations for their decisions. However, current MB methods suffer from some limitations due to network virtualization challenges: lack of network visibility and dynamic topologies. To tackle those limitations, we propose a self-modeling approach and an active diagnosis process for virtual networks that considers two types of knowledge to build the model: acquired knowledge and learned knowledge provided by fault injection to expand and validate the proposed model. The experimental results from their application to a real-world virtual IP Multimedia Subsystem (vIMS) use case show that the self-modeling and the active diagnosis procedures are effective in determining the root cause(s) of a failure and explaining fault propagation.

Call detail records (CDR) collected by mobile phone network providers have been largely used to model and analyze human-centric mobility. Despite their potential, they are limited in terms of both spatial and temporal accuracy thus being unable to capture detailed human mobility information. Network Signaling Data (NSD) represent a much richer source of spatio-temporal information currently collected by network providers, but mostly unexploited for fine-grained reconstruction of human-centric trajectories. In this paper, we present TRANSIT, TRAjectory inference from Network SIgnaling daTa, a novel framework capable of processing NSD to accurately distinguish mobility phases from stationary activities for individual mobile devices, and reconstruct, at scale, fine-grained human mobility trajectories, by exploiting, with a DBSCAN-based clustering approach, the inherent recurrence of human mobility and the higher sampling rate of NSD. The validation on a ground-truth dataset of GPS trajectories showcases the superior performance of TRANSIT (80% precision and 96% recall) with respect to state-of-the-art solutions in the identification of movement periods, as well as an average 190 m spatial accuracy in the estimation of the trajectories. We also leverage TRANSIT to process a unique large-scale NSD dataset of more than 10 millions of individuals and perform an exploratory analysis of city-wide transport mode shares, recurrent commuting paths, urban attractivity and analysis of mobility flows. TRUE pub

International audience; In this article, we propose a framework for seasonal time series probabilistic forecasting. It aims at forecasting (in a probabilistic way) the whole next season of a time series, rather than only the next value. Probabilistic forecasting consists in forecasting a probability distribution function for each future position. The proposed framework is implemented combining several machine learning techniques 1) to identify typical seasons and 2) to forecast a probability distribution of the next season. This framework is evaluated using a wide range of real seasonal time series. On the one side, we intensively study the alternative combinations of the algorithms composing our framework (clustering, classification), and on the other side, we evaluate the framework forecasting accuracy. As demonstrated by our experiences, the proposed framework outperforms competing approaches by achieving lower forecasting errors.

International audience; Recently, solutions based on mobile edge computing paradigm have been widely discussed in academia and industry. This paradigm offers solutions to address limitations, in terms of battery lifetime and processing power, of mobile and constrained devices. Despite the ever-increasing capabilities of these devices, resource requirements of applications can often transcend what is available within a single device. Offloading intensive computation tasks to a distant server can help applications reach their desired performances. In this work, we tackle the problem of offloading heavy computation tasks of unmanned aerial vehicles (UAVs) while achieving the best possible tradeoff between energy consumption, time delay, and computation cost. We focus on a scenario of a fleet of small UAVs performing an exploration mission. During their mission, these constrained devices have to carry-out highly intensive computation tasks such as pattern recognition and video preprocessing. We formulate the problem using a non-cooperative theoretical game with N players and three pure strategies. We provide a comprehensive proof for the existence of a Nash equilibrium and implement accordingly a distributed algorithm that converges to such an equilibrium. Extensive simulations are performed in order to provide thorough results and assess the performances of the approach compared to three other models. Results show that our algorithm outperforms all the three approaches. Our approach achieved in average about 19%, 58%, and 55% better results compared to local computing, offloading to the edge server, and offloading to base station, respectively.

International audience; Low latency targets for Ultra-Reliable Low Latency Communications (URLLC) may be conflicting with their stringent reliability requirements due to the need for re-transmissions. We explore in this paper the different resource allocation schemes for transmissions and re-transmissions depending on the requirements of the underlying service and on the traffic characteristics, focusing on Industrial Internet of Things (IIoT). We namely consider schemes with individual reservation versus a pool of contention-based reserved resources. We provide novel resource allocation schemes for initial transmissions and re-transmissions and derive corresponding analytical models for loss rates. We then show how to set the system parameters that allow meeting the URLLC requirements with low resource consumption.

International audience; In this paper, we consider a dynamic device-todevice (D2D) communication model where transmitters and receivers have multiple antennas and adopt beamforming (BF). A continuous spatio-temporal model for the wireless network is analyzed, which combines a spatial stochastic point process and a dynamic birth-death process. We model BF by using a uniform linear array (ULA) and extend the result of Sankararaman and Baccelli on the stability condition of such a network. We show that the critical arrival rate increases with the number of antennas at the transmitter and the receiver.

International audience; This paper aims to discuss issues raised by Artificial Intelligence/Machine Learning (AI/ML) in work situations, in the light of literature and an ongoing empirical study. Based on semi-structured interviews with 21 workers and 15 designers, this study explores stakeholders’ viewpoints and experiences of AI systems. The preliminary findings show the place and the use of AI systems by workers in different situations, the issues of explainability and trust, the preoccupations of workers about the introduction of AI systems in work activities and the transformations that it may cause. Finally, we discuss about an analytical framework for analyzing and anticipating the consequences of AI systems on work activities and for designing these systems through a Human-Centered approach.

