The conventional Mini and Large scale Unmanned Aerial Vehicle systems span anywhere from approximately 12 inches to 12 feet; endowing them with larger propulsion systems, batteries/fuel-tanks, which in turn provide ample power reserves for long-endurance flights, powerful actuators, on-board avionics, wireless telemetry etc. The limitations thus imposed become apparent when shifting to Micro Aerial Vehicles (MAVs) and trying to equip them with equal or near-equal flight endurance, processing, sensing and communication capabilities, as their larger scale cousins. The conventional MAV as outlined by The Defense Advanced Research Projects Agency (DARPA) is a vehicle that can have a maximum dimension of 6 inches and weighs no more than 100 grams. Under these tight constraints, the footprint, weight and power reserves available to on-board avionics and actuators is drastically reduced; the flight time and payload capability of MAVs take a massive plummet in keeping with these stringent size constraints. However, the demand for micro flying robots is increasing rapidly. The applications that have emerged over the years for MAVs include search&rescue operations for trapped victims in natural disaster succumbed urban areas; search&reconnaissance in biological, radiation, natural disaster/hazard succumbed/prone areas; patrolling&securing home/office/building premises/urban areas. VTOL capable rotary and fixed wing flying vehicles do not scale down to micro sized levels, owing to the severe loss in aerodynamic efficiency associated with low Reynolds number physics on conventional airfoils; whereas, present state of the art in flapping wing designs lack in one or more of the minimum qualities required from an MAV: Appreciable flight time, appreciable payload capacity for on-board sensors/telemetry and 6DoF hovering/VTOL performance. This PhD. work is directed towards overcoming these limitations. Firstly, this PhD thesis presents the advent of a novel Quad-Wing MAV configuration (called the QV). The Four-Wing configuration is capable of performing all 6DoF flight maneuvers including VTOL. The thesis presents the design, conception, simulation study and finally hardware design/development of the MAV. Secondly, this PhD thesis proves and demonstrates significant improvement in on-board Energy-Harvesting resulting in increased flight times and payload capacities of the order of even 200%-400% and more. Thirdly, this PhD thesis defines a new actuation principle called, Fixed Frequency, Variable Amplitude (FiFVA). It is demonstrated that by the use of passive elastic members on wing joints, a further significant increase in energy efficiency and consequently reduction in input power requirements is observed. An actuation efficiency increase of over 100% in many cases is possible. The natural evolution of actuation development led to invention of two novel actuation systems to illustrate the FiFVA actuation principle and consequently show energy savings and flapping efficiency improvement. Lastly, but not in the least, the PhD thesis presents supplementary work in the design, development of two novel Micro Architecture and Control (MARC) avionics platforms (autopilots) for the application of demonstrating flight control and communication capability on-board the Four-Wing Flapping prototype. The design of a novel passive feathering mechanism aimed to improve lift/thrust performance of flapping motion is also presented.

L’alimentation en énergie des réseaux de capteurs miniaturisés pose une question fondamentale, dans la mesure où leur autonomie est un critère de qualité de plus en plus important pour l’utilisateur. C’est même une question cruciale lorsque ces réseaux visent à assurer une surveillance d’infrastructure (avionique, machines, bâtiments…) ou une surveillance médicale ou environnementale. Les matériaux piézoélectriques permettent d’exploiter l’énergie mécanique inutilisée présente en abondance dans l’environnement (vibrations, déformations liées à des mouvements ou à des flux d’air…). Ils peuvent ainsi contribuer à rendre ces capteurs autonomes en énergie. Sous la forme de nanofils (NF), les matériaux piézoélectriques offrent une sensibilité qui permet d’exploiter des sollicitations mécaniques très faibles. Ils sont également intégrables, éventuellement sur substrat souple.Dans cette thèse nous nous intéressons au potentiel des nanofils de matériaux semi-conducteurs piézoélectriques, tels que ZnO ou les composés III-V, pour la conversion d’énergie mécanique en énergie électrique. Depuis peu, ceux-ci ont fait l’objet d’études relativement nombreuses, avec la réalisation