Rozprawy doktorskie na temat „Robust”

Este trabalho apresenta os resultados da aplicação de três técnicas de controle utilizadas no projeto e implementação do controle de um manipulador subatuado planar de três juntas em série e de elos rígidos, projetado e construído pela Universidade Carnegie Mellon, EUA. Devido ao alto grau de não-linearidade deste sistema, seria muito difícil implementar um controlador H2, H∞ ou H2/H∞ que atuasse sozinho. Assim, propõe-se a utilização de um método de controle combinado: torque computado/H2, H∞ ou H2/H∞. No controle combinado, a porção correspondente ao torque computado lineariza e pré-compensa a dinâmica do modelo da planta nominal, enquanto a porção correspondente ao controle H2, H∞ ou H2/H∞ realiza uma pós-compensação dos erros residuais, que não foram completamente eliminados pelo método torque computado. Testes de acompanhamento de trajetória e testes de robustez são realizados aqui para comprovar a eficiência destes controladores, com resultados de implementação bastante satisfatórios.
This work presents the application results of three control techniques used for the control design and implementation of a serial planar underactuated manipulator with three joints and rigid links, designed and built by the Carnegie Mellon University, USA. Due to the high non-linearity degree of this system, it would be very difficult to implement an H2, H∞ or H2/ H∞ control which would actuate on the system by itself. Therefore, it is proposed a combined control method: computed torque/ H2, H∞ or H2/H∞. In the combined control, the portion corresponding to the computed torque linearizes and pre-compensates the dynamics of the nominal model, while the portion corresponding to the H2, H∞ or H2/H∞ control realizes a pos-compensation of the residual errors, not completely removed by the computed torque method. Trajetory tracking and robustness tests are performed here to demonstrate the efficiency of these controllers, with very satisfatory implementation results.

Robust, reliable and safe navigation is one of the fundamental problems of robotics. Throughout the present thesis, we tackle the problem of navigation for robotic industrial mobile-bases. We identify its components and analyze their respective challenges in order to address them. The research work presented here ultimately aims at improving the overall quality of the navigation stack of a commercially available industrial mobile-base. To introduce and survey the overall problem we first break down the navigation framework into clearly identified smaller problems. We examine the Simultaneous Localization and Mapping (SLAM) problem, recalling its mathematical grounding and exploring the state of the art. We then review the problem of planning the trajectory of a mobile-base toward a desired goal in the generated environment representation. Finally we investigate and clarify the use of the subset of the Lie theory that is useful in robotics. The first problem tackled is the recognition of place for closing loops in SLAM. Loop closure refers to the ability of a robot to recognize a previously visited location and infer geometrical information between its current and past locations. Using only a 2D laser range finder sensor, we address the problem using a technique borrowed from the field of Natural Language Processing (NLP) which has been successfully applied to image-based place recognition, namely the Bag-of-Words. We further improve the method with two proposals inspired from NLP. Firstly, the comparison of places is strengthened by considering the natural relative order of features in each individual sensor reading. Secondly, topological correspondences between places in a corpus of visited places are established in order to promote together instances that are ‘close’ to one another. We then tackle the problem of motion model calibration for odometry estimation. Given a mobile-base embedding an exteroceptive sensor able to observe ego-motion, we propose a novel formulation for estimating the intrinsic parameters of an odometry motion model. Resorting to an adaptation of the pre-integration theory initially developed for inertial motion sensors, we employ iterative nonlinear on-manifold optimization to estimate the wheel radii and wheel separation. The method is further extended to jointly estimate both the intrinsic parameters of the odometry model together with the extrinsic parameters of the embedded sensor. The method is shown to accommodate to variation in model parameters quickly when the vehicle is subject to physical changes during operation. Following the generation of a map in which the robot is localized, we address the problem of estimating trajectories for motion planning. We devise a new method for estimating a sequence of robot poses forming a smooth trajectory. Regardless of the Lie group considered, the trajectory is seen as a collection of states lying on a spline with non-vanishing n-th derivatives at each point. Formulated as a multi-objective nonlinear optimization problem, it allows for the addition of cost functions such as velocity and acceleration limits, collision avoidance and more. The proposed method is evaluated for two different motion planning tasks, the planning of trajectories for a mobile-base evolving in the SE(2) manifold, and the planning of the motion of a multi-link robotic arm whose end-effector evolves in the SE(3) manifold. From our study of Lie theory, we developed a new, ready to use, programming library called `manif’. The library is open source, publicly available and is developed following good software programming practices. It is designed so that it is easy to integrate and manipulate, and allows for flexible use while facilitating the possibility to extend it beyond the already implemented Lie groups.
