Les avancées technologiques ont ouvert la voie à des approches de biopsie optique innovantes. Elles permettent à l'inverse des méthodes physiques de profiter des avantages d'une procédure mini-invasive, temps réels et répétitive. Le système de tomographie par cohérence optique (OCT) (la technique de biopsie optique utilisée dans cette thèse) propose des approches qui naviguent dans le corps humain grâce à des sondes endoscopiques robotisées. Toutefois, leur contrôle une fois à l’intérieur du corps devient difficile, surtout si l’objectif est de suivre l’évolution d'une zone cible, en faisant un travail de repositionnement dans le temps. L'asservissement visuel est un outil de choix pour le contrôle et le positionnement directement par l'image. Néanmoins, la richesse des informations présentes dans les images autorisent l'utilisation de plusieurs types d'information visuelle. Dans ce contexte, nous proposons l'utilisation de primitives visuelles innovantes fondées sur les ondelettes. Ainsi, deux approches d'asservissements visuels fondées sur les ondelettes ont été développées. La première approche est un asservissement visuel 2D pose fondé sur les ondelettes spectrales continues qui assure une convergence sur un espace plus important avec une bonne robustesse au bruit et une commande découplée. La deuxième est un asservissement visuel 2D direct fondé sur les ondelettes multirésolution, principalement pour faire du positionnement aux petits déplacements. Par ailleurs, la deuxième méthode couvre les 6 DDL quand la première se limite aux 3 DDL dans les images CCD. De plus, ces deux approches ont prouvé leurs aptitudes à faire du positionnement des coupes OCT. Mais encore, nous avons proposé une méthode de positionnement partitionnée que nous pouvons qualifier d'hybride, car elle exploite deux modalités d'images (OCT - CCD) pour assurer un positionnement sur SE(3) d'un échantillon. De même, nous avons proposé une méthode d'étalonnage des images de coupe et de volume OCT, liée aux distorsions générées par le chemin optique parcouru par le faisceau laser OCT. Finalement, ces travaux ouvrent la voie vers des applications dans le positionnement des volumes OCT, la compensation de mouvement physiologique et le suivi d'outils par des images OCT. / The technological advances have facilitated the optical biopsy approaches, unlike physical methods to take advantage of a minimally invasive, real time and repetitive procedure. The optical coherence tomography system is one of the optical biopsy techniques used in this thesis to prospect in the human body with robotized OCT endoscopic probes. Nevertheless, their control once inside the body becomes difficult, especially if the goal is following changes in the target area. The visual servoing is an ideal tool for the control and positioning of the robot. However, the amount of information present in the images allows the use of several types of visual features. In this thesis, we propose to use an innovative visual servoing feature based on wavelets. This representation developed as the evolution of the Fourier transform for non-stationary signals provides a time-frequency representation of the signal with a better extraction of the relevant information. Indeed, two visual servoing approaches based on wavelets were developed. The first approach is a 2D pose visual servoing based on spectral continuous wavelets, which ensures convergence over a larger area and decoupled control. The second is a direct 2D visual servoing based on multiresolution wavelets, mainly for small displacements positioning. However, the latter covers the 6 DOF when the previous one is limited to 3 DDL with a CCD camera. Both approaches have proven their ability to make the positioning of B-Scan OCT images. After that, we have proposed a method of partitioned positioning, that we can qualify by hybrid because it uses two image modalities to ensure SE(3) positioning of a sample. On the other side, we proposed a calibration method of B-Scan and 3D-Scan OCT images, due to the distortions generated by the optical path of the laser beam in OCT. Finally, these thesis is a beginning work for applications in positioning of 3D-Scan OCT, physiological motion compensation and monitoring tools by OCT images.

