The defense will be streamed live here: link

Title

Étalonnage in situ de l’instrumentation bas coût pour la mesure de grandeurs ambiantes : méthode d’évaluation des algorithmes et diagnostic des dérives

Abstract

In various fields going from agriculture to public health, ambient quantities have to be monitored in indoors or outdoors areas. For example, temperature, air pollutants, water pollutants, noise and so on have to be tracked. To better understand these various phenomena, an increase of the density of measuring instruments is currently necessary. For instance, this would help to analyse the effective exposure of people to nuisances such as air pollutants. The massive deployment of sensors in the environment is made possible by the decreasing costs of measuring systems, mainly using sensitive elements based on micro or nano technologies. The drawback of this type of instrumentation is a low quality of measurement, consequently lowering the confidence in produced data and/or a drastic increase of the instrumentation costs due to necessary recalibration procedures or periodical replacement of sensors. There are multiple algorithms in the literature offering the possibility to perform the calibration of measuring instruments while leaving them deployed in the field, called in situ calibration techniques.

The objective of this thesis is to contribute to the research effort on the improvement of data quality for low-cost measuring instruments through their in situ calibration. In particular, we aim at 1) facilitating the identification of existing in situ calibration strategies applicable to a sensor network depending on its properties and the characteristics of its instruments; 2) helping to choose the most suitable algorithm depending on the sensor network and its context of deployment; 3) improving the efficiency of in situ calibration strategies through the diagnosis of instruments that have drifted in a sensor network. Three main contributions are made in this work. First, a unified terminology is proposed to classify the existing works on in situ calibration. The review carried out based on this taxonomy showed there are numerous contributions on the subject, covering a wide variety of cases. Nevertheless, the classification of the existing works in terms of performances was difficult as there is no reference case study for the evaluation of these algorithms. Therefore in a second step, a framework for the simulation of sensors networks is introduced. It is aimed at evaluating in situ calibration algorithms. A detailed case study is provided across the evaluation of in situ calibration algorithms for blind static sensor networks. An analysis of the influence of the parameters and of the metrics used to derive the results is also carried out. As the results are case specific, and as most of the algorithms recalibrate instruments without evaluating first if they actually need it, an identification tool enabling to determine the instruments that are actually faulty in terms of drift would be valuable. Consequently, the third contribution of this thesis is a diagnosis algorithm targeting drift faults in sensor networks without making any assumption on the kind of sensor network at stake. Based on the concept of rendez-vous, the algorithm allows to identify faulty instruments as long as one instrument at least can be assumed as non-faulty in the sensor network. Across the investigation of the results of a case study, we propose several means to reduce false results and guidelines to adjust the parameters of the algorithm. Finally, we show that the proposed diagnosis approach, combined with a simple calibration technique, enables to improve the quality of the measurement results. Thus, the diagnosis algorithm opens new perspectives on in situ calibration.

Jury

• M. Jean-Luc Bertrand-Krajewski, Professeur des Universités, Université de Lyon, INSA Lyon, DEEP (Rapporteur)
• M. Romain Rouvoy, Professeur des Universités, Université de Lille, Spirals (Rapporteur)
• Mme Nathalie Redon, Maître de conférences, IMT Lille Douai, SAGE (Examinatrice)
• M. Gilles Roussel, Professeur des Universités, Université du Littoral Côte d’Opale, LISIC (Examinateur)
• Mme Bérengère Lebental, Directrice de recherche, Institut Polytechnique de Paris, École Polytechnique, LPICM (Directrice de thèse)
• M. Hervé Rivano, Univeristé de Lyon, INSA Lyon, CITI Lab (Co-directeur de thèse)
• M. Éric Peirano, Directeur général adjoint en charge de la R&D, Efficacity (Invité)
• M. Matthieu Puigt, Maître de conférences, Université du Littoral Côte d’Opale, LISIC (Invité)

The defense will take place in amphitheater Chappe and you are all welcome to attend as long as there are enough place (35 persons max).
It will be also streamed live on youtube at: https://youtu.be/fvCz5ZXYN_I

Title

Deep Multi-Agent Reinforcement Learning for Dynamic and Stochastic Vehicle Routing Problems