La navegación autónoma es uno de los problemas fundamentales de la robótica, y sus diferentes desafíos se han estudiado durante décadas. El desarrollo de métodos de navegación robusta, confiable y segura es un factor clave para la creación de funcionalidades de nivel superior en robots diseñados para operar en entornos con humanos. A lo largo de la presente tesis, abordamos el problema de navegación para bases robóticas móviles industriales; identificamos los elementos de un sistema de navegación; y analizamos y tratamos sus desafíos. El trabajo de investigación presentado aquí tiene como último objetivo mejorar la calidad general del sistema completo de navegación de una base móvil industrial disponible comercialmente. Para estudiar el problema de navegación, primero lo desglosamos en problemas menores claramente identificados. Examinamos el subproblema de mapeo del entorno y localización del robot simultáneamente (SLAM por sus siglas en ingles) y estudiamos el estado del arte del mismo. Al hacerlo, recordamos y detallamos la base matemática del problema de SLAM. Luego revisamos el subproblema de planificación de trayectorias hacia una meta deseada en la representación del entorno generada. Además, como una herramienta para las soluciones que se presentarán más adelante en el desarrollo de la tesis, investigamos y aclaramos el uso de teoría de Lie, centrándonos en el subconjunto de la teoría que es útil para la estimación de estados en robótica. Como primer elemento identificado para mejoras, abordamos el problema de reconocimiento de lugares para cerrar lazos en SLAM. El cierre de lazos se refiere a la capacidad de un robot para reconocer una ubicación visitada previamente e inferí información geométrica entre la ubicación actual del robot y aquellas reconocidas. Usando solo un sensor láser 2D, la tarea es desafiante ya que la percepción del entorno que proporciona el sensor es escasa y limitada. Abordamos el problema utilizando 'bolsas de palabras', una técnica prestada del campo de procesamiento del lenguaje natural (NLP) que se ha aplicado con éxito anteriormente al reconocimiento de lugares basado en imágenes. Nuestro método incluye dos nuevas propuestas inspiradas también en NLP. Primero, la comparación entre lugares candidatos se fortalece teniendo en cuenta el orden relativo natural de las características en cada lectura individual del sensor; y segundo, se establece un corpus de lugares visitados para promover juntos instancias que están "cerca" la una de la otra desde un punto de vista topológico. Evaluamos nuestras propuestas por separado y conjuntamente en varios conjuntos de datos, con y sin ruido, demostrando mejora en la detección de cierres de lazo para sensores láser 2D, con respecto al estado del arte. Luego abordamos el problema de la calibración del modelo de movimiento para la estimación de la edometría. Dado que nuestra base móvil incluye un sensor exteroceptivo capaz de observar el movimiento de la plataforma, proponemos una nueva formulación que permite estimar los parámetros intrínsecos del modelo cinemático de la plataforma durante el cómputo de la edometría del vehículo. Hemos recurrido a una adaptación de la teoría de reintegración inicialmente desarrollado para unidades inerciales de medida, y aplicado la técnica a nuestro modelo cinemático. El método nos permite, mediante optimización iterativa no lineal, la estimación del valor del radio de las ruedas de forma independiente y de la separación entre las mismas. El método se amplía posteriormente par idéntica de forma simultánea, estos parámetros intrínsecos junto con los parámetros extrínsecos que ubican el sensor láser con respecto al sistema de referencia de la base móvil. El método se valida en simulación y en un entorno real y se muestra que converge hacia los verdaderos valores de los parámetros. El método permite la adaptación de los parámetros intrínsecos del modelo cinemático de la plataforma derivados de cambios físicos durante la operación, tales como el impacto que el cambio de carga sobre la plataforma tiene sobre el diámetro de las ruedas. Como tercer subproblema de navegación, abordamos el reto de planificar trayectorias de movimiento de forma suave. Desarrollamos un método para planificar la trayectoria como una secuencia de configuraciones sobre una spline con n-ésimas derivadas en todos los puntos, independientemente del grupo de Lie considerado. Al ser formulado como un problema de optimización no lineal con múltiples objetivos, es posible agregar funciones de coste al problema de optimización que permitan añadir límites de velocidad o aceleración, evasión de colisiones, etc. El método propuesto es evaluado en dos tareas de planificación de movimiento diferentes, la planificación de trayectorias para una base móvil que evoluciona en la variedad SE(2), y la planificación del movimiento de un brazo robótico cuyo efector final evoluciona en la variedad SE(3). Además, cada tarea se evalúa en escenarios con complejidad de forma incremental, y se muestra un rendimiento comparable o mejor que el estado del arte mientras produce resultados más consistentes. Desde nuestro estudio de la teoría de Lie, desarrollamos una nueva biblioteca de programación llamada “manif”. La biblioteca es de código abierto, está disponible públicamente y se desarrolla siguiendo las buenas prácticas de programación de software. Esta diseñado para que sea fácil de integrar y manipular, y permite flexibilidad de uso mientras se facilita la posibilidad de extenderla más allá de los grupos de Lie inicialmente implementados. Además, la biblioteca se muestra eficiente en comparación con otras soluciones existentes. Por fin, llegamos a la conclusión del estudio de doctorado. Examinamos el trabajo de investigación y trazamos líneas para futuras investigaciones. También echamos un vistazo en los últimos años y compartimos una visión personal y experiencia del desarrollo de un doctorado industrial.