On rencontre aujourd'hui la vision par ordinateur employée pour la réalisation de nombreuses applications de la robotique moderne. L'un des axes de recherche actuel qui tend à améliorer ces systèmes est centré sur la commande. L'objectif est de proposer des schémas de commande originaux permettant de lier efficacement les informations mesurées par les capteurs de vision aux actions que l'on souhaite réaliser avec les robots. C'est dans cet aspect que s'inscrit ce document en apportant de nouvelles méthodes à la commande robotique classique faisant intervenir la vision, l'asservissement visuel. Le cas de pertes d'informations visuelles pendant la réalisation d'une tâche d'asservissement visuel est étudié. Dans ce sens, deux méthodes de prédiction qui permettent à la tâche d'être réalisée malgré ces pertes sont présentées. Puis une méthode inédite de correction est proposée. Celle-ci permet d'obtenir de meilleurs résultats de prédiction qu'une méthode classique, comme le démontrent des résultats obtenus en simulation et en condition réelle. Enfin, dans le contexte de la réalisation de plusieurs tâches d'asservissement visuel successives, une nouvelle méthode est présentée. Celle-ci exploite les caractéristiques d'un schéma de commande utilisé depuis quelques dizaines d'années dans l'industrie et la recherche, la commande prédictive basée modèle. Des résultats obtenus en simulation proposent de visualiser les effets de cette méthode sur le comportement d'un drone qui embarque une caméra. / The computer vision is used for the achievement of many applications of modern robotics. One of the current research topics that aims to improve these systems is focused on command. The objective consists to propose original control schemes to effectively link the information measured by the vision sensor to the actions that are to be achieved with the robots. This document is part of this look by bringing new methods to classical robotic control involving vision, the visual servoing.The case of visual information losses during the achievement of a visual servoing task is studied. In this sense, two prediction methods that allow the task to be achieved despite these losses are presented. Then a new method of correction is proposed. This provides better prediction results than a conventional method, as shown by the results obtained in simulation and in real conditions.Finally, in the context of the achievement of several successive visual servoing tasks, a new method is presented. This exploits the characteristics of a control scheme used for several decades in industry and research, model based predictive control. The results obtained in simulation propose to see the effect of this method on the behavior of a drone that embeds a camera.

Ce manuscrit traite de la conception d’un robot quadrirotor bio-inspiré. Ce robot, nommé QuadMorphing, s’inspire de l’oiseau et possède la capacité de se replier en vol afin de diminuer son envergure. Cette particularité est intéressante pour des problématiques d’évitement d’obstacles dans des milieux encombrés.Le travail présenté ici contient une présentation du robot où la plateforme mécatronique y est décrite en détails. Puis, des résultats expérimentaux sont présentés et commentés afin de quantifier les performances du prototype QuadMorphing durant des scénarios de franchissement d’obstacles.La deuxième partie de cette thèse traite de l’estimation de la taille d’obstacles en vol grâce à une perception visuelle monoculaire. Deux algorithmes d’estimation ont été simulés afin d’être validés pour être ensuite mis en place sur une nouvelle version du robot qui a été testée expérimentalement. Ces estimations permettent par la suite de rendre le robot plus autonome pour éviter les collisions avec son environnement et actionner son système de changement de forme si cela est nécessaire. / This paper describes a bio-inspired quadrotor design. This robot, called QuadMorphing, is inspired by birds and has the ability to fold its mechanical structure to reduce its wingspan during the flight. This feature could be useful for obstacle avoidance task in cluttered environments.The work presented here contains a full description of the mechatronic structure. Then, experimental results are presented and discussed in order to quantify the QuadMorphing performances during obstacle avoidance scenarios.The second part of this thesis deals with estimating obstacle size during flight using monocular visual perception. Two estimation algorithms were simulated in order to be validated and then implemented for experimental testing on a new version of the robot. In order to make the robot autonomous, the estimation of the size of the obstacle allows the robot to avoid collisions with its environment and to perform its morphological reduction if necessary.

Cette thèse porte sur la réalisation des tâches avec la locomotion sur des robots humanoïdes. Grâce à leurs nombreux degrés de liberté, ces robots possèdent un très haut niveau de redondance. D’autre part, les humanoïdes sont sous-actionnés dans le sens où la position et l’orientation ne sont pas directement contrôlées par un moteur. Ces deux aspects, le plus souvent étudiés séparément dans la littérature, sont envisagés ici dans un même cadre. En outre, la génération d’un mouvement complexe impliquant à la fois des tâches de manipulation et de locomotion, étudiée habituellement sous l’angle de la planification de mouvement, est abordée ici dans sa composante réactivité temps réel. En divisant le processus d’optimisation en deux étapes, un contrôleur basé sur la notion de pile de tâches permet l’adaptation temps réel des empreintes de pas planifiées dans la première étape. Un module de perception est également conçu pour créer une boucle fermée de perception-décision-action. Cette architecture combinant planification et réactivité est validée sur le robot HRP-2. Deux classes d’expériences sont menées. Dans un cas, le robot doit saisir un objet éloigné, posé sur une table ou sur le sol. Dans l’autre, le robot doit franchir un obstacle. Dans les deux cas, les condition d’exécution sont mises à jour en temps réel pour faire face à la dynamique de l’environnement : changement de position de l’objet à saisir ou de l’obstacle à franchir. / This thesis focuses on realization of tasks with locomotion on humanoid robots. Thanks to their numerous degrees of freedom, humanoid robots possess a very high level of redundancy. On the other hand, humanoids are underactuated in the sense that the position and orientation of the base are not directly controlled by any motor. These two aspects, usually studied separately in manipulation and locomotion research, are unified in a same framework in this thesis and are resolved as one unique problem. Moreover, the generation of a complex movement involving both tasks and footsteps is also improved becomes reactive. By dividing the optimization process into appropriate stages and by feeding directly the intermediate result to a task-based controller, footsteps can be calculated and adapted in real-time to deal with changes in the environment. A perception module is also developed to build a closed perception-decision-action loop. This architecture combining motion planning and reactivity validated on the HRP-2 robot. Two classes of experiments are carried out. In one case the robot has to grasp an object far away at different height level. In the other, the robot has to step over an object on the floor. In both cases, the execution conditions are updated in real-time to deal with the dynamics of the environment: changes in position of the target to be caught or of the obstacle to be stepped over.