de nanogénérateurs (NG) prometteurs. De nombreuses questions subsistent toutefois avec, par exemple, des contradictions notables entre prédictions théoriques et observations expérimentales.Notre objectif est d’approfondir la compréhension des mécanismes physiques qui définissent la réponse piézoélectrique des NF semi-conducteurs et des NG associés. Le travail expérimental s’appuie sur la fabrication de générateurs de type VING (Vertical Integrated Nano Generators) et sur leur caractérisation. Pour cela, un système de caractérisation électromécanique a été construit pour évaluer les performances des NG réalisés et les effets thermiques sous une force compressive contrôlée. Le module d’Young et les coefficients piézoélectriques effectifs de NF de GaN; GaAs et ZnO et de NF à structure cœur/coquille à base de ZnO ont été évalués également dans un microscope à force atomique (AFM). Les nanofils de ZnO sont obtenus par croissance chimique en milieu liquide sur des substrats rigides (Si) ou flexibles (inox) puis sont intégrés pour former un générateur. La conception du dispositif VING s’est appuyée sur des simulations négligeant l’influence des porteurs libres, comme dans la plupart des études publiées. Nous avons ensuite approfondi le travail théorique en simulant le couplage complet entre les effets mécaniques, piézoélectriques et semi-conducteurs, et en tenant compte cette fois des porteurs libres. La prise en compte du piégeage du niveau de Fermi en surface nous permet de réconcilier observations théoriques et expérimentales. Nous proposons notamment une explication au fait que des effets de taille apparaissent expérimentalement pour des diamètres au moins 10 fois plus grands que les valeurs prévues par simulation ab-initio ou au fait que la réponse du VING est dissymétrique selon que le substrat sur lequel il est intégré est en flexion convexe ou concave. / Energy autonomy in small sensors networks is one of the key quality parameter for end-users. It’s even critical when addressing applications in structures health monitoring (avionics, machines, building…), or in medical or environmental monitoring applications. Piezoelectric materials make it possible to exploit the otherwise wasted mechanical energy which is abundant in our environment (e. g. from vibrations, deformations related to movements or air fluxes). Thus, they can contribute to the energy autonomy of those small sensors. In the form of nanowires (NWs), piezoelectric materials offer a high sensibility allowing very small mechanical deformations to be exploited. They are also easy to integrate, even on flexible substrates.In this PhD thesis, we studied the potential of semiconducting piezoelectric NWs, of ZnO or III-V compounds, for the conversion from mechanical to electrical energy. An increasing number of publications have recently bloomed about these nanostructures and promising nanogenerators (NGs) have been reported. However, many questions are still open with, for instance, contradictions that remain between theoretical predictions and experimental observations.Our objective is to better understand the physical mechanisms which rule the piezoelectric response of semiconducting NWs and of the associated NGs. The experimental work was based on the fabrication of VING (Vertical Integrated Nano Generators) devices and their characterization. An electromechanical characterization set-up was built to evaluate the performance and thermal effects of the fabricated NGs under controlled compressive forces. Atomic Force Microscopy (AFM) was also used to evaluate the Young modulus and the effective piezoelectric coefficients of GaN, GaAs and ZnO NWs, as well as of ZnO-based core/shell NWs. Among them, ZnO NWs were grown using chemical bath deposition over rigid (Si) or flexible (stainless steel) substrates and further integrated to build VING piezoelectric generators. The VING design was based on simulations which neglected the effect of free carriers, as done in most publications to date. This theoretical work was further improved by considering the complete coupling between mechanical, piezoelectric and semiconducting effects, including free carriers. By taking into account the surface Fermi level pinning, we were able to reconcile theoretical and experimental observations. In particular, we propose an explanation to the fact that size effects are experimentally observed for NWs with diameters 10 times higher than expected from ab-initio simulations, or the fact that VING response is non-symmetrical according to whether the substrate on which it is integrated is actuated with a convex or concave bending.