Legged locomotion has several interesting challenges that need to be addressed, such as the ability of dynamically walk over rough terrain like stairs or stepping stones, as well as the ability to adapt to unexpected changes in the environment and the dynamic model of the robot. This thesis is driven towards solving these challenges and makes contributions on theoretical and experimental aspects to address: dynamic walking, model uncertainty, and rough terrain. On the theoretical front, we introduce and develop a unified robust and adaptive control framework that enables the ability to enforce stability and safety-critical constraints arising from robotic motion tasks under a high level of model uncertainty. We also present a novel method of walking gait optimization and gait library to address the challenge of dynamic robotic walking over stochastically generated stepping stones with significant variations in step length and step height, and where the robot has knowledge about the location of the next discrete foothold only one step ahead. On the experimental front, our proposed methods are successfully validated on ATRIAS, an underactuated, human-scale bipedal robot. In particular, experimental demonstrations illustrate our controller being able to dynamically walk at 0.6 m/s over terrain with step length variation of 23 to 78 cm, as well as simultaneous variation in step length and step height of 35 to 60cm and -22 to 22cm respectively. In addition to that, we also successfully implemented our proposed adaptive controller on the robot, which enables the ability to carry an unknown load up to 68 lb (31 kg) while maintaining very small tracking errors of about 0.01 deg (0.0017 rad) at all joints. To be more specific, we firstly develop robust control Lyapunov function based quadratic program (CLFQP) controller and L1 adaptive control to handle model uncertainty for bipedal robots. An application is dynamic walking while carrying an unknown load. The robust CLF-QP controller can guarantee robustness via a quadratic program that can be extended further to achieve robust safety-critical control. The L1 adaptive control can estimate and adapt to the presence of model uncertainty in the system dynamics. We then present a novel methodology to achieve dynamic walking for underactuated and hybrid dynamcal bipedal robots subject to safety-critical constraints. The proposed controller is based on the combination of control Barrier functions (CBFs) and control Lyapunov functions (CLFs) implemented as a state-based online quadratic program to achieve stability under input and state constraints. The main contribution of this work is the control design to enable stable dynamical bipedal walking subject to strict safety constraints that arise due to walking over a terrain with randomly generated discrete footholds. We next introduce Exponential Control Barrier Functions (ECBFs) as means to enforce high relativedegree safety constraints for nonlinear systems. We also develop a systematic design method that enables creating the Exponential CBFs for nonlinear systems making use of tools from linear control theory. Our method creates a smooth boundary for the safety set via an exponential function, therefore is called Exponential CBFs. Similar to exponential stability and linear control, the exponential boundary of our proposed method helps to have smoother control inputs and guarantee the robustness under model uncertainty. The proposed control design is numerically validated on a relative degree 4 nonlinear system (the two-link pendulum with elastic actuators and experimentally validated on a relative degree 6 linear system (the serial cart-spring system). Thanks to these advantages of Exponential CBFs, we then can apply the method to the problem of 3D dynamic walking with varied step length and step width as well as dynamic walking on time-varying stepping stones. For the work of using CBF for stepping stones, we use only one nominal walking gait. Therefore the range of step length variation is limited ([25 : 60](cm)). In order to improve the performance, we incorporate CBF with gait library and increase the step length range significantly ([10 : 100](cm)). While handling physical constraints and step transition via CBFs appears to work well, these constraints often become active at step switching. In order to resolve this issue, we introduce the approach of 2-step periodic walking. This method not only gives better step transitions but also offers a solution for the problem of changing both step length and step height. Experimental validation on the real robot was also successful for the problem of dynamic walking on stepping stones with step lengths varied within [23 : 78](cm) and average walking speed of 0:6(m=s). In order to address the problems of robust control and safety-critical