Cette thèse s’inscrit dans le domaine de l’automatique, et a pour but de fournir des outils utiles à l’atterrissage en mer d’hélicoptères (sur navire ou plateforme) et employés dans le cadre d’un potentiel pilotage automatique. L’objectif a donc été de développer une série de lois de commande pilotées manuellement, puis commandées de façon autonome à l’aide d’informations caméra. Les lois ainsi développées à l’aide de modèles dynamiques d’hélicoptères, limitations mécaniques incluses, se basent sur les critères de Qualité de Vol issus de la norme ADS-33. L’ensemble a fait appel à une approche anti-windup pour améliorer la robustesse face aux situations d’actionneurs en saturation. Enfin les lois issues de ces travaux ont été testées en temps-réel sur le banc de pilotage d’hélicoptères de l’ONERA Salon-de-Provence. / This thesis is related to the automatic & control engineering field, and itsmain goal is to provide useful tools for ship landing missions, tools that can be used fora potential autopilot. The objective has been to develop a series of control laws manuallypiloted, then automatically controlled by visual servoing using identified image features. Thelaws developped thanks to helicopter models with mechanical limitations were based on flyingqualities criteria from the ADS-33 standard. The process also defines an anti-windup approachto cope with actuator saturations. Finally the main results were assessed in real time withthe ONERA rotorcraft flight test bench at ONERA Salon-de-Provence.

[pt] Neste trabalho, apresenta-se diferentes esquemas de controle servovisuais para tarefas robóticas de colheita de fruta, na presença de incertezas paramétricas nos modelos do sistema. O primeiro esquema combina as abordagens de servovisão baseada em posição (PBVS) e servovisão baseada em imagem (IBVS) para realizar, respectivamente, a aproximação até a fruta e, em seguida, um ajuste fino para a colheita. O segundo esquema usa uma abordagem de servovisão híbrida (HVS) para realizar a tarefa de colheita completa, projetando uma lei de controle adequada que combina vetores de erro definidos no espaço operacional e no espaço da imagem. A fase de detecção utiliza um algoritmo baseado no espaço de cores OTHA e limiar da imagem Otsu para um rápido reconhecimento de frutos maduros em cenários complexos. Além disso, um método de detecção mais preciso emprega uma Rede Neural Convolucional Profunda (DCNN) pré-treinada baseada em uma versão Segnet minimizada para uma inferência rápida durante a execução da tarefa. A localização do objeto é realizada empregando uma técnica de triangulação de imagem, que combina os algoritmos SURF e RANSAC ou ORB e BF-Matcher para extrair a característica da imagem da fruta e associa-lo com o seu ponto correspondente na outra visualização. No entanto, como esses algoritmos exigem um elevado custo computacional para os requisitos da tarefa, um método de estimativa mais rápido utiliza o centróide da fruta e transformação homogênea para descobrir os pontos correspondentes. Finalmente, um esquema de controle em modos deslizantes (SMC) baseado em visão e uma função de monitoramento de comutação são empregados para lidar com incertezas nos parâmetros de calibração do sistema de câmera-robô. Nesse sentido, é possível garantir a estabilidade assintótica e a convergência do erro da característica da imagem, mesmo que o ângulo de desalinhamento, em torno do eixo z, entre os sistemas de coordenadas da câmera e do efetuador seja incerto. / [en] In this work, we present different eye-in-hand visual servoing control schemes applied to a robotic harvesting task of soft fruits in the presence of parametric uncertainties in the system models. The first scheme combines position-based visual servoing (PBVS) and image-based visual servoing (IBVS) approaches in order to perform respectively an approach phase to the fruit and then a fine tuning of the end-effector to harvest. The second scheme uses a hybrid visual servoing (HVS) approach to fulfill the complete harvesting task, by designing a suitable control law which combines error vectors defined in both the image and operational spaces. For detecting the fruits, an algorithm based on the combination of the OHTA color space and Otsu’s threshold method for a fast recognition of mature fruits in complex scenarios. In addition, a more accurate detection method employs a pre-trained deep encoder-decoder algorithm based on a minimized Segnet version for a fast and cheap inference during the task execution. The object localization is accomplished by employing an image triangulation technique, which combines the speeded-up-robust-features (SURF) and the-randomsample-consensus (RANSAC) or the Oriented FAST and Rotated BRIEF and the Brute-Force Matcher (BF-Matcher) algorithms to extract the fruit image feature and match it to its correspondent feature-point into the other view of the stereo camera. However, since these algorithms are computationally expensive for the task requirements, a faster estimation method uses the fruit centroid and a homogeneous transformation for discovering matching points. Finally, a vision-based sliding-mode-control scheme and a switching monitoring function are employed to cope with uncertainties in the calibration parameters of the camera-robot system. In this context, it is possible to guarantee the asymptotic stability and convergence of the image feature error, even if the misalignment angle, around the z-axis, between the camera and end-effector frames is uncertain. 3D computer simulations and preliminary experimental results, obtained with a Mitsubishi robot arm RV-2AJ carrying out a simple strawberry picking task, are included to illustrate the performance and effectiveness of the proposed control scheme.

