Le vieillissement des populations est une réalité incontournable des sociétés modernes, qui observent une augmentation inexorable de l’espérance de vie. L’un des obstacles à l’analyse de la mortalité chez l’humain tient au fait qu’il faut plus de cent ans pour que tous les individus d’une même génération décèdent. C’est pourquoi la biodémographie, un nouveau champ de recherche alliant la démographie et la biologie, se penche sur l’analyse des données de mortalité des primates, tels le lémurien genre Microcebus murinus. Les données pour cette espèce élevée en captivité proviennent principalement des livrets d’enregistrement des entrées et sorties de tous les individus du laboratoire MMDN du département de biologie de l’Université de Montpellier. L’objectif principal de ce mémoire est de comparer les distributions des décès par âge pour le lémurien par sexe, ainsi que pour des populations humaines de différentes époques. L’approche par P-splines utilisée permet de dériver ces distributions à partir des taux de mortalité lissés. Différents indices sont calculés sur la mortalité : l’espérance de vie à l’âge de la maturité sexuelle, l’espérance de vie aux grands âges, l’âge médian au décès, ainsi que l’âge modal (i.e. le plus fréquent) au décès. Nos résultats indiquent que les femelles lémurien genre Microcebus murinus élevées en captivité vivent plus longtemps que les mâles, et ce pour tous les différents indices utilisés, contrairement à ce qui avait été rapporté dans la littérature. Cela est toutefois cohérent avec les différentes hypothèses qui supposent des durées de vies plus longues chez les femelles primates et chez les humaines. De plus, la comparaison de la mortalité chez les lémuriens et les humains montrent que la distribution des décès du lémurien se rapproche des sociétés pré-industrielles européennes. Cette incursion dans la démographie d’une espèce animale contribuera l’avancement de ce tout nouveau champ de recherche qu’est la biodémographie. L’analyse plus approfondie de la longévité de primates à courtes durées de vie permettra de nous d’améliorer nos connaissances à long terme sur les mécanismes biologiques du vieillissement chez l’humain. / The aging of populations is an inescapable reality of modern societies, which observe an inexorable increase in life expectancy. One of the obstacles to the analysis of human mortality is the fact that it takes more than a hundred years for all individuals in a generation to die. For this reason, biodemography, a new field of research combining demography and biology, is looking at the analysis of mortality data from primates, such as the gray mouse lemur (Microcebus murinus). The data for this captive-bred species come mainly from the logbooks recording the entries and exits of all individuals in the MMDN laboratory of the Biology Department of the University de Montpellier. The main objective of this thesis is to compare the distributions of deaths by age for the lemur by sex, as well as for human populations of different ages. The P-splines approach used makes it possible to derive these distributions from smoothed mortality rates. Different indicators are calculated on mortality: life expectancy at the age of sexual maturity, life expectancy at old age, median age of survival, and modal age at death. Our results indicate that female captive-bred gray mouse lemurs live longer than males for all the different indices used, contrary to what has been reported in the literature. This is however consistent with the different hypotheses that assume longer life spans in primate and human females. Moreover, the comparison of mortality in lemurs and humans shows that the distribution of lemur deaths is close to the European pre-industrial societies. This incursion into the demography of an animal species will allow the advancement of this brand-new field of research that is biodemography. A more in-depth analysis of the longevity of short-lived primates will provide us with long-term information on the biological mechanisms of aging in humans.