control in a unified control framework, we present a novel method of optimal robust control through a quadratic program that offers tracking stability while subject to input and state-based constraints as well as safety-critical constraints for nonlinear dynamical robotic systems under significant model uncertainty. The proposed method formulates robust control Lyapunov and barrier functions to provide guarantees of stability and safety in the presence of model uncertainty. We evaluate our proposed control design on different applications ranging from a single-link pendulum to dynamic walking of bipedal robot subject to contact force constraints as well as safety-critical precise foot placements on stepping stones, all while subject to significant model uncertainty. We conduct preliminary experimental validation of the proposed controller on a rectilinear spring-cart system under different types of model uncertainty and perturbations. To solve this problem, we also present another solution of adaptive CBF-CLF controller, that enables the ability to adapt to the effect of model uncertainty to maintain both stability and safety. In comparison with the robust CBF-CLF controller, this method not only can handle a higher level of model uncertainty but is also less aggressive if there is no model uncertainty presented in the system.

Samhället och världen vi lever i är föränderlig, kanske mer än någonsin. Det finns en strävan efter tålighet och snabb återhämtning för att på ett så bra sätt som möjligt hantera händelser och förändringar som vi inte kunnat förutse. Robusta samhällen kan anses framstå som önskvärda i en föränderlig värld. Då samhällen består av människor kan det antas att robusta människor skapar goda förutsättningar för robusta samhällen. Syftet med denna studie är att få mer kännedom om hur en robust människa upplevs och om det finns likheter i beskrivningarna av ett robust samhälle och en robust människa. Studien genomfördes i form av tre semistrukturerade intervjuer med tolkande fenomenologisk analys. Tre stycken professionella, som i sin profession mött totalt ca 3000 människor som varit med om större händelser och/eller kriser, intervjuades. Studiens resultat visar att upplevelsen av en robust människa är kärleksfull, ansvarsfull i nuet och känslomedveten: Kärleksfull till sig själv, andra och livet. Ansvarsfull och i nuet, oavsett konsekvenserna Medveten om, och känner, sina känslor och gränser. Studien berör även likheter och skillnader mellan begreppen robust, resiliens, resilient och Känsla Av SAMmanhang (KASAM). Vidare visas att det finns likheter mellan hur robusta samhällen och hur robusta människor beskrivs. Robust samhälle          Robust människa Ekologisk robusthet        Strävar efter god hälsa. Social robusthet             Kan skilja på vad som är sina och andras känslor. Teknisk robusthet           Är kvar i sig själv vid yttre störningar. Studien kommer fram till att robusta människor troligen är en förutsättning för robusta samhällen.
The society and the world we live are in constant change, maybe more now than ever. There is an aim to be robust and to recover fast to be able meet unpredictable events and changes. Robust societies seem desirable in a changing world. Societies consist of individuals and robust individuals can be assumed to create good conditions for a robust society. The purpose of this study is to gain more knowledge regarding how a robust individual is perceived and if there are similarities in the way a robust society and a robust individual are described. The study was conducted using three semi-structured interviews and interpretive phenomenological analysis. Three professionals, who had meet in total about 3000 individuals who had experienced large events and/or crisis, were interviewed. The result of the study demonstrates the experience of a robust individual being loving, responsible in the present and aware of feelings. Loving to himself/herself, others and life. Responsible and in the present, disregarding the consequences Aware of, and feels, his/hers feelings and boundaries. Similarities and differences between robust, resilience, resilient and Sense Of Coherence (SOC) are briefly covered in the study. There are similarities between how a robust society and a robust individual are described. Robust society           Robust individual Ecological robustness    Aims at good health. Social robustness          Able to distinguish between his/her own feelings and others feelings. Technical robustness     Stays as himself/herself when disturbed. The result of the study shows that robust individuals are probably a prerequisite for a robust society.

Thesis: M. Eng., Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science, 2016.
Cataloged from PDF version of thesis.
Includes bibliographical references (pages 85-88).