This thesis focuses on the development of a distributed mobile robot localization and communication system for special interventions like those carried out by fire-fighters in fire ground search and rescue. The use case scenario is related to the one described for the GUARDIANS EU project, where a swarm formation of mobile robots accompany a fire fighter during a rescue intervention in a warehouse. In this line, localizing the robots and the fire fighter during an indoor intervention with the presence of smoke is one of the more interesting challenges in this scenario. Several localization techniques have been developed using ultrasonic sensors, radio frequency signals and visual information. It has also been studied several communication protocols that can help to improve the efficiency of the system in such scenario and a proposal for designing a cross-layer communication platform that improves the connectivity of the mobile nodes during an intervention and reduces the number of lost data packets.

Imagine that a robot fetched this thesis for you from a book shelf. How doyou think the robot would have been programmed? One possibility is thatexperienced engineers had written low level descriptions of all imaginabletasks, including grasping a small book from this particular shelf. A secondoption would be that the robot tried to learn how to grasp books from yourshelf autonomously, resulting in hours of trial-and-error and several bookson the floor.In this thesis, we argue in favor of a third approach where you teach therobot how to grasp books from your shelf through grasping by demonstration.It is based on the idea of robots learning grasping actions by observinghumans performing them. This imposes minimum requirements on the humanteacher: no programming knowledge and, in this thesis, no need for specialsensory devices. It also maximizes the amount of sources from which therobot can learn: any video footage showing a task performed by a human couldpotentially be used in the learning process. And hopefully it reduces theamount of books that end up on the floor. This document explores the challenges involved in the creation of such asystem. First, the robot should be able to understand what the teacher isdoing with their hands. This means, it needs to estimate the pose of theteacher's hands by visually observing their in the absence of markers or anyother input devices which could interfere with the demonstration. Second,the robot should translate the human representation acquired in terms ofhand poses to its own embodiment. Since the kinematics of the robot arepotentially very different from the human one, defining a similarity measureapplicable to very different bodies becomes a challenge. Third, theexecution of the grasp should be continuously monitored to react toinaccuracies in the robot perception or changes in the grasping scenario.While visual data can help correcting the reaching movement to the object,tactile data enables accurate adaptation of the grasp itself, therebyadjusting the robot's internal model of the scene to reality. Finally,acquiring compact models of human grasping actions can help in bothperceiving human demonstrations more accurately and executing them in a morehuman-like manner. Moreover, modeling human grasps can provide us withinsights about what makes an artificial hand design anthropomorphic,assisting the design of new robotic manipulators and hand prostheses. All these modules try to solve particular subproblems of a grasping bydemonstration system. We hope the research on these subproblems performed inthis thesis will both bring us closer to our dream of a learning robot andcontribute to the multiple research fields where these subproblems arecoming from. / QC 20111125