The main contribution of this thesis is the concept of Fenchel duality with a focus on its application in the field of machine learning problems and image restoration tasks. We formulate a general optimization problem for modeling support vector machine tasks and assign a Fenchel dual problem to it, prove weak and strong duality statements as well as necessary and sufficient optimality conditions for that primal-dual pair. In addition, several special instances of the general optimization problem are derived for different choices of loss functions for both the regression and the classifification task. The convenience of these approaches is demonstrated by numerically solving several problems. We formulate a general nonsmooth optimization problem and assign a Fenchel dual problem to it. It is shown that the optimal objective values of the primal and the dual one coincide and that the primal problem has an optimal solution under certain assumptions. The dual problem turns out to be nonsmooth in general and therefore a regularization is performed twice to obtain an approximate dual problem that can be solved efficiently via a fast gradient algorithm. We show how an approximate optimal and feasible primal solution can be constructed by means of some sequences of proximal points closely related to the dual iterates. Furthermore, we show that the solution will indeed converge to the optimal solution of the primal for arbitrarily small accuracy. Finally, the support vector regression task is obtained to arise as a particular case of the general optimization problem and the theory is specialized to this problem. We calculate several proximal points occurring when using difffferent loss functions as well as for some regularization problems applied in image restoration tasks. Numerical experiments illustrate the applicability of our approach for these types of problems.

This thesis is concerned with the development of novel numerical methods for solving nondifferentiable convex optimization problems in real Hilbert spaces and with the investigation of their asymptotic behavior. To this end, we are also making use of monotone operator theory as some of the provided algorithms are originally designed to solve monotone inclusion problems. After introducing basic notations and preliminary results in convex analysis, we derive two numerical methods based on different smoothing strategies for solving nondifferentiable convex optimization problems. The first approach, known as the double smoothing technique, solves the optimization problem with some given a priori accuracy by applying two regularizations to its conjugate dual problem. A special fast gradient method then solves the regularized dual problem such that an approximate primal solution can be reconstructed from it. The second approach affects the primal optimization problem directly by applying a single regularization to it and is capable of using variable smoothing parameters which lead to a more accurate approximation of the original problem as the iteration counter increases. We then derive and investigate different primal-dual methods in real Hilbert spaces. In general, one considerable advantage of primal-dual algorithms is that they are providing a complete splitting philosophy in that the resolvents, which arise in the iterative process, are only taken separately from each maximally monotone operator occurring in the problem description. We firstly analyze the forward-backward-forward algorithm of Combettes and Pesquet in terms of its convergence rate for the objective of a nondifferentiable convex optimization problem. Additionally, we propose accelerations of this method under the additional assumption that certain monotone operators occurring in the problem formulation are strongly monotone. Subsequently, we derive two Douglas–Rachford type primal-dual methods for solving monotone inclusion problems involving finite sums of linearly composed parallel sum type monotone operators. To prove their asymptotic convergence, we use a common product Hilbert space strategy by reformulating the corresponding inclusion problem reasonably such that the Douglas–Rachford algorithm can be applied to it. Finally, we propose two primal-dual algorithms relying on forward-backward and forward-backward-forward approaches for solving monotone inclusion problems involving parallel sums of linearly composed monotone operators. The last part of this thesis deals with different numerical experiments where we intend to compare our methods against algorithms from the literature. The problems which arise in this part are manifold and they reflect the importance of this field of research as convex optimization problems appear in lots of applications of interest.