Abstract

Routing delivery vehicles in dynamic and uncertain environments like dense city centers is a challenging task, which requires robustness and flexibility. Such logistic problems are usually formalized as Dynamic and Stochastic Vehicle Routing Problems (DS-VRPs) with a variety of additional operational constraints, such as Capacitated vehicles or Time Windows (DS-CVRPTWs). Main heuristic approaches to dynamic and stochastic problems simply consist in restarting the optimization process on a frozen (static and deterministic) version of the problem given the new information. Instead, Reinforcement Learning (RL) offers models such as Markov Decision Processes (MDPs) which naturally describe the evolution of stochastic and dynamic systems. Their application to more complex problems has been facilitated by recent progresses in Deep Neural Networks, which can learn to represent a large class of functions in high dimensional spaces to approximate solutions with high performances. Finding a compact and sufficiently expressive state representation is the key challenge in applying RL to VRPs. Recent work exploring this novel approach demonstrated the capabilities of Attention Mechanisms to represent sets of customers and learn policies generalizing to different configurations of customers. However, all existing work using DNNs reframe the VRP as a single-vehicle problem and cannot provide online decision rules for a fleet of vehicles.
In this thesis, we study how to apply Deep RL methods to rich DS-VRPs as multi-agent systems. We first explore the class of policy-based approaches in Multi-Agent RL and Actor-Critic methods for Decentralized, Partially Observable MDPs in the Centralized Training for Decentralized Control (CTDC) paradigm. To address DS-VRPs, we then introduce a new sequential multi-agent model we call sMMDP. This fully observable model is designed to capture the fact that consequences of decisions can be predicted in isolation. Afterwards, we use it to model a rich DS-VRP and propose a new modular policy network to represent the state of the customers and the vehicles in this new model, called MARDAM. It provides online decision rules adapted to the information contained in the state and takes advantage of the structural properties of the model. Finally, we develop a set of artificial benchmarks to evaluate the flexibility, the robustness and the generalization capabilities of MARDAM. We report promising results in the dynamic and stochastic case, which demonstrate the capacity of MARDAM to address varying scenarios with no re-optimization, adapting to new customers and unexpected delays caused by stochastic travel times. We also implement an additional benchmark based on micro-traffic simulation to better capture the dynamics of a real city and its road infrastructures. We report preliminary results as a proof of concept that MARDAM can learn to represent different scenarios, handle varying traffic conditions, and customers configurations.

Jury

• François Charpillet, Research Director at INRIA Nancy Grand Est, Reviewer
• Romain Billot, Professor at IMT Atlantique, Reviewer
• René Mandiau, Professor at Université Polytechnique des Hauts de France, Examiner
• Aurélie Beynier, Associate Professor at Sorbonne Université, Examiner
• Christian Wolf, Associate Professor at INSA de Lyon, Examiner
• Olivier Simonin, Professeur à l’INSA de Lyon, Thesis director
• Jilles Dibangoye, Associate Professor at INSA de Lyon, Co-supervisor
• Laëtitia Matignon, Associate Professor at Université Lyon 1, Co-supervisor
• Florian Pereyron, Research Engineer at Volvo Group, Co-supervisor

Title

Enhancing Transparency and Consent in the IoT

Abstract

In an increasingly connected world, the Internet permeates every aspect of our lives. The number of devices connected to the global network is rising, with prospects foreseeing 75 billions devices by 2025. The Internet of Things envisioned twenty years ago is now materializing at a fast pace, but this growth is not without consequence. The increasing number of devices raises the possibility of surveillance to a level never seen before. A major step has been taken in 2018 to safeguard privacy, with the introduction of the General Data Protection Regulation (GDPR) in the European Union. It imposes obligations to data controllers on the content of information about personal data collection and processing, and on the means of communication of this information to data subjects. This information is all the more important that it is required for consent, which is one of the legal grounds to process personal data. However, the Internet of Things can pose difficulties to implement lawful information communication and consent management. The tension between the requirements of the GDPR for information and consent and the Internet of Things cannot be easily solved. It is however possible. The goal of this thesis is to provide a solution for information communication and consent management in the Internet of Things from a technological point of view. To do so, we introduce a generic framework for information communication and consent management in the Internet of Things. This framework is composed of a protocol to communicate and negotiate privacy policies, requirements to present information and interact with data subjects, and requirements over the provability of consent. We support the feasibility of this generic framework with different options of implementation. The communication of information and consent through privacy policies can be implemented in two different manners: directly and indirectly. We then propose ways to implement the presentation of information and the provability of consent. A design space is also provided for systems designers, as a guide for choosing between the direct and the indirect implementations. Finally, we present fully functioning prototypes devised to demonstrate the feasibility of the framework’s implementations. We illustrate how the indirect implementation of the framework can be developed as a collaborative website named Map of Things. We then sketch the direct implementation combined with the agent presenting information to data subjects under the mobile application CoIoT.