Licensed Shared Access (LSA) is a new sharing approach that aims to optimize the use of the 2.3-2.4 Ghz frequency band in order to support the deployment of 5G systems. Under LSA, Mobile Network Operators (MNOs) can share the 2.3-2.4 band with the incumbent of that band under some guarantees, specified in a license attributed by the regulator. In this paper, we focus on ascending auctions to allocate and price licenses. We first show how to implement an ascending version of the well-known Vickrey-Clarke-Groves (VCG) mechanism in the LSA context, but highlight that it may introduce some computational complexity problems. We therefore propose another ascending mechanism, called C-LSA, based on the clinching approach. We also design the one-shot equivalent (in terms of allocations and payments) of C- LSA that we use to compare the performance of VCG and C-LSA through simulations. Our simulations suggest that C-LSA is an interesting candidate to allocate LSA licenses, since the mechanism yields considerably larger revenues than VCG, for a very limited loss of allocation efficiency (around 4%).

In this work, we study a new model for continuum line-of-sight percolation in a random environment driven by the Poisson-Voronoi tessellation in the $d$-dimensional Euclidean space. The edges (one-dimensional facets, or simply 1-facets) of this tessellation are the support of a Cox point process, while the vertices (zero-dimensional facets or simply 0-facets) are the support of a Bernoulli point process. Taking the superposition $Z$ of these two processes, two points of $Z$ are linked by an edge if and only if they are sufficiently close and located on the same edge (1-facet) of the supporting tessellation. We study the percolation of the random graph arising from this construction and prove that a 0-1 law, a subcritical phase as well as a supercritical phase exist under general assumptions. Our proofs are based on a coarse-graining argument with some notion of stabilization and asymptotic essential connectedness to investigate continuum percolation for Cox point processes. We also give numerical estimates of the critical parameters of the model in the planar case, where our model is intended to represent telecommunications networks in a random environment with obstructive conditions for signal propagation., 30 pages, 4 figures. Accepted for publication in Advances in Applied Probability

International audience; Current generation of switches are highly programmable and able to support stateful. However, these switches cannot perform floating point operations. As a result several network applications have to be run on external servers or middleboxes in the network. We introduce InREC, a system that extends the capabilities of programmable switches to support in-network real-valued operations using the IEEE half-precision floating point representation. Our demo on a Barefoot Tofino switches demonstrates the efficiency of InREC for in-network computation by computing a logistic regression function to classify devices (IoT and laptop) in a simulated home environment.

International audience; Current generation of Reconfigurable Match-Action Tables switches are highly programmable, able to support stateful operations and pipeline specifications using languages like P4. Nevertheless, these switches do not offer primitives to support real-valued operations on the data plane, thus requiring support from external servers or middleboxes to perform advanced operations. We introduce InREC, a system that extends the capabilities of programmable switches to support in-network real-valued operations using the IEEE half-precision floating point representation. It relies on decomposing real-valued functions into lookup tables taking into account the RMT model constraints to reach the right trade-off between accuracy and resource usage. InREC prototype on Barefoot Tofino switches demonstrates the efficiency of InREC for in-network computation of different types of operations and its application for in-network logistic regression models used for classification problems. Our evaluation of InREC shows that it is possible to implement complex in-network applications with high accuracy and low latency.

International audience; The high data rates and diversity of services in 5G require a flexible and efficient use of all the available frequencies. In 5G networks, new approaches of dynamic spectrum sharing will be deployed, allowing Mobile Network Operators (MNOs) to access other incumbents' spectrum, after obtaining a license from the regulator. The attribution of licenses will be made via auction mechanisms valid for given geographical areas and durations. To determine how to bid, each MNO has to estimate his valuation for spectrum i.e., how much he is willing to pay for spectrum. In this paper, we propose a model for estimating that valuation. The model is based on Markov chain modeling of user behavior, to compute the MNO satisfaction as a function of the obtained spectrum. We then illustrate our method by applying it to real operator data.

International audience; We consider a network with Massive Multiple Input Multiple Output (M-MIMO) base stations using a Grid of Beams (GoB) for data and control channels. 5G allows to establish interference relations between beams of neighboring cells. Such relations can be used to automatically generate a beam relation matrix, denoted as Automatic Neighbor Beam Relation (ANBR) matrix that can be very useful for optimizing different resource allocation processes. This paper shows how the ANBR matrix can be used to coordinate scheduling of neighboring cells with a small amount of information exchange. The coordination is performed by judicially muting or reducing the band-width of certain beams in the process of MultiUser (MU) Proportional Fair (PF) scheduling. Numerical results show how the coordination approach can bring about significant performance gain.

This paper proposes a framework to extract dynamic trip flows and travel demand patterns from large-scale 2 G and 3 G cellular signaling data. Novel data pre-processing techniques based on cell phone activity metrics are presented. The trip extraction method relies on the detection of stationary activities to form trip sequences related to resident users. A probabilistic solution is introduced to estimate the trip starting time, allowing to aggregate trips by time of the day and reconstruct hourly travel flows. To better characterize these flows, a spatial clustering process combined with land-use data is proposed based on the temporal demand profile of each zone. Empirical comparisons have been performed showing that the resulting dynamic travel demand patterns are consistent with those obtained from travel survey data with high correlation coefficients of about 0.9. The results prove the potential of signaling data to generate low-cost valuable information for large-scale travel demand modeling.

