Le travail effectué au cours de cette thèse tente d’apporter une solution algorithmique à la problématique de l’estimation de l’état d’un mini-drone en vol qui soit compatible avec les exigences d’embarquabilité inhérentes au système. Il a été orienté vers les méthodes d’estimation non linéaire à base de modèles. Les algorithmes d’estimation, d’état ou de paramètres, et de contrôle apparaissent primordiaux, lorsque la technologie des capteurs et des actionneurs, pour des raisons de coût et d’encombrement essentiellement, ne permet pas de disposer de capacités illimitées. Ceci est particulièrement vrai dans le cas des micro- et des mini-drones. L’estimation permet de fusionner en temps réel les informations imparfaites provenant des différents capteurs et de fournir une estimation, par exemple de l’état du drone (orientation, vitesse, position) au calculateur embarqué où sont implémentés les algorithmes de commande de l’engin. Ce contrôle de l’appareil doit garantir sa stabilité en boucle fermée quelque soit l’ordre de consigne fourni directement par l’opérateur ou par tout système automatique de gestion du vol et assurer que celle-ci soit correctement recopiée. Estimation et commande participent donc grandement au succès de toute mission. Une dimension extrêmement importante qui a conditionné les travaux entrepris tout au long de cette thèse concerne la capacité d’emport des mini-drones que nous considérons. En effet, celle-ci, relativement limitée, et couplée à la volonté de ne pas grever les budgets de développement de tout mini-drone, autorise uniquement l’intégration de matériels dits bas-coûts. Malgré les progrès significatifs de la miniaturisation et l’augmentation continuelle des capacités de calcul embarqué (loi de Moore), les mini-drones d’intérêt considérés ici n’embarquent donc que des capteurs aux performances limitées dans un contexte où cette catégorie d’engins autonomes est amenée à être de plus en plus fréquemment exploitée pour remplir des missions elles-mêmes toujours plus nombreuses. Celles-ci requièrent notamment que de tels drones puissent de manière sûre s’insérer et partager l’espace aérien civil moyennant le passage d’une certification de leur vol au même titre que pour les avions de transport des différentes compagnies aériennes. Dès lors, face à cet enjeu de sécurisation des vols de mini-drones, la consolidation de la connaissance de l’état de l’aéronef par des techniques d’estimation devient un tâche essentielle pour en assurer le contrôle, y compris en situations dégradées (pannes capteurs, perte occasionnelle de signaux, bruit et perturbations environnantes, imperfections des moyens de mesure, etc). Tenter de répondre à cet enjeu conduit naturellement le chercheur à s’attaquer à des problèmes relativement nouveaux, en tout cas pas forcément aussi proches de ceux qui se posent dans le secteur de l’aéronautique civile ou militaire, où le système avionique est sans commune mesure avec celui sur lequel nous avons travaillé dans cette thèse. Ce travail à tout d’abord consisté à définir une modélisation dynamique descriptive du vol des mini-drones étudiés, suffisamment générique pour être appliquée à différents types de minidrones (voilure fixe, multirotor, etc). Par la suite, deux algorithmes d’estimation originaux, dénommés IUKF et -IUKF, exploitant ce modèle, ont été développés avant d’être testés en simulation puis sur données réelles pour la version -IUKF. Ces deux méthodes transposent le cadre générique des observateurs invariants au cas de l’estimation non linéaire de l’état d’un système dynamique par une technique de type Unscented Kalman Filter (UKF) qui appartient à la classe plus générale des algorithmes de filtrage non linéaire de type Sigma Point (SP). La solution proposée garantit un plus grand domaine de convergence de l’estimé que les techniques plus traditionnelles. / This thesis presents the study of an algorithmic solution for state estimation problem of unmanned aerial vehicles, or UAVs. The necessary resort to multiple miniaturized low-cost and low-performance sensors integrated into mini-RPAS, which are obviously subjected to hardspace requirements or electrical power consumption constraints, has led to an important interest to design nonlinear observers for data fusion, unmeasured systems state estimation and/or flight path reconstruction. Exploiting the capabilities of nonlinear observers allows, by generating consolidated signals, to extend the way mini-RPAS can be controlled while enhancing their intrinsic flight handling qualities.That is why numerous recent research works related to RPAS certification and integration into civil airspace deal with the interest of highly robust estimation algorithm. Therefore, the development of reliable and performant aided-INS for many nonlinear dynamic systems is an important research topic and a major concern in the aerospace engineering community. First, we have proposed a novel approach for nonlinear state estimation, named pi-IUKF (Invariant Unscented Kalman Filter), which is based on both invariant filter estimation and UKF theoretical principles. Several research works on nonlinear invariant observers have been led and provide a geometrical-based constructive method for designing filters dedicated to nonlinear state estimation problems while preserving the physical properties and systems symmetries. The general invariant observer guarantees a straightforward form of the nonlinear estimation error dynamics whose properties are remarkable. The developed pi-IUKF estimator suggests a systematic approach to determine all the symmetry-preserving correction terms, associated with a nonlinear state-space representation used for prediction, without requiring any linearization of the differential equations. The exploitation of the UKF principles within the invariant framework has required the definition of a compatibility condition on the observation equations. As a first result, the estimated covariance matrices of the pi-IUKF converge to constant values due to the symmetry-preserving property provided by the nonlinear invariant estimation theory. The designed pi-IUKF method has been successfully applied to some relevant practical problems such as the estimation of Attitude and Heading for aerial vehicles using low-cost AH reference systems (i.e., inertial/magnetic sensors characterized by low performances). In a second part, the developed methodology is used in the case of a mini-RPAS equipped with an aided Inertial Navigation System (INS) which leads to augment the nonlinear state space representation with both velocity and position differential equations. All the measurements are provided on board by a set of low-cost and low-performance sensors (accelerometers, gyrometers, magnetometers, barometer and even Global Positioning System (GPS)). Our designed pi-IUKF estimation algorithm is described and its performances are evaluated by exploiting successfully real flight test data. Indeed, the whole approach has been implemented onboard using a data logger based on the well-known Paparazzi system. The results show promising perspectives and demonstrate that nonlinear state estimation converges on a much bigger set of trajectories than for more traditional approaches.