Jury

• Patricia Serrano Alvarado, Maître de conférences HDR à l’Université de Nantes, rapporteur
• Gerardo Schneider, Professor at the University of Gothenburg, rapporteur
• Félicien Vallet, Docteur ingénieur au sein du service de l’expertise technologique de la CNIL, examinateur
• Hervé Rivano, Professeur des universités à l’Insa de Lyon, examinateur
• Daniel Le Métayer, Directeur de recherche à Inria, Directeur de thèse
• Claude Castelluccia, Directeur de recherche à Inria, co-Directeur de thèse

The defense will be available at https://join.skype.com/dbnmuw06rBou

Title

Medium Access Control Layer for Dedicated IoT Networks

Abstract

Les réseaux dédiés pour l’Internet des Objets sont apparus avec la promesse de connecter des milliers de nœuds, voire plus, à une seule station de base dans une topologie en étoile. Cette nouvelle logique représente un changement fondamental dans la façon de penser les réseaux, après des décennies pendant lesquelles les travaux de recherche se sont focalisés sur les réseaux multi-sauts.
Les réseaux pour l’Internet des Objets se caractérisent par la longue portée des transmissions, la vaste couverture géographique, une faible consommation d’énergie et un bas coût de mise en place. Cela a rendu nécessaire des adaptations à tous les niveaux protocolaires afin de satisfaire les besoins de ces réseaux. Plusieurs acteurs sont en concurrence sur le marché de l’Internet des Objets, essayant chacun d’établir la solution la plus efficiente. Ces acteurs se sont concentrés sur la modification de la couche physique, soit au niveau de la partie matérielle, soit par la proposition de nouvelles techniques de modulation. Toutefois, en ce qui concerne la solution de contrôle d’accès au canal (connue sous le nom de couche MAC), toutes les solutions proposées par ces acteurs se fondent sur des approches classiques, tel que Aloha et CSMA.
L’objectif de cette thèse est de proposer une solution MAC dynamique pour les réseaux dédiés à l’Internet des Objets. La solution proposée a la capacité de s’adapter aux conditions du réseau. Cette solution est basée sur un algorithme d’apprentissage automatique, qui apprend de l’historique du réseau afin d’établir la relation entre les conditions du réseau, les paramètres de la couche MAC et les performances du réseau en termes de fiabilité et de consommation d’énergie. La solution possède également l’originalité de faire coexister des nœuds utilisant de différentes configurations MAC au sein du même réseau. Les résultats de simulations ont montré qu’une solution MAC basée sur l’apprentissage automatique pourrait tirer profit des avantages des différents protocoles MAC classiques. Les résultats montrent aussi qu’une solution MAC cognitive offre toujours le meilleur compromis entre fiabilité et consommation d’énergie, tout en prenant en compte l’équité entre les nœuds du réseau. La solution MAC cognitive testée pour des réseaux à haute densité a prouvé des bonnes propriétés de passage à l’échelle par rapport aux protocoles MACs classiques, ce qui constitue un autre atout important de notre solution.

Jury

• M. Antoine GALLAIS, Professeur Université Polytechnique Hauts de France, Rapporteur
• M. Congduc PHAM, Professeur Université de Pau et des Pays de l’Adour, Rapporteur
• Mme. Nancy EL RACHKIDY, Maître de Conférences Université Clermont Auvergne, Examinateur
• M. Mickael MAMAN, Ingénieur de Recherche CEA LETI, Examinateur
• M. Hervé RIVANO, Professeur INSA Lyon, Directeur de thèse
• M. Razvan STANICA, Maître de Conférences INSA Lyon, Co-directeur de thèse

The defense will take place at the University of Yaoundé I and will be available at https://join.skype.com/EVc3J1aGhASf

Title

Autonomous Wireless Sensor Network Architecture for Vehicular traffic monitoring at an Intersection