We propose to demonstrate a network slice placement optimization solution based on Deep Reinforcement Learning (DRL), referred to as Heuristically-controlled DRL, which uses a heuristic to control the DRL algorithm convergence. The solution is adapted to realistic networks with large scale and under non-stationary traffic conditions (namely, the network load). We demonstrate the applicability of the proposed solution and its higher and stable performance over a non-controlled DRL-based solution. Demonstration scenarios include full online learning with multiple volatile network slice placement request arrivals., Comment: arXiv admin note: substantial text overlap with arXiv:2010.08295

This PhD thesis investigates how to optimize Network Slice Placement in distributed large-scale infrastructures focusing on online heuristic and Deep Reinforcement Learning (DRL) based approaches. First, we rely on Integer Linear Programming (ILP) to propose a data model for enabling on-Edge and on-Network Slice Placement. In contrary to most studies related to placement in the NFV context, the proposed ILP model considers complex Network Slice topologies and pays special attention to the geographic location of Network Slice Users and its impact on the End-to-End (E2E) latency. Extensive numerical experiments show the relevance of taking into account the user location constraints. Then, we rely on an approach called the “Power of Two Choices"(P2C) to propose an online heuristic algorithm for the problem which is adapted to support placement on large-scale distributed infrastructures while integrating Edge-specific constraints. The evaluation results show the good performance of the heuristic that solves the problem in few seconds under a large-scale scenario. The heuristic also improves the acceptance ratio of Network Slice Placement Requests when compared against a deterministic online ILP-based solution. Finally, we investigate the use of ML methods, more specifically DRL, for increasing scalability and automation of Network Slice Placement considering a multi-objective optimization approach to the problem. We first propose a DRL algorithm for Network Slice Placement which relies on the Advantage Actor Critic algorithm for fast learning, and Graph Convolutional Networks for feature extraction automation. Then, we propose an approach we call Heuristically Assisted Deep Reinforcement Learning (HA-DRL), which uses heuristics to control the learning and execution of the DRL agent. We evaluate this solution trough simulations under stationary, cycle-stationary and non-stationary network load conditions. The evaluation results show that heuristic control is an efficient way of speeding up the learning process of DRL, achieving a substantial gain in resource utilization, reducing performance degradation, and is more reliable under unpredictable changes in network load than non-controlled DRL algorithms., Cette thèse examine comment optimiser le placement de tranches (slices) de réseau dans les infrastructures distribuées à grande échelle en se concentrant sur des approches heuristiques en ligne et basées sur l'apprentissage par renforcement profond (DRL). Tout d'abord, nous nous appuyons sur la programmation linéaire en nombre entiers (ILP) pour proposer un modèle de données permettant le placement de tranches de réseau sur le bord et le cœur du réseau. Contrairement à la plupart des études relatives au placement de fonctions réseau virtualisées, le modèle ILP proposé prend en compte les topologies complexes des tranches de réseau et accorde une attention particulière à l'emplacement géographique des utilisateurs des tranches réseau et à son impact sur le calcul de la latence de bout en bout. Des expérimentations numériques nous ont permis de montrer la pertinence de la prise en compte des contraintes de localisation des utilisateurs.Ensuite, nous nous appuyons sur une approche appelée "Power of Two Choices" pour proposer un algorithme heuristique en ligne qui est adapté à supporter le placement sur des infrastructures distribuées à grande échelle tout en intégrant des contraintes spécifiques au bord du réseau. Les résultats de l'évaluation montrent la bonne performance de l'heuristique qui résout le problème en quelques secondes dans un scénario à grande échelle. L'heuristique améliore également le taux d'acceptation des demandes de placement de tranches de réseau par rapport à une solution déterministe en ligne en utilisant l'ILP.Enfin, nous étudions l'utilisation de méthodes de ML, et plus particulièrement de DRL, pour améliorer l'extensibilité et l'automatisation du placement de tranches réseau en considérant une version multi-objectif du problème. Nous proposons d'abord un algorithme DRL pour le placement de tranches réseau qui s'appuie sur l'algorithme "Advantage Actor Critic" pour un apprentissage rapide, et sur les réseaux convolutionels de graphes pour l'extraction de propriétés. Ensuite, nous proposons une approche que nous appelons "Heuristically Assisted DRL" (HA-DRL), qui utilise des heuristiques pour contrôler l'apprentissage et l'exécution de l'agent DRL. Nous évaluons cette solution par des simulations dans des conditions de charge de réseau stationnaire, ensuite cyclique et enfin non-stationnaire. Les résultats de l'évaluation montrent que le contrôle par heuristique est un moyen efficace d'accélérer le processus d'apprentissage du DRL, et permet d'obtenir un gain substantiel dans l'utilisation des ressources, de réduire la dégradation des performances et d'être plus fiable en cas de changements imprévisibles de la charge du réseau que les algorithmes DRL non contrôlés.