This work presents a soft hand capable of robustly grasping and identifying objects based on internal state measurements along with a combined system which autonomously performs grasps. A highly compliant soft hand allows for intrinsic robustness to grasping uncertainties; the addition of internal sensing allows the configuration of the hand and object to be detected. The hand can be configured in different ways using finger unit modules. The finger module includes resistive force sensors on the fingertips for contact detection and resistive bend sensors for measuring the curvature profile of the finger. The curvature sensors can be used to estimate the contact geometry and thus to distinguish between a set of grasped objects. With one data point from each finger, the object grasped by the hand can be identified. A clustering algorithm to find the correspondence for each grasped object is presented for both enveloping grasps and pinch grasps. This hand is incorporated into a full system with vision and motion planning on the Baxter robot to autonomously perform grasps of objects placed on a table. This hand is a first step towards proprioceptive soft grasping.
by Bianca Homberg.
M. Eng.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Sloan School of Management, Operations Research Center, 2004.
Includes bibliographical references (p. 169-171).
We propose new methodologies in robust optimization that promise greater tractability, both theoretically and practically than the classical robust framework. We cover a broad range of mathematical optimization problems, including linear optimization (LP), quadratic constrained quadratic optimization (QCQP), general conic optimization including second order cone programming (SOCP) and semidefinite optimization (SDP), mixed integer optimization (MIP), network flows and 0 - 1 discrete optimization. Our approach allows the modeler to vary the level of conservatism of the robust solutions in terms of probabilistic bounds of constraint violations, while keeping the problem tractable. Specifically, for LP, MIP, SOCP, SDP, our approaches retain the same complexity class as the original model. The robust QCQP becomes a SOCP, which is computationally as attractive as the nominal problem. In network flows, we propose an algorithm for solving the robust minimum cost flow problem in polynomial number of nominal minimum cost flow problems in a modified network. For 0 - 1 discrete optimization problem with cost uncertainty, the robust counterpart of a polynomially solvable 0 - 1 discrete optimization problem remains polynomially solvable and the robust counterpart of an NP-hard o-approximable 0-1 discrete optimization problem, remains a-approximable.
(cont.) Under an ellipsoidal uncertainty set, we show that the robust problem retains the complexity of the nominal problem when the data is uncorrelated and identically distributed. For uncorrelated, but not identically distributed data, we propose an approximation method that solves the robust problem within arbitrary accuracy. We also propose a Frank-Wolfe type algorithm for this case, which we prove converges to a locally optimal solution, and in computational experiments is remarkably effective.
by Melvyn Sim.
Ph.D.

Detta arbete behandlar framtagning av en robust sköljtank till en diskmaskin för grovdisk. Sköljtanken har en viktig roll i diskprocessen med ansvarar för att säkerställa att eventuella bakterier, diskmedel och matrester rensas i slutprocessen av diskningen. Därför ställs krav på produkten genom olika standarder för att vara godkänd för användning i diskmaskinen. Kraven berör främst vattentemperaturen samt dimensioner på ett bräddavlopp, som förhindrar att vatten från sköljtanken återströmmas till inloppsvattnet. Syftet med arbetet är att eliminera problem med överhettade värmeelment, opålitliga nivåvakter och korrosion som finns i befintlig sköljtank. Detta uppnås med framtagning av ett nytt sköljtanksystem genom en systematisk produktutvecklingsmetodik som inkluderar tillverkningsanpassad konstruktion (DFM) och miljöanpassad design (DFE). Arbetet resulterar i ett nytt system för sköljtanken med ett ovalformat tvärsnitt och beröringsfria nivåvakter. Detta nya system har ett värmeelement som alltid är täckt med vatten, ger tillförlitlig nivåmätning samt god härdighet mot korrosion. Andra viktiga förbättringar är lägre totalkostnad samt en miljöanpassad design. Det nya sköljtanksystemet lever upp till uppställda krav.
This paper is focusing on the development of a robust rinse tank in a dishwashing machine made for pot washing. The rinse tank plays an important role in the dishwashing process, that is to eliminate any bacteria, dish soap or left-overs, in the last step in the finishing process. Therefore demands on the product is set up, based on the required standards. The demands moststly affect the temperature of the water and the dimensions of an overflow, that will prevent the water from flowing back into the supply water pipelines. The purpose of this paper is to eliminate the problems that occur in the existing rinse tank, such as overheated heat element, unreliable level indicators and corrosion. To achieve this, a new robust rinse tank system is developted, by using an systematic product development method that includes design for manufacturing (DFM) and design for environment (DFE). The work results in a new system for the rinse tank, with a oval-shaped cross-section and contactless level sensors. This new system has a heat element that always is underwater, gives a reliable level indication and has a good resistance against corrosion. Other important improvements are a lower total cost together with a environmental design. The new rinse tank system satisfies the set demands.