Les progrès technologiques considérables menés depuis ces dernières décennies nous permettent aujourd’hui de disséminer dans notre environnement une nuée de noeuds de capteurs communicants combinant la taille micrométrique et la consommation dérisoire caractéristiques des MEMS avec la puissance des protocoles de communications Internet. L’Internet des Objets, formé par ce réseau de capteurs, possède le potentiel d‘optimiser un grand panel d’applications industrielles et domotiques. Le nouveau défi, que la communauté du Energy Harvesting tente de relever depuis une décennie maintenant, est de rendre ces noeuds de capteurs autonomes en les alimentant grâce à l’énergie perdue dans leur environnement. Dans ces travaux de recherche, nous explorons le potentiel d’un principe thermo-fluidique auto-oscillant pour la génération d’énergie utile à partir d’une source thermique de faible qualité. L’implémentation de cette technologie en tant que machine thermique est étudiée et mène à la caractérisation d’un nouveau cycle thermodynamique caractéristique du SOFHE (Self Oscillating Fluidic Heat Engine). Nous montrons, par une approche phénoménologique, que notre machine thermique se comporte comme un oscillateur mécanique, excité par les évaporations et condensations successives du fluide de travail. Ces changements de phase alternatifs mettent en mouvement une colonne d’eau, jouant le rôle de masse, couplée à une zone de vapeur, jouant le rôle d’un ressort. Une étude de l’influence du couplage du SOFHE avec un transducteur électromécanique, représenté par un oscillateur, mène à la conception et la fabrication d’une spirale piézoélectrique. L’intégration de cette spirale à notre machine thermique forme un générateur thermo-électrique dont les capacités de conversion sont démontrées par la charge d’une capacité. Finalement, la miniaturisation du principe thermo-fluidique SOFHE est rendue possible par la réalisation d’un procédé de fabrication utilisant les techniques MEMS. Des dispositifs miniatures parviennent à exhiber un comportement oscillatoire montrant le potentiel d’intégration de cette technologie. / Abstract : The tremendous technological progresses realized in the last decades allow us to swarm our environment with Wireless Sensors Networks. These WSNs combine the MEMS’ miniature size and low energy consumption, and the powerful Internet communication protocols. This Internet of Things shows great potential in many applications such as industry or housing. For a decade now, the Energy Harvesting community wants to build autonomous WSNs by enabling them to feed off energy wastes. In this work, we study the electricity generation capabilities of a Self-Oscillating Fluidic Heat Engine (SOFHE) and present its characteristic thermodynamic cycle. Our model shows that the SOFHE acts as a mechanical resonator excited by the successive evaporation and condensation processes underwent by the working fluid. These phase changes put a liquid mass in motion, coupled with a vapor spring. The coupling of our heat engine with an electromechanical transducer is studied and leads to a piezoelectric spiral conception and fabrication. Their association forms a thermo-electrical generator able to power and charge an electrical capacitor. Eventually, we demonstrate the miniaturization prospects and integration potential of this SOFHE technology. A micro-fabrication process enables a SOFHE MEMS implementation. Our process includes a deep glass wet etching step as well as a Au-Si eutectic wafer bonding.

Le travail de cette thèse propose une nouvelle méthode de contrôle appelée SSDH (Synchronized Switch Damping and Harvesting) basée sur l’idée de redistribution de l’énergie récupérée pour réduire l’énergie vibratoire d’une structure. De nombreuses recherches ont concerné le contrôle de vibration des structures souples. L’utilisation de l’approche modale pour ce genre de structure présente de nombreux intérêts. Dans le cadre de cette thèse l’idée est de récupérer l’énergie des modes qui ne sont pas contrôlés de façon à améliorer l’effet d’amortissement des modes ciblés par le contrôle sur une même structure. Pour cela, sur la base de la technique semi-active de contrôle, un circuit de contrôle modal a été conçu pour être compatible, via un convertisseur, avec des techniques semi-active de récupération d’énergie qui ont elles même été adaptées en modal. Plusieurs variantes de la méthode SSDH ont été testées en simulation. De façon à estimer l’efficacité du concept, une application sur un modèle expérimental d’une smart structure simple est proposée. Actionneurs et capteurs utilisent des matériaux piézoélectriques qui présentent les effets directs et inverses utiles pour la récupération d’énergie et le contrôle vibratoire. Après optimisation des différents paramètres électromécaniques et électriques, les résultats des simulations menées sous excitations bisinusoidale ou en bruit blanc, montrent que la nouvelle méthode de contrôle autoalimentée SSDH est efficace et robuste. Elle améliore sensiblement l’amortissement produit par les techniques semi-actives modales de base (SSDI) grâce à l’utilisation de l’énergie modale récupérée. / In a context of embedded structures, the next challenge is to develop an efficient, energetically autonomous vibration control technique. Synchronized Switch Damping techniques (SSD) have been demonstrated interesting properties in vibration control with a low power consumption. For compliant or soft smart structures, modal control is a promising way as specific modes can be targetted. This Ph-D work examines a novel energy transfer concept and design of simultaneous energy harvesting and vibration control on the same host structure. The basic idea is that the structure is able to extract modal energy from the chosen modes, and utilize this harvested energy to suppress the target modes via modal control method. We propose here a new technique to enhance the classic SSD circuit due to energy harvesting and energy transfer. Our architecture called Modal Synchronized Switching Damping and Harvesting (Modal SSDH) is composed of a harvesting circuit (Synchronized Switch Harvesting on Inductor SSHI), a Buck-Boost converter and a vibration modal control circuit (SSD). Various alternatives of our SSDH techniques were proposed and simulated. A real smart structure is modeled and used as specific case to test the efficiency of our concept. Piezoelectric sensors and actuators are taken as active transducers, as they develop the direct and inverse effects useful for the energy harvesting and the vibration damping. Optimization are running out and the basic design factors are discussed in terms of energy transfer. Simulations, carried out under bi-harmonic and noise excitation, underline that our new SSDH concept is efficient and robust. Our technique improve the damping effect of semi-active method compared to classic SSD method thanks to the use of harvested modal energy.