This thesis deals with the filter algorithm design, implementing mathematical model to improve algorithm performance. Designed algorithms are implemented in a control unit of the experimental vehicle (filters signal used in the closed-loop controller). The improvement of the position estimation using Kalman Filter is demonstrated on the experimental vehicle. In the next part the design process of algorithm developing for dsPIC microcontroller using Matlab is described.

This work deals with the orientation of the camera in real-time with a single camera. Offline methods are described and used as a reference for comparison of a real-time metods. Metods work in real-time Monocular SLAM and PTAM methods are there described and compared. Further, paper shows hints of advanced methods whereas future work is possible.

In underwater tracking and surveillance, the active towed array sonar presents a way of discovering and tracking adversarial submerged targets that try to stay hidden. The configuration consist of listening and emitting hydrophones towed behind a ship. Moreover, it has inherent limitations, and the characteristics of sound in the ocean are complex. By varying the pulse form emitted and the trajectory of the ship the measurement accuracy may be improved. This type of optimization constitutes a sensor management problem. In this thesis, a model of the tracking scenario has been constructed derived from Cramér-Rao bound analyses. A model predictive control approach together with information measures have been used to optimize a filter's estimated state of the target. For the simulations, the MATLAB environment has been used. Different combinations of decision horizons, information measures and variations of the Kalman filter have been studied. It has been found that the accuracy of the Extended Kalman filter is too low to give consistent results given the studied information measures. However, the Unscented Kalman filter is sufficient for this purpose.

The thesis work investigates brake friction estimation strategies. The friction between the brake disc and brake pads is not constant during the braking application and contributes to the amount of brake torque achieved at the wheels. In this study, it is considered that any change in the brake torque between the requested and achieved values is only due to the varying brake friction coefficient. The work gives three different approaches to estimate the brake friction coefficient using two prominent state estimation strategies, Unscented Kalman Filter and Moving Horizon Estimation. The inputs to the estimators are obtained from a Vehicle model, which is built using the wheel balance equations. The estimators have been tuned to minimize the estimation error in nominal conditions and tested for their robustness through a wide analysis, where the sensitivity of the strategies is checked against a spectra of potential system parameters and boundary conditions. Throughout all the analysis, the developed models estimate the brake friction coefficient within an acceptable error range. This work opens up opportunities for further studies that can be performed using the built estimator models. / Detta examensarbete studerar strategier för skattning av bromsfriktion. Friktionen mellan bromsskivan och bromsbeläggen är inte konstant under bromsförloppet och det är denna som genererar bromsmomentet för varje hjul. I detta arbete så antas att förändringen i bromsmoment mellan begärd och uppnått endast är på grund av varierande bromsfriktion mellan bromsbelägg och bromsskiva. Arbetet presenterar tre olika sätt att skatta bromsfriktionen genom användning av två kända skattningsmetoder, Uncented Kalman Filter och Moving Horizon Estimation. Ingående värden till skattningsmetoderna fås från en fordonsmodell som är byggd med hjälp av hjulbalansekvationer. Skattningsmetoderna har justerats så att de minimerar skattningsfelet i nominella fall och de är testade för robusthet genom en bred analys där känsligheten hos metoderna testas genom en flora av potentiella systemparametrar och gränsvärden. Genom hela analysen så uppnår de utvecklade skattningsmetoderna bromsfriktionsvärden med acceptabla felnivåer. Detta arbete öppnar upp för möjligheter för vidare analyser där de utvecklade metoderna kan användas.