Private and public clouds require users to specify requests for resources such as CPU and memory (RAM) to be provisioned for their applications. The values of these requests do not necessarily relate to the application’s run-time requirements, but only help the cloud infrastructure resource manager to map requested virtual resources to physical resources. If an application exceeds these values, it might be throttled or even terminated. Consequently, requested values are often overestimated, resulting in poor resource utilization in the cloud infrastructure. Autoscaling is a technique used to overcome these problems. In this research, we formulated two new predictive CPU autoscaling strategies forKubernetes containerized applications, using time-series analysis, based on Holt-Winters exponential smoothing and long short-term memory (LSTM) artificial recurrent neural networks. The two approaches were analyzed, and their performances were compared to that of the default Kubernetes Vertical Pod Autoscaler (VPA). Efficiency was evaluated in terms of CPU resource wastage, and insufficient CPU percentage and amount for container workloads from Alibaba Cluster Trace 2018, and others. In our experiments, we observed that Kubernetes Vertical Pod Autoscaler (VPA) tended to perform poorly on workloads that periodically change. Our results showed that compared to VPA, predictive methods based on Holt- Winters exponential smoothing (HW) and Long Short-Term Memory (LSTM) can decrease CPU wastage by over 40% while avoiding CPU insufficiency for various CPU workloads. Furthermore, LSTM has been shown to generate stabler predictions compared to that of HW, which allowed for more robust scaling decisions. / Privata och offentliga moln kräver att användare begär mängden CPU och minne (RAM) som ska fördelas till sina applikationer. Mängden resurser är inte nödvändigtvis relaterat till applikationernas körtidskrav, utan är till för att molninfrastrukturresurshanteraren ska kunna kartlägga begärda virtuella resurser till fysiska resurser. Om en applikation överskrider dessa värden kan den saktas ner eller till och med krascha. För att undvika störningar överskattas begärda värden oftast, vilket kan resultera i ineffektiv resursutnyttjande i molninfrastrukturen. Autoskalning är en teknik som används för att överkomma dessa problem. I denna forskning formulerade vi två nya prediktiva CPU autoskalningsstrategier för containeriserade applikationer i Kubernetes, med hjälp av tidsserieanalys baserad på metoderna Holt-Winters exponentiell utjämning och långt korttidsminne (LSTM) återkommande neurala nätverk. De två metoderna analyserades, och deras prestationer jämfördes med Kubernetes Vertical Pod Autoscaler (VPA). Prestation utvärderades genom att observera under- och överutilisering av CPU-resurser, för diverse containerarbetsbelastningar från bl. a. Alibaba Cluster Trace 2018. Vi observerade att Kubernetes Vertical Pod Autoscaler (VPA) i våra experiment tenderade att prestera dåligt på arbetsbelastningar som förändras periodvist. Våra resultat visar att jämfört med VPA kan prediktiva metoder baserade på Holt-Winters exponentiell utjämning (HW) och långt korttidsminne (LSTM) minska överflödig CPU-användning med över 40 % samtidigt som de undviker CPU-brist för olika arbetsbelastningar. Ytterligare visade sig LSTM generera stabilare prediktioner jämfört med HW, vilket ledde till mer robusta autoskalningsbeslut.

Denna studie undersöker användningen av maskininlärningsmodeller för att predicera arbetskraftstrender inom hälso- och sjukvården i Sverige. Med hjälp av en linjär regressionmodell, en Gradient Boosting Regressor-modell och en Exponential Smoothing-modell syftar forskningen för detta arbete till att ge viktiga insikter för underlaget till makroekonomiska överväganden och att ge en djupare förståelse av Beveridge-kurvan i ett sammanhang relaterat till hälso- och sjukvårdssektorn. Trots vissa utmaningar med datan är målet att förbättra noggrannheten och effektiviteten i beslutsfattandet rörande arbetsmarknaden. Resultaten av denna studie visar maskininlärningspotentialen i predicering i ett ekonomiskt sammanhang, även om inneboende begränsningar och etiska överväganden beaktas. / This study examines the use of machine learning models to predict workforce trends in the healthcare sector in Sweden. Using a Linear Regression model, a Gradient Boosting Regressor model, and an Exponential Smoothing model the research aims to grant needed insight for the basis of macroeconomic considerations and to give a deeper understanding of the Beveridge Curve in the healthcare sector’s context. Despite some challenges with data, the goal is to improve the accuracy and efficiency of the policy-making around the labor market. The results of this study demonstrates the machine learning potential in the forecasting within an economic context, although inherent limitations and ethical considerations are considered.