Les progrès réalisés durant ces dernières années dans le domaine de la microélectronique et notamment vis-à-vis de l’augmentation exponentielle de la densité d’intégration des composants et des systèmes a participé activement à l’apparition et au développement de systèmes portables communicants de plus en plus performants et polyvalents. La R&D dans les technologies de stockage d’énergie n’a pas suivi cette tendance d’évolution très rapide ; ce qui constitue un handicap majeur dans les évolutions futures des systèmes portables. La transmission d’énergie sans fils sur des distances considérables (plusieurs dizaines de mètres) grâce aux microondes constitue une solution très prometteuse pour pallier aux problèmes d’autonomie dans le cas des systèmes sans fils communicants. De plus, du fait de l’omniprésence des ondes électromagnétiques dans notre environnement avec des niveaux plus ou moins importants, la récupération et l’exploitation de cette énergie libre est également possible. La rectenna (Rectifying Antenna) est le dispositif permettant de capter et de convertir une onde électromagnétique en une tension continue. Plusieurs travaux de thèse axés sur l’étude et l’optimisation de la rectenna ont été réalisés au sein du laboratoire. Ces travaux avaient montré que pour des faibles niveaux de champs les tensions délivrées par la rectenna sont généralement très faibles et inexploitables. Aussi, comme la majorité des micro-sources d’énergie et à cause de son impédance interne, les performances de la rectenna dépendent fortement de sa charge de sortie. Ainsi, le développement d’un système d’interfaçage de la rectenna est nécessaire afin de pallier ces manquements inhérents du convertisseur RF/DC. Ce genre de système d’interfaçage est généralement absent dans la littérature à cause des faibles niveaux de puissance exploités. Par conséquent, la rectenna est très souvent utilisée tel quelle ; ce qui limite fortement le champ applicatif. Dans ce projet de recherche, un système de gestion énergétique de la rectenna complètement autonome a été conçu, développé et optimisé afin de garantir les performances optimales de la rectenna quelques soient les fluctuations de la puissance d’entrée et celles de la charge de sortie. Le circuit d’interfaçage permet également de fournir à la charge des niveaux de tension utilisables. Le système réalisé est basé tout d’abord sur l’utilisation d’un convertisseur DC/DC résonant pouvant fonctionner d’une manière complètement autonome à partir de niveaux très bas de la tension et de la puissance de la source. Ce convertisseur permet donc de garantir l’autonomie du système en éliminant la nécessité d’une source d’énergie auxiliaire. A cause de ses faibles performances énergétiques, ce convertisseur ne sera utilisé que durant la phase de démarrage. L’efficacité du système en termes de rendement énergétique et d’adaptation d’impédance est garantie grâce à l’utilisation d’un convertisseur Flyback fonctionnant dans son régime de conduction discontinu. Ainsi, une adaptation d’impédance très efficace est réalisée entre la rectenna et la charge de sortie. Ce convertisseur principal fonctionnera durant le régime permanent. Les deux convertisseurs ont été optimisés pour des niveaux de tension et de puissance aussi bas que quelques centaines de mV et quelques μW respectivement. Des mesures expérimentales réalisées sur plusieurs prototypes ont démontré le bon fonctionnement et les excellentes performances prédites par la procédure de conception ; ce qui nous permet de valider notre approche. De plus, les performances obtenues se distinguent parfaitement vis-à-vis de l’état de l’art. Enfin, en fonction de l’application désirée, plusieurs synoptiques d’association des deux structures sont proposés. Ceci inclut également la gestion énergétique de la charge de sortie. / Latest advancements in microelectronic technologies and especially with the exponential increase of components and devices integration density have yield novel high technology and polyvalent portable systems. Such polyvalent communication devices need more and more available energy. Nonetheless, research in energy storage technology did not evolve with a similar speed. This constitutes a substantial handicap for the future evolution of portable devices. Wireless energy transfer through large distances such as tens of meters using microwaves is a very promising solution in order to deal with the autonomy problem in portable devices. In addition, since electromagnetic waves are ubiquitous in our environment, harvesting and using this free and available energy is also possible. Rectenna (Rectifying Antenna) is the device that allows to collect and to convert an electromagnetic wave into DC power. Several thesis research projects focusing on studying and optimizing the rectenna was carried-out into the Ampere laboratory. It has been shown that for a low level of the electromagnetic field the voltage provided by the rectenna is ultra-low and thus impractical. Further, as it is the case for the majority of energy harvesting micro-sources, the performances of