Abstract

In many countries, because of the limited financial budget, the growth of road infrastructures is low compared to the growth of population and the number of vehicles in urban areas. Such a context does not make the task easy for authorities in charge of the management of transportation systems. The introduction of information and communication technologies (ICT) allows to better address these issues. Vehicular traffic management at intersections has an impact on the traffic jam observed in the whole city. In this thesis, our goal is to propose a low-cost, lightweight and autonomous Wireless Sensors Network (WSN) architecture for vehicular traffic monitoring, especially at an intersection. Vehicular traffic data collected can be used, for instance, for intelligent traffic lights management. In the WSN architecture proposed in the literature for vehicular traffic monitoring, underground sensors are used. In terms of network communication, these architectures are not realistic. Nowadays, surface-mounted sensors are proposed by manufacturers.

The first contribution of this thesis is an experimental characterization of wireless links in a WSN with sensors deployed at the ground level. We evaluate the impact of several parameters like the proximity of the ground surface, the communication frequency and the message size on the link quality. Results show a poor link quality at ground level. Based on the conclusions of the experiments, the second contribution of this thesis is WARIM, a new WSN architecture for vehicular traffic monitoring at an intersection. In WARIM, the sensors deployed on a lane form a multi-hop WSN with a linear topology (LWSN). In this network, all the data are forwarded toward the sink. In a network with such properties, the computation and communication requirements are highest in the neighborhood of the sink. Thus, the third contribution of this thesis is a virtual nodes-based and energy efficient sensors deployment strategy for LWSN. Compared to a uniform deployment, this deployment improves the network lifetime by 40%. In our intersection monitoring application, it is important to correlate the messages generated by a sensor to its position with respect to the intersection. Therefore,the fourth contribution of this thesis is, a centroid-based algorithm for sensors ranking in a LWSN. We evaluate the performance of this algorithm considering a realistic channel model, a uniform deployment, as well as the virtual nodes based-deployment proposed in this thesis. Finally, putting all our contributions together, simulations show that WARIM can be used for reliable and real-time vehicular traffic monitoring at an intersection.

Jury

• Marcel FOUDA, Professor, Université de Yaoundé I, President
• Thomas DJOTIO, Associate Professor, Université de Yaoundé I, Reviewer
• Nathalie MITTON, Research Director, INRIA, Reviewer
• Bernard TOURANCHEAU, Professor, Université Grenoble Alpes, Reviewer
• André-Luc BEYLOT, Professor, ENSEEIHT Toulouse, Examinator
• René NDOUNDAM, Associate Professor, Université de Yaoundé I, Examinator
• Razvan STANICA, HDR, INSA Lyon, Examinator
• Maurice TCHUENTE, Professor, Université de Yaoundé I, Co-Supervisor
• Fabrice VALOIS, Professor, INSA Lyon, Co-Supervisor

Title

Dynamic Heterogeneous Memory Allocation for embedded devices

Abstract

Reducing energy consumption is a key challenge to the realisation of the Internet of Things. While emerging memory technologies may offer power reduction and high integration density, they come with major drawbacks such as high latency or limited endurance. As a result, system designers tend to juxtapose several memory technologies on the same chip. We aim to provide the embedded application programmer with a transparent software mechanism to leverage this memory heterogeneity. This work studies the interaction between dynamic memory allocation and memory heterogeneity. We provide cycle accurate simulation of embedded platforms with various memory technologies and we show that different dynamic allocation strategies have a major impact on performance. We demonstrates that interesting performance gains can be achieved even for a low fraction of memory using low latency technology, but only with a clever placement strategy between memory banks. We propose an efficient strategy based on application profiling in our simulator.

Jury

• Olivier Sentieys, Professeur des Universités, Université de Rennes – Examinateur
• Cécile Belleudy, Maitre de Conférence HDR, Université de Nice Sophia Antipolis – Rapporteure
• Lionel Torres, Professeur des Universités, Université de Montpellier – Rapporteur
• Guillaume Salagnac, Maitre de Conférences, INSA de Lyon – Examinateur, Encadrant
• Tanguy Risset, Professeur des Universités, INSA de Lyon – Examinateur, Co-directeur de thèse
• Matthieu Moy, Maitre de Conférences HDR, Université Claude Bernard Lyon 1 – Examinateur, Co-directeur de thèse