This thesis studies the transport of critical services in 5G networks, where unlicensed spectrum is advocated to minimize the cost and to cope with the high demand for frequency resources. We first evaluate the performance of Ultra-Reliable Low-Latency Communication (URLLC) which has stringent requirements on reliability and delay, on the order of 99.999% and 1 ms, respectively, transported in unlicensed spectrum. We propose a model based on a Markov chain to quantify the reliability within a delay constraint under Listen-Before-Talk (LBT) medium access procedure, then we deduce the maximum number of stations that can be handled at the same time, while respecting URLLC constraints. This analysis is then used to investigate novel methods for the joint transmission of URLLC over unlicensed and licensed spectrum. We propose three methods for the joint access to available resources, and demonstrate that the optimal method to access the resources is by using licensed ones only when unlicensed transmission fails within a given time budget. This method is then studied in the case of multiple tenants in proximity competing over the same unlicensed channel. If all tenants try to maximize their usage of unlicensed resources then everyone will end up in a tragedy of the commons type of situation. We show that at least one equilibrium point exists for this system which minimizes the cost for all tenants. We study later the coexistence of URLLC with other 5G services, such as enhanced Mobile Broadband (eMBB), in unlicensed spectrum. eMBB has large packets and its multiplexing with URLLC may entail a large degradation in URLLC performance. For that, we propose a new technique to prioritize URLLC packets by transmitting them with higher power. However, high power transmission is not systematically performed to reduce the interference on other users and also to reduce the energy consumption, which is very important for battery-powered devices. In this case, two methods were proposed to transmit with high power, leaving it as a last resort. One is LBT-agnostic and transmits whenever the packet delay approaches time-out, while the other respects LBT and uses high power only when transmission opportunities occur beyond a time threshold. We then propose a decentralized implementation of the time-threshold approach. We formulate the problem as an optimization problem where transmitters are to choose the optimal policy (time threshold) which minimizes the energy consumption while preserving URLLC requirements. We then solve the optimization problem using a learning approach and show a slow convergence to the optimal policy due to the fact that losses are rare events. To remedy to this, we employ the optimization framework and the prior knowledge of the system to accelerate this learning. We finally study the decentralized approach for a different type of critical services which focuses on the freshness of the information, known as the Age of Information (AoI). In this context, instead of guaranteeing a reliability target within a delay, the packet must be delivered as soon as it is generated, or else it loses its value. We demonstrate that optimal policies in the AoI context tend to start aggressively, and reduce the transmission power when the age of the packet increases.; Cette thèse étudie le transport de services critiques dans les réseaux 5G, où le spectre non-licencié est préconisé pour minimiser le coût et faire face à la forte demande de ressources en fréquences. Nous évaluons d'abord les performances des services critiques type URLLC (Ultra-Reliable Low-Latency Communication) qui a des exigences strictes en matière de fiabilité et de latence, de l'ordre de 99,999% et 1 ms, respectivement, transporté dans le spectre non-licencié. Nous proposons un modèle basé sur une chaîne de Markov pour quantifier la fiabilité sous contrainte de délai, sous procédure d'accès au support Listen-Before-Talk (LBT), puis nous en déduisons le nombre maximum de stations pouvant être servies en même temps, tout en respectant l'URLLC contraintes. Cette analyse est ensuite utilisée pour étudier de nouvelles méthodes pour la transmission conjointe d'URLLC sur les spectres non-licencié et licencié. Nous proposons trois méthodes pour l'accès conjoint aux ressources disponibles et démontrons que la méthode optimale pour accéder aux ressources consiste à utiliser des ressources licenciées, uniquement lorsque la transmission dans le système non-licencié échoue dans un budget de temps donné. Cette méthode est ensuite étudiée dans le cas de plusieurs tenants à proximité en concurrence sur le même canal non-licencié. Si tous les tenants essaient de maximiser leur utilisation des ressources non-licenciées, tout le monde se retrouvera dans une situation type « tragédie des biens communs ». Nous montrons qu'au moins un point d'équilibre existe pour ce système qui minimise le coût pour tous les tenants. Nous étudions ensuite la coexistence d'URLLC avec d'autres services 5G, tels que le haut débit mobile amélioré eMBB (enhanced Mobile Broadband), dans le spectre non-licencié. eMBB a de grandes paquets et son multiplexage avec URLLC peut entraîner une forte dégradation des performances d'URLLC. Pour cela, nous proposons une nouvelle technique pour prioriser les paquets URLLC en les transmettant avec une puissance plus élevée. Cependant, la transmission à haute puissance n'est pas systématiquement effectuée afin de réduire les interférences sur les autres utilisateurs et aussi pour réduire la consommation d'énergie, ce qui est très important pour les appareils alimentés par batterie. Dans ce cas, deux méthodes ont été proposées pour transmettre avec une puissance élevée, en ne le laissant qu'en dernier recours. L'un est indépendant du LBT et transmet une fois le délai de paquet approche de l'expiration, tandis que l'autre respecte le LBT et n'utilise une puissance élevée que lorsque les opportunités de transmission se produisent au-delà d'un seuil de temps. Nous proposons ensuite une mise en œuvre décentralisée de l'approche par seuil de temps décrit ci-dessus. Nous formulons le problème dans le cadre d'optimisation où les émetteurs doivent choisir la politique optimale (seuil de temps) qui minimise la consommation d'énergie tout en préservant les exigences d'URLLC. Nous résolvons ensuite le problème d'optimisation en utilisant une approche d'apprentissage et montrons une lente convergence vers la politique optimale du fait que les pertes sont des événements rares. Pour y remédier, nous utilisons le cadre d'optimisation et la connaissance préalable du système pour accélérer cet apprentissage. Nous étudions enfin l'approche décentralisée pour un type différent de services critiques qui met l'accent sur la fraîcheur de l'information, connue sous le nom d'Age de l'Information (AoI). Dans ce contexte, au lieu de garantir une cible de fiabilité dans un délai, le paquet doit être livré dès sa génération, sinon sa valeur se dégrade. Nous démontrons que les politiques optimales dans le contexte AoI ont tendance à démarrer de manière agressive et à réduire la puissance de transmission lorsque l'âge du paquet augmente.