the rectenna depend highly with the loading conditions. So, the development of an interfacing circuit for the rectenna is a necessary task in order to relieve the RF/DC converter inherent flaws. As it is pointed out into the literature, such power management circuit is in most cases absent due to the ultra-low power levels. In most cases, the rectenna is used as it; which reduces strongly the applications area. Within this research project, an ultra-low power and fully-autonomous power management system dedicated to rectennas was developed and optimized. It allows to guarantee highest performances of the rectenna whatever are the fluctuation of the input power level and the output load conditions. In addition, this power management system allows to provide a conventional voltage level to the load. The first part of the developed system is composed by a resonant DC/DC converter which plays the role of start-up circuit. In this case, no external energy source is required even with low voltage and ultra-low power source conditions. Because of its general poor energetic performances, this resonant converter will be used only during the start-up phase. The second part of the developed system is composed by a Flyback converter operating in its discontinuous conduction mode. Using this mode, the converter realizes static and very effective impedance matching with the rectenna in order to extract the maximum available power whatever are the input and the output conditions. Furthermore, thanks to the optimization procedure, the converter shows excellent efficiency performances even for μW power levels based on a discrete demonstrator. Finally, the converter provides conventional voltage levels allowing to power standard electronics. Experimental tests based on discrete prototypes for the both converters show distinguish results for the start-up voltage, the impedance matching effectiveness and the efficiency as regard to the state of the art.

Les matériaux à gradient d'indice de réfraction (GRIN) présentent des propriétés mécaniques variant en temps ou/et en espace. Ils ont été testés pour des applications prometteuses dans de nombreuses applications d'ingénierie, comme pour le contrôle santé structurale ou la surveillance de structure (SHM), le contrôle des vibrations et bruit, la récupération d'énergie, etc. D'un autre côté, les matériaux piézoélectriques offrent la possibilité de réaliser des cellules composites dont les propriétés mécaniques peuvent être contrôlées en ligne. Motivé par ces deux approches, cette thèse étudie la mise en œuvre de structures GRIN adaptatifs pour le contrôle des ondes élastiques. Deux types de structures GRIN adaptatifs sont étudiés dans ce travail. Le premier exemple concerne la mise en œuvre d'une lentille piézoélectrique dans une plaque. Il est composé de patchs piézoélectriques shuntés, collés périodiquement en surface du guide d'ondes. Les circuits de shunt utilisés permettent d'émuler une capacité négative (NC). En accordant les valeurs de NC on peut ajuster l'indices de réfraction du milieu à l'intérieur de la lentille piézoélectrique et pour satisfaire une fonction sécante hyperbolique. Les résultats numériques montrent que les lentilles piézoélectriques peuvent alors focaliser les ondes de flexion de la plaque sur les points focaux. La lentille piézoélectrique est efficace dans une grande bande de fréquences et efficace dans une grande plage de fonctionnement. Ainsi elle peut focaliser des ondes sur différent points par simple ajustement des valeurs de NC réalisés par le circuit. Cette focalisation adaptative la rend très intéressante pour de nombreuses applications comme la récupération d'énergie ou le SHM. La mise en œuvre de ces techniques pour la récupération d'énergie est discutée dans cette thèse. Le second exemple concerne l'étude d'une structure dont les propriétés mécaniques sont contrôlées en temps et en espace. En particulier, une modulation périodique permet de créer une onde artificielle se propageant dans la structure. L'interaction avec des ondes mécaniques entraîne une rupture de réciprocité visible dans un diagramme de bande non symétrique. De nombreux phénomènes inhabituels sont observés dans ce type de structures variables : fractionnement des fréquences, conversion d'ondes et transmission unidirectionnelles. Deux types de conversion fréquentielle sont démontrés et expliqués. Le premier est induit par la transmission d'énergie entre les différents modes Bloch et le second type est dû à la diffusion de Bragg dans les structures modulées. La transmission unidirectionnelle des ondes pourrait être exploitée pour réaliser des diodes dans des systèmes infinis ou semi-infinis. Cependant, la transmission unidirectionnelle n'existe pas dans les systèmes finis en raison des phénomènes de conversion de fréquence. / GRadient INdex (GRIN) media are those whose properties smoothly vary in space or/and time. They have shown promising effects in many engineering applications, such as Structural Health Monitoring (SHM), vibration and noise control, energy harvesting, etc. On the other