The simulation of physical systems in the early stages of conceptual designs has shown to be a key factor for adequate decision making and avoiding big and expensive issues downstream in engineering projects. In the case of aircraft cabin design, taking into account the thermal comfort of the passengers as well as the proper air circulation and renovation can make this difference. However, current numerical fluid simulations (CFD) are too computationally expensive for integrating them in early design stages where extensive comparative studies have to be performed. Instead, Zonal Models (ZM) appear to be a fast-computation approach that can provide coarse simulations for aircraft cabin flows. In this thesis, a Zonal Model solver is developed as well as a geometry-definition and meshing framework, both in Matlab®, for performing coarse, flexible and computationally cheap flow simulations of user-defined cabin designs. On one hand, this solver consists of a Fractional Step approach for coarse unstructured bi-dimensional meshes. On the other, the cabin geometry can be introduced by hand for simple shapes, but also with Computational Aided Design tools (CAD) for more complex designs. Additionally, it can be chosen to generate the meshes from scratch or morph them from previously generated ones. / <p>The presentation was online</p>

This project investigated the implementation and evaluation of various collision-free path planning algorithms for the Saab Seaeye eM1-7 6-DOF Electric Manipulator (eManip). The primary goal was to enhance the autonomous performance of the eManip by integrating efficient path planning methodologies, ultimately ensuring the avoidance of collisions and manipulator singularities during underwater operations. Key algorithms examined included the Rapidly-exploring Random Trees (RRT) algorithm and its enhanced variants. Through simulation tests in MATLAB and Gazebo, metrics such as planning time, path length, and the number of explored nodes were evaluated. The results highlighted the robustness of Goal-biased and Bidirectional RRT* (Gb-Bd-RRT*), which consistently performed well across various environments. The research also highlighted the correlation between algorithm effectiveness and specific task attributes, emphasizing their adaptability to complex environments. This research contributes valuable insights into the effectiveness of path planning algorithms, informing the selection and integration of viable strategies for 6-DOF robotic manipulators.

[ES] La necesidad de mejorar el consumo de combustible y las emisiones de los sistemas propulsivos de automoción en condiciones reales de conducción es la base de esta tesis. Para ello, se exploran dos ejes: En primer lugar, el control de los sistemas de propulsión. El estado del arte de control en los sistemas propulsivos de automoción se basa en gran medida en el uso de técnicas de optimización que buscan las leyes de control que minimizan una función de coste en un conjunto de condiciones de operación denidas a priori. Estas leyes se almacenan en las ECUs de producción en forma de mapas de calibración de los diferentes actuadores del motor. Las incertidumbres asociadas al conjunto limitado de condiciones en el proceso de calibración dan lugar a un funcionamiento subóptimo del sistema de propulsión en condiciones de conducción real. Por lo tanto, en este trabajo se proponen métodos de control adaptativo que optimicen la gestión de la planta propulsiva a las condiciones esperadas de funcionamiento para un usuario y un caso determinado en lugar de a un conjunto genérico de condiciones. El segundo eje se reere a optimizar, en lugar de los parámetros de control del sistema propulsivo, la demanda de potencia de este, introduciendo al propio conductor en el bucle de control, sugiriéndole las acciones a tomar. En particular, este segundo eje se reere al control de la velocidad del vehículo (conocido popularmente como Eco-Driving en la literatura) en condiciones reales de conducción. Se proponen sistemas de aviso en tiempo real al conductor acerca de la velocidad óptima para minimizar el consumo del vehículo. Los métodos de control desarrollados para cada aplicación se describen en detalle en la tesis y se muestran ensayos experimentales de validación en los casos de estudio diseñados. Ambos ejes representan un problema de control óptimo, denido por un sistema dinámico, unas restricciones a cumplir y un coste a minimizar, en este sentido las herramientas desarrolladas en la tesis son comunes a los dos ejes: Un modelo de vehículo, una herramienta de predicción del ciclo de conducción y métodos de control óptimo (Programación Dinámica, Principio Mínimo de Pontryagin y Estrategia de Consumo Equivalente Mínimo). Dependiendo de la aplicación, los métodos desarrollados se implementaron en varios entornos experimentales: un motor térmico en sala de