International audience; In this paper, we exploit the use of unlicensed spectrum in industrial scenarios for UltraReliable Low-Latency Communication (URLLC). We consider a set of automated machines communicating with a central base station and competing to the radio resources among themselves and with enhanced Mobile Broadband (eMBB) traffic. Observing the impact of Listen-Before-Talk (LBT) on the stringent reliability and delay requirements, we propose a progressive bilevel power transmission policy for URLLC, where a high power transmission can be regarded as a preemption of eMBB packets with lower power. We formulate the objective as a stochastic optimization problem under constraints and solve it in the simple Aloha-like case, and in the more complex LBT case. We then use this model to build an online (re-)transmission algorithm which enables the URLLC stations to determine in a distributed manner the optimal policy using observations of the environment parameters.

Background: On March 17, 2020, French authorities implemented a nationwide lockdown to respond to the COVID-19 epidemic and curb the surge of patients requiring critical care. Assessing the effect of lockdown on individual displacements is essential to quantify achievable mobility reductions and identify the factors driving the changes in social dynamics that affected viral diffusion. We aimed to use mobile phone data to study how mobility in France changed before and during lockdown, breaking down our findings by trip distance, user age and residency, and time of day, and analysing regional data and spatial heterogeneities.Methods: For this population-based study, we used temporally resolved travel flows among 1436 administrative areas of mainland France reconstructed from mobile phone trajectories. Data were stratified by age class (younger than 18 years, 18-64 years, and 65 years or older). We distinguished between residents and non-residents and used population data and regional socioeconomic indicators from the French National Statistical Institute. We measured mobility changes before and during lockdown at both local and country scales using a case-crossover framework. We analysed all trips combined and trips longer than 100 km (termed long trips), and separated trips by daytime or night-time, weekdays or weekends, and rush hours.Findings: Lockdown caused a 65% reduction in the countrywide number of displacements (from about 57 million to about 20 million trips per day) and was particularly effective in reducing work-related short-range mobility, especially during rush hour, and long trips. Geographical heterogeneities showed anomalous increases in long-range movements even before lockdown announcement that were tightly localised in space. During lockdown, mobility drops were unevenly distributed across regions (eg, Île-de-France, the region of Paris, went from 585 000 to 117 000 outgoing trips per day). They were strongly associated with active populations, workers employed in sectors highly affected by lockdown, and number of hospitalisations per region, and moderately associated with the socioeconomic level of the regions. Major cities largely shrank their pattern of connectivity, reducing it mainly to short-range commuting (95% of traffic leaving Paris was contained in a 201 km radius before lockdown, which was reduced to 29 km during lockdown).Interpretation: Lockdown was effective in reducing population mobility across scales. Caution should be taken in the timing of policy announcements and implementation, because anomalous mobility followed policy announcements, which might act as seeding events. Conversely, risk aversion might be beneficial in further decreasing mobility in highly affected regions. We also identified socioeconomic and demographic constraints to the efficacy of restrictions. The unveiled links between geography, demography, and timing of the response to mobility restrictions might help to design interventions that minimise invasiveness while contributing to the current epidemic response., Agence Nationale de la Recherche, EU, REACTing.

International audience; The rise of new technologies, like serious games, opens up new horizons in terms of training, skills development and knowledge acquisition. The design of these serious games which have educational goals is based on various requirements. The retrospective simulation approach that is the subject of this paper was developed as part of the design of this type of innovative serious game. It is guided by ergonomic and work psychology. The innovative dimension involves designing realistic behaviours, emotions and profiles of non-player characters (NPCs). The retrospective simulation allows to access and model real situations of change involving emotions and social behaviours. The purpose of this paper is to report the contributions and limitations of the retrospective simulation of social-emotional activity. The deployed methods have three different goals: comprehensive goal, re-situating goal and finally projective goal. The sample consists of 27 participants, 13 women and 14 men, with an average age of 45 years. They have experienced all types of change: professional, structural, organizational and technological. Thanks to the results, we were able to reconstitute for each participant chronicles of socio-emotional experience. These chronicles make it possible to recompose the situation of change as well as to associate the various events to the participant's emotional experiences and social behaviours. In addition, personas – archetypal representations of different profiles of people with similar characteristics – have been developed. They are resources for designing NPCs and for the script of the serious game. In this way, the retrospective simulation, turned towards a past activity, allows on the one hand, accessing and modeling the emotional and social experiences and on the other hand, to contribute to the development of the prospective simulation that is serious game.; La démarche de simulation rétrospective présentée dans cet article a été développée dans le cadre de la conception d’un serious game innovant. Cet environnement de formation a pour objectif de sensibiliser des managers chargés d’accompagner un changement. Sa dimension innovatrice renvoie à l’exigence du réalisme des comportements sociaux, des émotions et des profils psychologiques des personnages non joueurs (PNJ). L’objectif de cet article est de relater la démarche ainsi que l’intérêt de la simulation rétrospective qui est utilisée pour accéder et modéliser l’expérience socio-émotionnelle vécue dans une situation de changement. La méthodologie déployée comporte trois approches complémentaires à visée : (i) compréhensive, (ii) restituante et resituante et (iii) projective et resituante. L’échantillon est composé de 27 sujets, qui ont vécu des situations de changement. Grâce aux résultats obtenus, nous avons pu reconstituer pour chaque sujet des chroniques d’expérience socio-émotionnelles. Ces chroniques permettent de scénariser le processus de changement (par situations et évènements marquants) et d’apparier les éprouvés émotionnels et les conduites sociales – qui ont pu se manifester – avec chaque évènement. Par ailleurs, la méthode des personas a permis de guider la conception des PNJ et la scénarisation du serious game. Les bénéfices et limites de cette simulation sont enfin discutés.