hand, piezoelectric materials provide the possibility to build unit cells, whose mechanical properties can be controlled on-line. Motivated by these two facts, adaptive GRIN structures, which can be realized using shunted piezoelectric materials, are explored in this dissertation to control guided elastic waves. Two types of adaptive GRIN structures are studied in this work. The first type is a piezo-lens. It is composed of shunted piezoelectric patches bonded on the surfaces of plates. To control the mechanical properties of the piezoelectric composite, the piezoelectric patches are shunted with Negative Capacitance (NC). By tuning the shunting NC values, refractive indexes inside the piezo-lens are designed to satisfy a hyperbolic secant function in space. Numerical results show that the piezo-lens can focus waves by smoothly bending them toward the designated focal point. The piezo-lens is effective in a large frequency band and is efficient in many different working conditions. Also the same piezo-lens can focus waves at different locations by tuning the shunting NC values. The focusing effect and tunable feature of piezo-lens make it useful in many applications like energy harvesting and SHM. The former application is fully discussed in this thesis. The focusing effect at the focal point results in a known point with high energy density, therefore harvesting at the focal point can yield more energy. Besides, the tunable ability makes the harvesting system adaptive to environment changes. The second type is the time-space modulated structure. Its properties are modulated periodically both in time and space. Particularly, the modulation works like a traveling wave in the structure. Due to the time-varying feature, time-space modulated structures break the reciprocity theorem, i.e., the wave propagation in them is nonreciprocal. Many unusual phenomena are observed during the interaction between waves and time-space modulated structures: frequency splitting, frequency conversion and one-way wave transmission. Two types of frequency conversion are demonstrated and explained. The first type is caused by energy transmission between different orders Bloch modes. The second type is due to the Bragg scattering effect inside the modulated structures. The one-way wave transmission could be exploited to realize one-way energy insulation in equivalent infinite or semi-inffnite systems. However, the one-way energy insulation fails in finite systems due to the frequency conversion phenomenon.

The present research work mainly focuses on the fabrication of 0.85PMN-0.15PT thin film relaxor ferroelectrics for energy harvesting applications. Chapter 1 gives a brief review about why energy harvesting is required and the different ways it can be scavenged. An introduction to relaxor ferroelectrics and their characteristics structural features are discussed. A brief introduction is given about charge storage, electrocaloric effect , DC-EFM and integration over Si substrate is discussed. Finally, the specific objectives of the current research are outlined. Chapter 2 deals with the various experimental studies carried out in this research work. It gives the details of the experimental set up and the basic operation principles of various structural and physical characterizations of the materials prepared. A brief explanation of material fabrication, Microstructural and physical property measurements is discussed. Chapter3 involves the optimization process carried out to contain a phase pure PMN-PT structure without any pyrochlore phase. The optimization process is an important step in the fabrication of a thin film as the quality of any device is determined by their structural and Microstructural features. XRD, SEM, AFM were used to characterize the observed phase formation in these films. The optimizing domain images of polycrystalline 0.85PMN-0.15PT thin films on La0.5Sr0.5CoO3/ (111) Pt/TiO2/SiO2/Si substrates deposited at different oxygen partial pressures are presented. The oxygen pressure has a drastic influence on the film growth and grain morphology which are revealed through XRD and SEM characterization techniques. The presence of oxygen vacancies have found to influence the distribution of polar nanoregions and their dynamics which are visualized using domain images acquired by DC-EFM In Chapter 7 the piezoelectric response of 0.85PMN-0.15PT thin films are studied due to the electric field induced bias. From this the d33 value is calculated. d33 value is an important parameter which determines whether a material is suitable for device application (PZT). But, for a device fabrication it is important to integrate them with Si wafer which is not a straightforward work .Hence, buffer layers are used to obtain a pure perovskite PMN-PT film. We have deposited 0.85PMN-0.15PT thin films of 500 nm on a SOI wafer and tried to investigate their piezoelectric application. Chapter 8 summarizes the present study and discusses about the future work that could give more insight into the understanding of the0.85PMN-0.15 PT relaxor ferroelectric thin film.