<p dir="ltr">This thesis aims to assess the efficacy of adaptive Gaussian mixture filtering for an inertial navigation-based cislunar application. The thesis focuses on a fully autonomous system, where the navigation system is solely reliant on onboard sensors and receives no navigation information from external tracking systems. The proposed adaptive filter is tested under non-ideal conditions. Specifically, this thesis considers the challenging case where range information is unavailable, and instead, only bearings angles with respect to illuminated celestial bodies are measured. The performance of the adaptive filter is compared to the unscented Kalman filter (UKF), and the filter consistency and errors are compared. The proposed filter addresses challenges in linearization errors that accrue in the UKF measurement update equations. The adaptive filter is shown to be a consistent estimator, significantly outperforming the UKF. Considering design requirements for similar navigation missions, recommendations and practical considerations are suggested for future cislunar autonomous navigation applications</p>

This thesis addresses three problems related to image registration, prediction and tracking, applied to Angiography and Oncology. For image analysis, various probabilistic models have been employed to characterize the image deformations, target motions and state estimations. (i) In Digital Subtraction Angiography (DSA), having a high quality visualization of the blood motion in the vessels is essential both in diagnostic and interventional applications. In order to reduce the inherent movement artifacts in DSA, non-rigid image registration is used before subtracting the mask from the contrast image. DSA image registration is a challenging problem, as it requires non-rigid matching across spatially non-uniform control points, at high speed. We model the problem of sub-pixel matching, as a labeling problem on a non-uniform Markov Random Field (MRF). We use quad-trees in a novel way to generate the non uniform grid structure and optimize the registration cost using graph-cuts technique. The MRF formulation produces a smooth displacement field which results in better artifact reduction than with the conventional approach of independently registering the control points. The above approach is further improved using two models. First, we introduce the concept of pivotal and non-pivotal control points. `Pivotal control points' are nodes in the Markov network that are close to the edges in the mask image, while 'non-pivotal control points' are identified in soft tissue regions. This model leads to a novel MRF framework and energy formulation. Next, we propose a Gaussian MRF model and solve the energy minimization problem for sub-pixel DSA registration using Random Walker (RW). An incremental registration approach is developed using quad-tree based MRF structure and RW, wherein the density of control points is hierarchically increased at each level M depending of the features to be used and the required accuracy. A novel numbering scheme of the control points allows us to reuse the computations done at level M in M + 1. Both the models result in an accelerated performance without compromising on the artifact reduction. We have also provided a CUDA based design of the algorithm, and shown performance acceleration on a GPU. We have tested the approach using 25 clinical data sets, and have presented the results of quantitative analysis and clinical assessment. (ii) In External Beam Radiation Therapy (EBRT), in order to monitor the intra fraction motion of thoracic and abdominal tumors, the lung diaphragm apex can be used as an internal marker. However, tracking the position of the apex from image based observations is a challenging problem, as it undergoes both position and shape variation. We propose a novel approach for tracking the ipsilateral hemidiaphragm apex (IHDA) position on CBCT projection images. We model the diaphragm state as a spatiotemporal MRF, and obtain the trace of the apex by solving an energy minimization problem through graph-cuts. We have tested the approach using 15 clinical data sets and found that this approach outperforms the conventional full search method in terms of accuracy. We have provided a GPU based heterogeneous implementation of the algorithm using CUDA to increase the viability of the approach for clinical use. (iii) In an adaptive radiotherapy system, irrespective of the methods used for target observations there is an inherent latency in the beam control as they involve mechanical movement and processing delays. Hence predicting the target position during `beam on target' is essential to increase the control precision. We propose a novel prediction model (called o set sine model) for the breathing pattern. We use IHDA positions (from CBCT images) as measurements and an Unscented Kalman Filter (UKF) for state estimation. The results based on 15 clinical datasets show that, o set sine model outperforms the state of the art LCM model in terms of prediction accuracy.