International audience; We focus in this paper on the transport of critical services in unlicensed spectrum, where stringent constraints on latency and reliability are to be met, in the context of Ultra-Reliable Low Latency Communication (URLLC). Since contention-based medium access performs poorly in the case of high traffic load, we propose a new transmission scheme where the transmitter can increase its transmission power when the delay of the packet approaches the delay constraint, increasing by that its chance of being decoded even in case of collision with other lower-power packets. We are however interested in minimizing the usage of high power transmissions, mainly to conserve energy for battery-powered devices and to limit the range of interference. Therefore, we define a transmission policy that makes use of a delay threshold after which the high-power transmission starts, and propose a new online-learning approach based on Multi-Armed Bandit (MAB) in order to identify the policy which achieves minimum energy consumption while guaranteeing reliability. However, we observe that the MAB converges slowly to the optimal policy because the loss event is rare in the load regime of interest. We then propose a model-aided learning approach where a simple analytical model helps estimating the longterm reliability resulting from an action and thus its reward. Our results show a significant enhancement of the convergence towards the optimal policy.

International audience; In this work, we model the network slice provi-sioning as an optimization problem including novel mapping and provisioning requirements rising with new 5G radio and core function placement policies. We propose an MILP-based formulation that joins different functional splitting strategies with different network function sharing policies and novel mapping continuity constraints from 5G specifications. We show by numerical simulations the impact of taking into full and partial consideration these peculiar sets of novel technical constraints.

International audience; Recent developments in orchestration and machine learning have made network automation more feasible, allowing the transition from error-prone and time-consuming manual manipulations to fast and refined automated responses in areas such as security and management. This article investigates the capabilities of a deep reinforcement learning agent to learn how to automatically share prefix announcements of an Autonomous System to its neighbors, in order to mitigate undesired network behaviors and therefore increase network resiliency and security. Our work focuses on network saturation, tackling the problem of network responsiveness in today's massive content delivery context. Results not only prove feasibility of such an agent, but also demonstrate its ability to minimize traffic loss as well as the number of actions to be performed by the automation process.

International audience; Network slicing has emerged as a promising technical solution to ensure the coexistence of the various 5G services. While the evolution of the 5G architecture for supporting slicing has been thoroughly studied, the impact of the architectural options on RAN resource allocation efficiency is still unclear. This article fills a gap in this area by assessing the impact of architecture choices on the quality of service of different services, with a focus on ultra-reliable low-latency communication applications. We propose architectural options based on the placement of the entities responsible for implementing these functions. We then assess their impact on the radio resource allocation flexibility when slices span two radio access technologies with redundant coverage. Our numerical experiments show that the placement of the slice management functions plays a crucial role in the choice of the radio resource allocation scheme that best fits URLLC slices. Index Terms-5G, network slicing, quality of service, ultra-reliable low-latency communication (URLLC), resource allocation , redundancy, join the shortest queue.

International audience; Network Slicing has its roots in Network Function Virtualization (NFV) allowing high flexibility in the delivery of end-to-end network services. To achieve Network Slicing promises on efficiency, Network Slice Providers have to ensure optimized resource utilization and to guarantee Quality of Service when managing the life-cycle of a Network Slice. We focus in this paper on Network Slice Placement, intimately related to the VNF Placement and Chaining problem. In contrary to most studies related to VNF placement, we deal with the most complete and complex Network Slice topologies and we pay special attention to the geographic location of Network Slice Users. We propose a data model adapted to Integer Linear Programming. Extensive numerical experiments assess the relevance of taking into account the user location constraints.