Vibration generated from main rotor unbalance has a major impact on the aerospace gas turbine industry. A combination of logistical and technological drivers differentiate aerospace gas turbines and their derivatives from other types of gas turbines with respect to the approach for managing vibration, leading to a bespoke approach for the design, build, and balancing. The key drivers are: limited accessibility to rotors within the engine, the requirement for exchangeable modules (sub-assemblies) of major parts of the rotor without rebalancing, the use of only low-speed balancing on pseudo-rigid rotors with bladed assemblies, very low vibration limits leading to very tight balancing limits, extreme weight limits for design solutions, and the need for highly accurate, repeatable, and stable rotor joints. This study proposes and demonstrates a novel and rapid robust design system that has been created to deal with these unique challenges. The system comprises an overarching process and a set of novel tools and methods that have been created to support the process. These tools comprise an Unbalance Response Function (URF) design method that effectively delivers a preliminary design assessment and a fast Monte-Carlo simulation and comparison method with supporting software for comparing and improving build and balance design solutions. The aim of the overall process is to generate a system that identifies and controls critical parameters, and alleviates time wasted controlling non-critical parameters. The target outcome is therefore the most cost effective, predictable and repeatable solution with respect to rotor generated vibration (i.e. robust). Two novel methods of informing and improving the outcome of the low-speed balancing process using extra information available about the rotor are also introduced.

Robots would be much more useful if they could be more robust. Systems that can tolerate variability and uncertainty are called robust and the design of robust feedback controllers is a difficult problem that has been extensively studied for the past several decades. In this thesis, we aim to provide a quantitative measure of performance and robustness in control design under an optimization framework, producing controllers robust against parametric system uncertainties, external disturbances, and unmodeled dynamics. Under the H1 framework, we formulate the nonlinear robust control problem as a noncooperative, two-player, zero-sum, differential game, with the Hamilton-Jacobi-Isaacs equation as a necessary and sufficient condition for optimality. Through a spectral approximation scheme, we develop approximate algorithms to solve this differential game on the foundation of three ideas: global solutions through value function approximation, local solutions with trajectory optimization, and the use of multiple models to address unstructured uncertainties. Our goal is to introduce practical algorithms that are able to address complex system dynamics with high dimensionality, and aim to make a novel contribution to robust control by solving problems on a complexity scale untenable with existing approaches in this domain. We apply this robust control framework to the control of humanoid robots and manipulation in tasks such as operational space control and full-body push recovery.

Les robots mobiles à roues orientables gagnent de la mobilité en employant des roues conventionnelles entièrement orientables, comportant deux joints actifs, un pour la direction et un autre pour la conduite. En dépit d'avoir seulement un degré de mobilité (DOM) (défini ici comme degrés de liberté instantanément autorisés DOF), correspondant à la rotation autour du centre de rotation instantané (ICR), ces robots peuvent effectuer des trajectoires planaires complexes de $ 2D $. Ils sont moins chers et ont une capacité de charge plus élevée que les roues non conventionnelles (par exemple, Sweedish ou Omni-directional) et, en tant que telles, préférées aux applications industrielles. Cependant, ce type de structure de robot mobile présente des problèmes de contrôle textit {basic} difficiles de la coordination de la direction pour éviter les combats d'actionneur, en évitant les singularités cinématiques (ICR à l'axe de la direction) et les singularités de représentation (du modèle mathématique). En plus de résoudre les problèmes de contrôle textit {basic}, cette thèse attire également l'attention et présente des solutions aux problèmes de textit {niveau d'application}. Plus précisément, nous traitons deux problèmes: la première est la nécessité de reconfigurer "de manière discontinue" les articulations de direction, une fois que la discontinuité dans la trajectoire du robot se produit. Une telle situation - la discontinuité dans le mouvement du robot - est plus susceptible de se produire de nos jours, dans le domaine émergent de la collaboration homme-robot. Les robots mobiles qui fonctionnent à proximité des travailleurs humains en mouvement rapide rencontrent généralement une discontinuité dans la trajectoire calculée en ligne. Le second apparaît dans les applications nécessitant que l'angle de l'angle soit maintenu, certains objets ou fonctionnalités restent dans le champ de vision (p. Ex., Pour les tâches basées sur la vision) ou les changements de traduction. Ensuite, le point ICR est nécessaire pour déplacer de longues distances d'un extrême de l'espace de travail à l'autre, généralement en passant par le centre géométrique du robot, où la vitesse du robot est limitée. Dans ces scénarios d'application, les contrôleurs basés sur l'ICR à l'état de l'art conduiront à des comportements / résultats insatisfaisants. Dans cette thèse, nous résolvons les problèmes de niveau d'application susmentionnés; à savoir la discontinuité dans les commandes de vitesse du robot et une planification meilleure / efficace pour le contrôle du mouvement du point ICR tout en respectant les limites maximales de performance des articulations de direction et en évitant les singularités cinématiques et représentatives. Nos résultats ont été validés expérimentalement sur une base mobile industrielle