ensayos simulando el resto del vehículo, incluyendo el resto del sistema de propulsión híbrido y en un vehículo real. Los resultados muestran mejoras signicativas en el rendimiento del sistema de propulsión en términos de ahorro de combustible y emisiones en comparación con los métodos empleados en el estado del arte actual. / [CA] La necessitat de millorar el consum de combustible i les emissions dels sistemes propulsius d'automoció en condicions reals de conducció és la base d'aquesta tesi. Per a això, s'exploren dos eixos: En primer lloc, el control dels sistemes de propulsió. L'estat de l'art de control en els sistemes propulsius d'automoció es basa en gran manera en l'ús de tècniques d'optimització que busquen les lleis de control que minimitzen una funció de cost en un conjunt de condicions d'operació denides a priori. Aquestes lleis s'emmagatzemen en les Ecus de producció en forma de mapes de calibratge dels diferents actuadors del motor. Les incerteses associades al conjunt limitat de condicions en el procés de calibratge donen lloc a un funcionament subòptim del sistema de propulsió en condicions de conducció real. Per tant, en aquest treball es proposen mètodes de control adaptatiu que optimitzen la gestió de la planta propulsiva a les condicions esperades de funcionament per a un usuari i un cas determinat en lloc d'un conjunt genèric de condicions. El segon eix es refereix a optimitzar, en lloc dels paràmetres de control del sistema propulsiu, la demanda de potència d'aquest, introduint al propi conductor en el bucle de control, suggerint-li les accions a prendre. En particular, aquest segon eix es refereix al control de la velocitat del vehicle (conegut popularment com Eco-*Driving en la literatura) en condicions reals de conducció. Es proposen sistemes d'avís en temps real al conductor sobre la velocitat òptima per a minimitzar el consum del vehicle. Els mètodes de control desenvolupats per a cada aplicació es descriuen detalladament en la tesi i es mostren assajos experimentals de validació en els casos d'estudi dissenyats. Tots dos eixos representen un problema de control òptim, denit per un sistema dinàmic, unes restriccions a complir i un cost a minimitzar, en aquest sentit les eines desenvolupades en la tesi són comunes als dos eixos: Un model de vehicle, una eina de predicció del cicle de conducció i mètodes de control òptim (Programació Dinàmica, Principi Mínim de *Pontryagin i Estratègia de Consum Equivalent Mínim). Depenent de l'aplicació, els mètodes desenvolupats es van implementar en diversos entorns experimentals: un motor tèrmic en sala d'assajos simulant la resta del vehicle, incloent la resta del sistema de propulsió híbrid i en un vehicle real. Els resultats mostren millores signicatives en el rendiment del sistema de propulsió en termes d'estalvi de combustible i emissions en comparació amb els mètodes emprats en l'estat de l'art actual. / [EN] The need of improving the real-world fuel consumption and emission of automotive applications is the basis of this thesis. To this end, two verticals are explored: First is the online control of the powertrain systems. In state-of-the-art Optimal Control techniques (such as Dyanmic Programming, Pontryagins Minimum Principle, etc...) are extensively used to formulate the optimal control laws. These laws are stored in the production ECUs in the form of feedforward calibration maps. The unaccounted uncertainities related to the real-world during the powertrain calibration result in suboptimal operations of the powertrain in actual driving. Therefore, adaptive control methods are proposed in this work which, optimise the energy management of the conventional and the HEV powertrain control on real driving mission. The second vertical is regarding the vehicle speed control (popularly known as Eco-Driving in the literature) methods in real driving condition. In particular, speed advisory systems are proposed for real time application on a vehicle. The control methods developed for each application are described in details with their verication and validation on the designed case studies. Apart from the developed control methods, there are three tools that were developed and used at various stages of this thesis: A vehicle model, A driving cycle prediction tool and optimal control methods (dynamic programming, PMP and ECMS). Depending on the application, the developed methods were implemented on the Hardware-In-Loop Internal Combustion Engine testing setup or on a real vehicle. The results show signicant improvements in the performance of the powertrain in terms of fuel economy and emissions in comparison to the state-of-the-art methods. / Pandey, V. (2021). Online Control of Automotive systems for improved Real-World Performance [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/173716