The fifth generation of cellular networks is expected to provide coverage for an unprecedented number of devices over large areas. One of the main paradigms investigated to address this challenge, called Device-to-Device (D2D) communication, consists in allowing for short-range direct communications between network devices. An application of significant economic interest for operators is the one of the uberisation of networks, where an operator having no (or very few) network infrastructure could build a mobile network relying only on its end-devices (users). In this thesis, we study new mathematical models of D2D networks in urban environments. We see the street system of a city as a planar Poisson-Voronoi tessellation (PVT). Network users are given by a Cox process supported by the edges of the PVT while additional network relays are given by a Bernoulli process on the vertices of the PVT. The network is then modelled by a connectivity graph as follows: vertices are the atoms of both these processes and fixed-range connections between them possible only along the PVT edges or between network nodes located on adjacent PVT edges. Percolation of this random graph (existence of an infinite connected component with positive probability) is interpreted as good connectivity of the network. Using renormalisation techniques, we prove the existence of phase transitions between different connectivity regimes, in particular those where percolation can be solely ensured by the relays or, on the contrary, where a sufficient density of users is essential. Performing numerical simulations with original path-finding algorithms, we estimate critical parameters (e.g. the density of relays and users) allowing for good connectivity of the network. Finally, we also introduce appropriate cost models and use our numerical estimates to study the economic feasibility of uberisation scenarios of telecommunications networks., La cinquième génération de réseaux mobiles devrait être en mesure de servir un nombre jamais vu d'appareils sur de vastes étendues. Un des principaux paradigmes étudiés pour répondre à ce défi est celui des communications Device-to-Device (D2D), c’est-à-dire de communications directes et de courte portée entre les appareils d'un réseau. Un cas d'usage d'intérêt économique significatif pour les opérateurs est celui de l'übérisation des réseaux : grâce au D2D, un nouvel opérateur n'ayant pas (ou presque pas) d'infrastructures réseau pourrait construire un réseau mobile fonctionnel reposant uniquement sur des terminaux mobiles. Dans cette thèse, nous nous intéressons à de nouveaux modèles mathématiques de réseaux D2D en environnement urbain. Nous modélisons la voirie d'une ville par une mosaïque de Poisson-Voronoi plane. Les utilisateurs du réseau sont modélisés par un processus ponctuel de Cox sur les arêtes de cette mosaïque tandis que des relais supplémentaires sont modélisés par un processus ponctuel de Bernoulli sur ses sommets. Le réseau D2D est alors représenté par un graphe de connectivité dont les sommets sont les atomes des deux processus ponctuels précédents et où les connexions sont possibles uniquement entre des noeuds du réseau situés sur la même arête ou sur deux arêtes incidentes de la mosaïque de Poisson-Voronoi. Nous interprétons la percolation de ce graphe aléatoire (c'est-à-dire une probabilité positive d'existence d'une composante connexe infinie) comme signe d'une bonne connectivité du réseau. À l'aide de techniques de renormalisation, nous prouvons l'existence de transitions de phases entre différents régimes de connectivité : ceux où la percolation peut être assurée seulement par les relais ou, a contrario, ceux où une densité suffisante d'utilisateurs est nécessaire. À l'aide de simulations numériques et de nouveaux algorithmes de détection de chemins, nous estimons des paramètres critiques (par exemple la densité minimale d'utilisateurs) permettant une connectivité à grande échelle du réseau. Enfin, nous introduisons des modèles de coûts et utilisons les estimations précédentes pour étudier la faisabilité économique de scénarios d'übérisation des réseaux de télécommunications.

International audience; Virtualization methods and techniques play an important role in the development of cloud infrastructures and their services. They enable the decoupling of virtualized resources from the underlying hardware, and facilitate their sharing amongst multiple users. They contribute to the building of elaborated cloud services that are based on the instantiation and composition of these resources. Different models may support such a virtualization, including virtualization based on type-I and type-II hypervisors, OS-level virtualization, and unikernel virtualization. These virtualization models pose a large variety of security issues, but also offer new opportunities for the protection of cloud services. In this article, we describe and compare these virtualization models, in order to establish a reference architecture of cloud infrastructure. We then analyze the security issues related to these models from the reference architecture, by considering related vulnerabilities and attacks. Finally, we point out different recommendations with respect to the exploitation of these models for supporting cloud protection.

International audience; Cloud-Radio Access Network is a promising mobile network architecture that centralizes the computing resources in the Baseband Unit pool which adds more flexibility and increases network performance. However, as computing resources are shared among the Radio Remote Heads (RRH) connected to the Baseband Unit pool, efficient scheduling algorithms should be explored to meet the deadlines requirements of RRHs' subframes and to increase the network throughput. In this paper, we propose optimal scheduling algorithms for computing resources along with three heuristics. We test the different algorithms as a function of three metrics. The evaluation is performed under a real traffic model and the results highlight the importance of choosing the appropriate scheduling algorithm to increase network performance.

International audience; The growing diversification of mobile services imposes requirements on network performance that are ever more stringent and heterogeneous. Network slicing aligns mobile network operation to this context, by enabling operators to isolate and customize network resources on a per-service basis. A key input for provisioning resources to slices is real-time information about the traffic demands generated by individual services. Acquiring such knowledge is however challenging, as legacy approaches based on in-depth inspection of traffic streams have high computational costs, which inflate with the widening adoption of encryption over data and control traffic. In this paper, we present a new approach to service-level demand estimation for slicing, which hinges on decomposition, i.e., the inference of per-service demands from traffic aggregates. By operating on total traffic volumes only, our approach overcomes the complexity and limitations of legacy traffic classification techniques , and provides a suitable input to recent 'Network Slice as a Service' (NSaaS) models. We implement decomposition through Microscope, a novel framework that uses deep learning to infer individual service demands from complex spatiotemporal features hidden in traffic aggregates. Microscope (i) transforms traffic data collected in irregular radio access deployments in a format suitable for convolutional learning, and (ii) can accommodate a variety of neural network architectures, including original 3D Deformable Convolutional Neural Networks (3D-DefCNNs) that we explicitly design for decomposition. Experiments with measurement data collected in an operational network demonstrate that Microscope accurately estimates per-service traffic demands with relative errors below 1.2%. Further, tests in practical NSaaS management use cases show that resource allocations informed by decomposition yield affordable costs for the mobile network operator.