Steerable wheeled mobile robots gain mobility by employing fully steerable conventional wheels, having two active joints, one for steering, and another for driving. Despite having only one degree of mobility (DOM) (defined here as the instantaneously accessible degrees of freedom DOF), corresponding to the rotation about the instantaneous center of rotation (ICR), such robots can perform complex $2D$ planar trajectories. They are cheaper and have higher load carrying capacity than non-conventional wheels (e.g., Sweedish or Omni-directional), and as such preferred for industrial applications. However, this type of mobile robot structure presents challenging textit{basic} control issues of steering coordination to avoid actuator fighting, avoiding kinematic (ICR at the steering joint axis) and representation (from the mathematical model) singularities. In addition to solving the textit{basic} control problems, this thesis also focuses attention and presents solutions to textit{application level} problems. Specifically we deal with two problems: the first is the necessity to "discontinuously" reconfigure the steer joints, once discontinuity in the robot trajectory occurs. Such situation - discontinuity in robot motion - is more likely to happen nowadays, in the emerging field of human-robot collaboration. Mobile robots working in the vicinity of fast moving human workers, will usually encounter discontinuity in the online computed trajectory. The second appears in applications requiring that some heading angle is to be maintained, some object or feature stays in the field of view (e.g., for vision-based tasks), or the translation verse changes. Then, the ICR point is required to move long distances from one extreme of the workspace to the other, usually passing by the robot geometric center, where the feasible robot velocity is limited. In these application scenarios, the state-of-art ICR based controllers will lead to unsatisfactory behavior/results. In this thesis, we solve the aforementioned application level problems; namely discontinuity in robot velocity commands, and better/efficient planning for ICR point motion control while respecting the maximum steer joint performance limits, and avoiding kinematic and representational singularities. Our findings has been validated experimentally on an industrial mobile base

L’utilisation des AUV pour une exploitation durable des ressources océaniques est pertinente. Un robot sous-marin peut être utilisé comme plateforme pour observer, recueillir des informations sur l’environnement marin. Afin d’améliorer la qualité des observations et d’augmenter la capacité de navigation, de nombreuses questions doivent être abordées et examinées simultanément. Nous abordons ici le problème du pilotage de ces robots autonomes.Atteindre la maniabilité nécessaire dépend de deux facteurs clés: un modèle hydrodynamique précis et un système de contrôle performant. Cependant, le coût de développement d’un modèle précis est généralement très élevé. De plus, lorsque la géométrie du robot est complexe, il devient très difficile d’identifier de manière pertinente les paramètres dynamiques et hydrodynamiques. En outre, du point de vue de la commande, les modèles obtenus sont non linéaire, en particuliers pour les amortissements.De nombreux phénomènes dynamiques ne sont pas modélisés: dynamiques internes au robot, environnementales, liées aux bruits des capteurs, aux retards intrinsèques. Dans les concours de robotique sous-marine, il est confirmé que le traditionnel régulateur Proportionnel-Intégral-Dérivé (PID) est peu efficace pour les robots légers. Dans ce cas, notre champ d’application est plus axé sur la combinaison des approches de modélisation numérique et la commande robuste.Dans ce travail, nous proposons un schéma de régulation basé sur la commande robuste et la modélisation. La régulation robuste a été mise en place et validée en mer sur un AUV de la marque CISCREA et la solution proposée utilise Computational Fluid Dynamic (CFD) pour caractériser les deux paramètres hydrodynamiques (matrice de masse ajoutée et matrice d’amortissement). Puis un modèle à quatre degrés de liberté est construit pour le CISCREA. Les résultats numériques et expérimentaux sont alors comparés.La commande robuste proposée est basée sur une compensation non linéaire et de la commande H∞. La validation de la robustesse a été testée par simulation en Matlab et finalement validée par des essais en mer à Brest. La simulation et l’expérience montrent que l’approche en plus d’être robuste est plus rapide que les régulateurs précédemment proposés