Many sequential decision making problems under uncertainty arising in engineering, science and economics are often modelled as Markov Decision Processes (MDPs). In the setting of MDPs, the goal is to and a state dependent optimal sequence of actions that minimizes a certain long-term performance criterion. The standard dynamic programming approach to solve an MDP for the optimal decisions requires a complete model of the MDP and is computationally feasible only for small state-action MDPs. Reinforcement learning (RL) methods, on the other hand, are model-free simulation based approaches for solving MDPs. In many real world applications, one is often faced with MDPs that have large state-action spaces whose model is unknown, however, whose outcomes can be simulated. In order to solve such (large) MDPs, one either resorts to the technique of function approximation in conjunction with RL methods or develops application specific RL methods. A solution based on RL methods with function approximation comes with the associated problem of choosing the right features for approximation and a solution based on application specific RL methods primarily relies on utilizing the problem structure. In this thesis, we investigate the problem of choosing the right features for RL methods based on function approximation as well as develop novel RL algorithms that adaptively obtain best features for approximation. Subsequently, we also develop problem specie RL methods for applications arising in the areas of wireless sensor networks and road traffic control. In the first part of the thesis, we consider the problem of finding the best features for value function approximation in reinforcement learning for the long-run discounted cost objective. We quantify the error in the approximation for any given feature and the approximation parameter by the mean square Bellman error (MSBE) objective and develop an online algorithm to optimize MSBE. Subsequently, we propose the first online actor-critic scheme with adaptive bases to find a locally optimal (control) policy for an MDP under the weighted discounted cost objective. The actor performs gradient search in the space of policy parameters using simultaneous perturbation stochastic approximation (SPSA) gradient estimates. This gradient computation however requires estimates of the value function of the policy. The value function is approximated using a linear architecture and its estimate is obtained from the critic. The error in approximation of the value function, however, results in sub-optimal policies. Thus, we obtain the best features by performing a gradient descent on the Grassmannian of features to minimize a MSBE objective. We provide a proof of convergence of our control algorithm to a locally optimal policy and show numerical results illustrating the performance of our algorithm. In our next work, we develop an online actor-critic control algorithm with adaptive feature tuning for MDPs under the long-run average cost objective. In this setting, a gradient search in the policy parameters is performed using policy gradient estimates to improve the performance of the actor. The computation of the aforementioned gradient however requires estimates of the differential value function of the policy. In order to obtain good estimates of the differential value function, the critic adaptively tunes the features to obtain the best representation of the value function using gradient search in the Grassmannian of features. We prove that our actor-critic algorithm converges to a locally optimal policy. Experiments on two different MDP settings show performance improvements resulting from our feature adaptation scheme. In the second part of the thesis, we develop problem specific RL solution methods for the two aforementioned applications. In both the applications, the size of the state-action space in the formulated MDPs is large. However, by utilizing the problem structure we develop scalable RL algorithms. In the wireless sensor networks application, we develop RL algorithms to find optimal energy management policies (EMPs) for energy harvesting (EH) sensor nodes. First, we consider the case of a single EH sensor node and formulate the problem of finding an optimal EMP in the discounted cost MDP setting. We then propose two RL algorithms to maximize network performance. Through simulations, our algorithms are seen to outperform the algorithms in the literature. Our RL algorithms for the single EH sensor node do not scale when there are multiple sensor nodes. In our second work, we consider the problem of finding optimal energy sharing policies that maximize the network performance of a system comprising of multiple sensor nodes and a single energy harvesting (EH) source. We develop efficient energy sharing algorithms, namely Q-learning algorithm with exploration mechanisms based on the -greedy method as well as upper confidence bound (UCB). We extend these algorithms by incorporating state and action space aggregation to tackle state-action space explosion in the MDP. We also develop a cross entropy based method that incorporates policy parameterization in order to find near optimal energy sharing policies. Through numerical experiments, we show that our algorithms yield energy sharing policies that outperform the heuristic greedy method. In the context of road traffic control, optimal control of traffic lights at junctions or traffic signal control (TSC) is essential for reducing the average delay experienced by the road users. This problem is hard to solve when simultaneously considering all the junctions in the road network. So, we propose a decentralized multi-agent reinforcement learning (MARL) algorithm for solving this problem by considering each junction in the road network as a separate agent (controller) to obtain dynamic TSC policies. We propose two approaches to minimize the average delay. In the first approach, each agent decides the signal duration of its phases in a round-robin (RR) manner using the multi-agent Q-learning algorithm. We show through simulations over VISSIM (microscopic traffic simulator) that our round-robin MARL algorithms perform significantly better than both the standard fixed signal timing (FST) algorithm and the saturation balancing (SAT) algorithm over two real road networks. In the second approach, instead of optimizing green light duration, each agent optimizes the order of the phase sequence. We then employ our MARL algorithms by suitably changing the state-action space and cost structure of the MDP. We show through simulations over VISSIM that our non-round robin MARL algorithms perform significantly better than the FST, SAT and the round-robin MARL algorithms based on the first approach. However, on the other hand, our round-robin MARL algorithms are more practically viable as they conform with the psychology of road users.