For several years, there has been a remarkable increase in efforts to develop an autonomous car. Autonomous car systems combine various techniques of recognizing the environment with the help of the sensors and could drastically bring down the number of accidents on road by removing human conduct errors related to driver inattention and poor driving choices. In this research thesis, an algorithm for jointly ego-vehicle motion and road geometry estimation for Advanced Driver Assistance Systems (ADAS) is developed. The measurements are obtained from the inertial sensors, wheel speed sensors, steering wheel angle sensors, and camera. An Unscented Kalman Filter (UKF) is used for estimating the states of the non-linear system because UKF estimates the state in a simplified way without using complex computations. The proposed algorithm has been tested on a winding and straight road. The robustness and functioning of our algorithm have been demonstrated by conducting experiments involving the addition of noise to the measurements, reducing the process noise covariance matrix, and increasing the measurement noise covariance matrix and through these tests, we gained more trust in the working of our tracker. For evaluation, each estimated parameter has been compared with the reference signal which shows that the estimated signal matches the reference signal very well in both scenarios. We also compared our joint algorithm with individual ego-vehicle and road geometry algorithms. The results clearly show that better estimates are obtained from our algorithm when estimated jointly instead of estimating separately.

Actuellement, les machines Asynchrones à Double Alimentation (MADA) sont omniprésentes dans le secteur éolien, grâce à leur simplicité de construction, leur faible coût d’achat et leur robustesse mécanique ainsi que le nombre faible d’interventions pour la maintenance. Cependant, comme toute autre machine électrique, ces génératrices sont sujettes aux défauts de différent ordre (électrique, mécanique, électromagnétique…) ou de différents types (capteur, actionneur ou composants du système). C’est pourquoi, il est primordial de concevoir une approche de diagnostic permettant de manière anticipée, de détecter, localiser et identifier tout défaut ou anomalie pouvant altérer le fonctionnement sain de ce type de machine. Motivés par les points forts des méthodes de diagnostic de défauts à base d’observateurs, nous proposons d’une part, dans cette thèse, une approche de détection, localisation et identification des défauts de la MADA d’une éolienne à vitesse variable, à base des observateurs de Kalman, performants et largement utilisés. Les erreurs d’estimation d’état du filtre de Kalman linéaire et de ses variantes non-linéaires, à noter : le Filtre de Kalman Etendu (EKF) et le Filtre de Kalman sans-Parfum (UKF), sont utilisés comme résidus sensibles aux défauts. En vue d’éviter les fausses alarmes et de découpler les défauts des perturbations et des bruits, l’analyse des résidus générés est réalisée par des tests statistiques tels que : Test de Page Hinkley (PH) et Test DCS (Dynamic Cumulative Sum). Pour la localisation des défauts multiples et simultanés, la Structure d’Observateurs Dédiés (DOS) et la Structure d’Observateurs Généralisés (GOS) sont appliquées. De plus, l’amplitude du défaut est déterminée dans l’étape d’identification de défaut. Les défauts capteurs, actionneurs et composants de la MADA, sont traités dans ce travail de recherche. D’autre part, une étude comparative entre les différents observateurs de Kalman, est élaborée. La comparaison porte sur les critères suivants : le temps de calcul, la précision et la vitesse de convergence des estimations. / Actually, the Doubly Fed Induction Generators (DFIG) are omnipresent in the wind power market, owing to their construction simplicity, their low purchase cost and their mechanical robustness. However, as any other electrical machine, these generators are subject to defects of different order (electrical, mechanical, electromagnetic ...) or of different type (sensor, actuator or system). That’s why, it is important to design an effective diagnostic approach, able to early detect, locate and identify any defect or abnormal behavior, which could undermine the healthy operation of this machine On the one hand, motivated by the observer-based fault diagnosis methods strengths, we proposed, in this thesis, a diagnostic approach for the faults detection, localization and identification of the DFIG used in variable speed wind turbine. This approach is based on the use of the efficient and widely used Kalman observers. The state estimation errors of the linear Kalman filter and the non-linear Kalman filters, named: The Extended Kalman Filter (EKF) and the Unscented Kalman Filter (UKF) are used as faults sensitive residuals. In order to avoid false alarms and to decouple faults from disturbances and noises, the faults detection is carried out by the analysis of the residuals generated, by the mean of statistical tests such as: Hinkley Page Test (PH) and DCS Test (Dynamic) Cumulative Sum). For the localization step in case of multiple and simultaneous faults, the Dedicated Observer scheme (DOS) and the Generalized Observer scheme (GOS) are applied. In addition, the fault level is determined in the fault identification step. Sensor faults, actuator and system faults of DFIG, are treated in this research work. On the other hand, a comparative study between the three Kalman observers proposed is performed. The comparison was done in terms of (1) the computation time, (2) the estimation accuracy, and (3) the convergence speed.