In this paper, we study the performance of packet scheduling schemes for Ultra-Reliable Low-Latency Communications (URLLC) services. In order to ensure a low queuing time, we exploit the redundant coverage in many 5G Radio Access Networks (RAN) scenarios, where two frequency layers or Radio Access Technologies (RAT) are integrated. We consider three packet scheduling and redundancy schemes, namely Jointhe- Shortest-Queue (JSQ), systematic Redundancy (RED), and redundancy with Cancellation upon completion (CAN). On the basis of queuing theory results, we develop expressions for the reliability, defined as the probability that the packet is transmitted before some given target delay. We show that RED has a good performance at low load, while JSQ is better when the load increases; CAN outperforms all other schemes. We then show how the obtained results can be used for the dimensioning of needed 5G radio resources and discuss the trade-off between performance and implementation complexity.

The use of batteries as a backup in case of power outages is frequent in telecommunications companies that need to keep their services always active. Also, these batteries can be used when the energy costs more, to reduce the electricity bill, as long as the safety usage rules are respected. A new context of use is to participate in the curtailment market, which is a key financial lever for the transmission grid to balance energy demand and offer. The company would use its batteries if called upon to reduce its energy consumption when the power demand is greater than a country's production, receiving a reward in exchange. Our focus in this work is to optimize over time the use of batteries installed for backup in case of power outages to participate in the curtailment market and also to reduce the energy OPEX for the company. We formally state the related optimization problem and propose a mixed-integer program to address it. Such a model is investigated with tests based on real data of a telecommunications operator in France. Simulation results prove the efficiency of our solving approach and the relevance of using batteries to participate in the curtailment market and also to reduce the energy cost for the company.

International audience; Recently, UAVs or Unnamed Aerial Vehicles have been proposed as flexible aerial support to assist ground vehicles for different applications such as rescue and traffic surveillance missions. UAVs can collect different data information about the road/traffic state usually as aerial photography and videos. The processing of this kind of data consists usually on pattern recognition and video processing which are complex tasks that necessitate powerful computing and energy resources. Unfortunately, the moderate UAV's computational and energy capabilities restrict local data processing. Fortunately, UAVs can leverage the computation resources of the surrounding edge network entities to enhance their computational capabilities. In this paper, we aim to achieve efficient data processing for the data collected by UAVs in the context of UAVs-aided vehicular networks for traffic monitoring missions. For this purpose, we proposed a new system model where UAVs can offload and/or share intensive computation tasks with other nearby network nodes. Then, we used the computation response time, the energy consumed for the computation, the cost of cellular communication and the computation cost as the main system metrics to make any computation offloading/sharing decisions that optimize the system performance. We then modeled the offloading/sharing decision-making problem as a sequential game, where we provide complete proof of the existence of the Nash equilibrium and proposed an algorithm to reach such an equilibrium. The simulation results showed that the proposed game-based model outperforms other approaches by delivering better performance in terms of overall system utility with a data processing efficiency that varies between 43% and 97% depending on the computation approach, and provides a more efficient computation time and energy average.

International audience

International audience; 5G RAN aims to evolve new technologies spanning the Cloud infrastructure, virtualization techniques and Software Defined Network capabilities. Advanced solutions are introduced to split the functions of the Radio Access Network (RAN) between centralized and distributed locations. Such paradigms improve RAN flexibility and reduce the infrastructure deployment cost without impacting the user quality of service. We propose a novel functional split orchestration scheme that aims at minimizing the RAN deployment cost, while considering the requirements of its processing network functions and the capabilities of the Cloud infrastructure. With a fine grained approach on user basis, we show that the proposed solution optimizes both processing and bandwidth resource usage, while minimizing the overall energy consumption compared to i) cell-centric, ii) distributed and iii) centralized Cloud-RAN approaches. Moreover, we evaluate the effectiveness of our proposal in a 5G experimental prototype, based on Open Air Interface (OAI). We show that our solution achieves good performance in terms of total deployment cost and resolution time.

International audience; We study in this paper the coexistence of Ultra Reliable Low Latency Communications (URLLC) and enhanced Mobile Broadband (eMBB) services in unlicensed spectrum for the uplink transmission in a 5G smart-factory scenario. We first model the medium access for both services coexisting in unlicensed spectrum and evaluate their performance metrics: high reliability and stringent delay budget for URLLC and throughput for eMBB. The results show that URLLC requirements cannot be met even for low eMBB traffic load. In order to cope with this, we explore a preemptive approach where URLLC packets are transmitted with high power when their delay approaches the delay constraint, increasing their chance of being successfully received. This approach enhances URLLC performance, with a little impact on that of eMBB. We finally show that with a good calibration of some system parameters can lead to an optimal performance for both services.

In this paper, we consider the Delay Constrained Unsplittable Shortest Path Routing problem which arises in the field of traffic engineering for IP networks. This problem consists, given a directed graph and a set of commodities, to compute a set of routing paths and the associated administrative weights such that each commodity is routed along the unique shortest path between its origin and its destination, according to these weights. We present a compact MILP formulation for the problem, extending the work in [5] along with some valid inequalities to strengthen its linear relaxation. This formulation is used as the bulding block of an iterative approach that we develop to tackle large scale instances. We further propose a dynamic programming algorithm based on a tree decomposition of the graph. To the best of our knowledge, this is the first exact combinatorial algorithm for the problem. Finally, we assess the efficiency of our approaches through a set of experiments on state-of-the-art instances.