This work studies the influence of field effect on large area 2 dimensional ribbons of fluids. A fluid of choice is confined in the channel of a metal-insulator-channel-insulator-metal architecture and is subjected to constant (d.c) or alternating (a.c) fields (de-pending on the application) along with a pressure drive flow. A general fluid would be composed of molecules having certain polarizability and be a dispersion of non-ionic and ionic particulates. The field effect response under pressure driven flow for this fluid would result in electrophoresis, electro osmosis, dielectrophoresis, dipole-dipole interaction and inverse electro osmosis phenomena. Using some of these phenomena we study applications related to desalination and energy harvesting with saline water as the ex-ample fluid for the former case, and solution processed poly vinyldene fluoride (PVDF) for the latter case. The geometrical features of \large area" and the \ribbon shape" can be taken advantage of to influence the design and performance for both applications. With regards to desalination, it is shown via experiments and theoretical models that the presence of alternating electric fields aid in ion separation along the flow when the saline water is subjected to laminar flow. Moreover, the power consumption is low due to the presence of the insulator. An average of 30% ion removal efficiency and 15% throughput is observed in the systems fabricated. Both performance parameters are discussion can be improved upon with larger channel lengths. The \2-D ribbon" and alternating field effect aid in achieving this by patterning the randomly distributed ions in the bulk into a smooth sheet charge and then repelling this sheet charge back into the bulk. The electric field exhibited by this sheet charge helps trap more ion sheets near the interface, thereby converting a surface ion trapping phenomena (when d.c is used) to a bulk phenomena and thereby improving efficiency. With regards to energy harvesting, a solution of PVDF in methyl ethyl ketone and 1-methyl-2-pyrollidone is confined to the \2-D ribbon" geometry and subject to high d.c fields. This aids in combining the fabrication, patterning and poling process for PVDF into one setup. Since the shape of the ribbon is defined by the shape of the channel, the ribbons (straight or serrated) can be used to sense forces of various magnitudes. More importantly experiments and theoretical models are studied for energy harvesting. Since the ribbon geometry defines the resonant frequency, large PVDF ribbon can be used to harvest energy from low frequency vibrations. Experiments show that up to 60 microwatt power can be harvested at 200 Hz and is sufficient to supplement the power for ICs.

Ces travaux s'inscrivent dans la problématique de l'alimentation autonome de systèmes électroniques communicants fondée sur la récupération de l'énergie disponible dans l'environnement humain. Cette thèse traite du dimensionnement d'un générateur multisource (thermique, photovoltaïque et cinétique) avec stockage d'énergie. Afin d'optimiser le dimensionnement d'un tel système dans un contexte de ressources paramétrables, des modèles génériques, adaptés à une large plage de dimensions, ont été établis, à partir de technologies déjà existantes, et validés expérimentalement. L'approche système a permis d'étudier les nombreux couplages multiphysiques existants et de mieux dimensionner le système. Ainsi, il a été montré qu'optimiser le rendement global de toute la chaîne de conversion d'énergie conduit à des solutions différentes de celles résultant d'une optimisation du dimensionnement de chaque organe pris séparément. Enfin, dans la dernière partie de cette thèse, une étude a été menée sur l'impact du profil de consommation sur le dimensionnement du système. Cette étude a permis, sur le cas particulier d'une application réelle, de mettre en évidence le potentiel d'une gestion d'énergie en cas de ressources faibles notamment par l'adaptation des profils de consommation. / This work deals with the problematic of self-powered communicating electronic systems based onenergy harvesting in the human environment. This thesis addresses the sizing of a multisource generator(thermal, photovoltaic and kinetic) with energy storage. To optimize the sizing of such a systemin the context of configurable resources, generic models, adapted to a wide range of dimensions, wereestablished from existing technology, and validated experimentally. The system-level approach wasused to study the many existing multiphysics couplings to better size the system. Thus, it was shownthat optimizing the global efficiency of the whole energy conversion chain leads to solutions differentfrom those resulting from sizing optimization of each component separately. Finally, in the latter partof this thesis, a study was conducted on the impact of load profil on the sizing of the system. Thisstudy, on the particular case of a real application, highlight the potential for energy management inthe case of poor ressources, notably by adapting the load profils.