Les Energies Marines Renouvelables (EMR) se développent aujourd’hui très vite tant au niveau de la recherche amont que de la R&D, et même des premiers démonstrateurs à la mer. Parmi ces EMR, l'énergie des vagues présente un potentiel particulièrement intéressant. Avec une ressource annuelle brute moyenne estimée à 40 kW/m au large de la côte atlantique, le littoral français est plutôt bien exposé. Mais l’exploitation à grande échelle de cette énergie renouvelable ne sera réalisable et pertinente qu'à condition d'une bonne intégration au réseau électrique (qualité) ainsi que d'une gestion et d'un dimensionnement optimisé au sens du coût sur cycle de vie. Une première solution de génération tout électrique pour un houlogénérateur a d’abord été évaluée dans le cadre de la thèse de Marie RUELLAN menée sur le site de Bretagne du laboratoire SATIE (ENS de Cachan). Ces travaux ont mis en évidence le potentiel de viabilité économique de cette chaîne de conversion et ont permis de poser la question du dimensionnement de l’ensemble convertisseur-machine et de soulever les problèmes associés à la qualité de l’énergie produite. Puis une seconde thèse a été menée par Judicaël AUBRY dans la même équipe de recherche. Elle a consisté, entre autres, en l’étude d’une première solution de traitement des fluctuations de la puissance basée sur un système de stockage par supercondensateurs. Une méthodologie de dimensionnement de l’ensemble convertisseur-machine et de gestion de l’énergie stockée fut également élaborée, mais en découplant le dimensionnement et la gestion de la production d’énergie et de ceux de son système de stockage. Le doctorant devra donc : 1. S’approprier les travaux antérieurs réalisés dans le domaine de la récupération de l’énergie des vagues ainsi que les modèles hydrodynamiques et mécaniques réalisés par notre partenaire : le LHEEA de l’Ecole Centrale de Nantes - 2. Résoudre le problème du couplage entre dimensionnement/gestion de la chaîne de conversion et dimensionnement/gestion du système de stockage. 3. Participer à la réalisation d’un banc test à échelle réduite de la chaine électrique et valider expérimentalement les modèles énergétiques du stockage et des convertisseurs statiques associés - 4. Proposer une méthodologie de dimensionnement de la chaine électrique intégrant le stockage et les lois de contrôle préalablement élaborées 5. Déterminer les gains en termes de capacités de stockage obtenus grâce à la mutualisation de la production (parc de machines) et évaluer l’intérêt d’un stockage centralisé - 6. Analyser l’impact sur le réseau d’une production houlogénérée selon divers scenarii, modèles et outils développés par tous les partenaires dans le cadre du projet QUALIPHE. L’exemple traité sera celui de l’Ile d’Yeu (en collaboration avec le SyDEV. / The work of this PhD thesis deals with the minimization of the per-kWh cost of direct-drive wave energy converter, crucial to the economic feasibility of this technology. Despite the simplicity of such a chain (that should provide a better reliability compared to indirect chain), the conversion principle uses an oscillating system (a heaving buoy for example) that induces significant power fluctuations on the production. Without precautions, such fluctuations can lead to: a low global efficiency, an accelerated aging of the fragile electrical components and a failure to respect power quality constraints. To solve these issues, we firstly study the optimization of the direct drive wave energy converter control in order to increase the global energy efficiency (from wave to grid), considering conversion losses and the limit s from the sizing of an electrical chain (maximum force and power). The results point out the effect of the prediction horizon or the mechanical energy into the objective function. Production profiles allow the study of the flicker constraint (due to grid voltage fluctuations) linked notably to the grid characteristics at the connection point. Other models have also been developed to quantify the aging of the most fragile and highly stressed components, namely the energy storage system used for power smoothing (with super capacitors or electrochemical batteries Li-ion) and power semiconductors.Finally, these aging models are used to optimize key design parameters using life-cycle analysis. Moreover, the sizing of the storage system is co-optimized with the smoothing management.

The purpose of this work is to introduce explicit finite element method (FEM) and to familiarize with commercial software tools witch are capable to perform simulations. The technological conditions and the scope of application of roller burnishing are described in subsequent part. The simulation model of roller burnishing was created. Software ANSYS LS-DYNA was used to make computations. The results of simulation and technical and economical benefits of roller burnishing are presented in the conclusion.