Neste trabalho é apresentado o desenvolvimento de um sistema de posicionamento dinâmico para uma pequena embarcação baseado em controle a estrutura variável com realimentação por visão computacional. Foram investigadas, na literatura, diversas técnicas desenvolvidas e escolheu-se o controle a estrutura variável devido, principalmente, ao modo de acionamento dos propulsores presentes no barco utilizado para os experimentos. Somando-se a isto, foi considerada importante a robustez que a técnica de controle escolhida apresenta, pois o modelo utilizado conta com incerteza em sua dinâmica. É apresentado ainda o projeto da superfície de deslizamento para realizar o controle a estrutura variável. Como instrumento de medição optou-se por utilizar técnicas de visão computacional em imagens capturadas a partir de uma webcam. A escolha por este tipo de sistema deve-se a alta precisão das medições aliada ao seu baixo custo. São apresentadas simulações e experimentos com controle a estrutura variável em tempo discreto utilizando a integral do erro da posição visando eliminar o erro em regime. Para realizar o controle que demanda o estado completo, são comparados quatro estimadores de estado realizados em tempo discreto: derivador aproximado; observador assintótico com uma frequência de amostragem igual a da câmera; observador assintótico com uma frequência de amostragem maior que a da câmera; e filtro de Kalman. / The design of a dynamic positioning system for a small boat based on variable strutucture control and visual-servoing is presented. Many control tecniques have been investigated and the variable structure control was chosen, mainly, due the operation mode of the motor drivers installed on the boat applied in the experiments. The robustness of this control technique was also considered since the available dynamic model of the boat is uncertain. The design of the sliding surface is shown as well. Computer vision techniques were used to measure the position of the boat from images taken with a webcam, this kind of measure system was chosen due to its high accuracy and low cost. Simulation and experimental results of discrete time variable structure control with integral action of the boats postion, included in order to eliminate steady state error, are shown. To develop this controller four state estimators, in discrete time, are compared: derivative of position; asymptotic observer at a high sampling rate; asymptotic observer at webcams sampling rate; and Kalman filter.

Une maladie appelée cholestéatome affecte l'oreille moyenne, en absence de traitement, elle pourrait conduire à des complications graves. Le seul traitement dans la pratique médicale actuelle est une procédure chirurgicale. Les incidences de cholestéatome résiduelle ou récurrente sont élevés et le patient doit subir plus d'une intervention chirurgicale. Par conséquent, un système robotique original a été proposé pour d'éliminer l'incidence du cholestéatome résiduel en enlevant efficacement toutes les cellules infectées de la première intervention chirurgicale, et de faire une chirurgie moins invasive. Ainsi, ce manuscrit montre les différents défis auxquels fait face le chirurgien à travers une telle micro-procédure. Il est également défini le cahier de charge pour la réalisation d'un système futuriste dédié à la chirurgie du cholestéatome. En outre, un contrôleur est proposé comme un première étape vers le système idéal. Un tel contrôleur permet de guider un outil chirurgical rigide afin de suivre un chemin de référence sous les contraintes du trou d'incision. Le contrôleur proposé peut guider soit un outil droit, soit un outil courbe. En effet, le contrôleur proposé est une commande de haut niveau qui es formulé dans l'espace de tâche (ou espace Cartésien). Ce contrôleur est une couche modulaire qui peut être ajoutée à différentes structures robotiques. Le contrôleur proposé a montré de bons résultats en termes de précision tout en étant évalué sur un robot parallèle et un robot en série. / A disease called cholesteatoma affects the middle ear, in the absence of treatment, it could lead to serious complications. The only treatment in current medical practice is a surgical procedure. Incidences of residual or recurrent cholesteatoma are high and the patient may have more than one surgical procedure. Therefore, a novel robotic system was proposed to eliminate the incidence of residual cholesteatoma by removing efficiently all infected cells from the first surgery, and make a less invasive surgery. Thus, this manuscript shows the different challenges that face the surgeon through such a micro-procedure. It also is specified the requirements for achieving a futuristic system dedicated to cholesteatoma surgery. In addition, a controller is proposed as a first step toward the ideal system. Such a controller allows to guide a rigid surgical tool for following a reference path under the constraints of the incision hole. The proposed controller can guide either a straight tool or a curved one. Indeed, the proposed controller is a high level control which is formulated in the task-space (or Cartesian-space). This controller is a modular layer which can be added to different robotics structures. The proposed controller showed a good results in term of accuracy while assessed on a parallel robot and a serial one.