Autonomous Underwater Vehicle (AUV) is a relevant technology for the sustainable use of ocean resources. AUV can be used as an important ocean observing platform to collect information on marine environmental characteristics for research and industry fields. In order to improve the observation quality and increase the navigation ability, many issues should be addressed and considered simultaneously. Achieve necessary maneuverability depends on two key factors: an accurate hydrodynamic model and an advanced control system. However, the cost to develop an accurate hydrodynamic model, which shrinks the uncertainty intervals, is usually high. Meanwhile, when the robot geometry is complex, it becomes very difficult to identify its dynamic and hydrodynamic parameters. In addition, according to the quadratic damping factor, underwater vehicle dynamic and hydrodynamic model is nonlinear from the control point of view. Moreover, unmodeled dynamics, parameter variations and environmental disturbances create significant uncertainties among the nominal model and the reality. Sensor noise, signal delay as well as unmeasured states also affect the stability and control performance of the motion control system. In many of our underwater competitions, it has been confirmed that the traditional Proportional-Integral-Derivative (PID) regulation is less efficient for low mass AUV. In this case, our scope is more focused on the combination of numerical modeling approaches and robust control schemes. In this work, we proposed a model based robust motion control scheme. Without loss of generality, a robust heading controller was implemented and validated in the sea on cubic-shaped CISCREA AUV. The proposed solution uses cost efficient Computational Fluid Dynamic (CFD) software to predict the two hydrodynamic key parameters: The added mass matrix and the damping matrix. Four Degree of Freedom (DOF) model is built for CISCREA from CFD calculation. Numerical and experimental results are compared. Besides, the proposed control solution inherited the numerically obtained model from previous CFD calculation. Numerically predicted the actuator force compensates the nonlinear damping behavior result in a linear model with uncertainties. Based on the bounded linear nominal model, we proposed H∞ approach to handle the uncertainties, we used kalman filter to estimate unmeasured states such as angular velocity and we developed smith compensator to compensate the sensor signal delay. The proposed robust heading control application uses only one compass as feedback sensor. This is important while AUV is working at certain depth where only magnetic sensors still work. Our robust control scheme was simulated in Matlab and validated in the sea near Brest. Simulation shows obvious advantage of the proposed robust control approach. Meanwhile, the proposed robust heading control is much faster than PID controller. The robust controller is insensitive to uncertainties and has no overshot. From both simulations and real sea experiments, we found our proposed robust control approach and the one compass heading control applications are efficient for low mass and complex-shaped AUV CISCREA

A central problem in control system design is how to design a controller to guarantee that the closed-loop system is robustly stable and that performance requirements are satisfied despite the presence of model uncertainties and exogenous disturbance signals. The analysis problem, that is the assessment of control systems with respect to robust stability and robust performance, can be adequately solved using the structured singular value u as introduced by Doyle. The corresponding design problem (how to choose a controller K to minimize u) is still largely unsolved, but an approximate solution can be found using Doyle's D - K iteration. In this thesis we present an alternative algorithm, called u - K iteration, which works by flattening the structured singular value u over frequency. As a prelude to this a classical loop shaping approach to robust performance is presented for SISO systems, and is also based on flattening u. In u-synthesis it is often the case that real uncertainties are modelled as complex perturbations but the conservatism so introduced can be severe. On the other hand, if real uncertainties are modelled as real perturbations then D - K iteration is not relevant. It is shown that u - K iteration still works for real perturbations. In addition, a geometric approach for computing the structured singular value for a scalar problem with respect to real and/or complex uncertainty is described. This provides insight into the relationship between real u and complex u. A robust performance problem is considered for a 2-input 2-output high purity distillation column which is an ill-conditioned plant. Analysis reveals the potentially damaging effects on robustness of ill-conditioning. A design is carried out using u - K iteration and the "optimum" u compared with that obtained by Doyle and by Freudenberg for the same problem.

Le domaine de recherche de cette thèse est l'optimisation linéaire robuste en deux étapes. Nous sommes intéressés par des algorithmes d'exploration de sa structure et aussi pour ajouter des alternatives afin d'atténuer le conservatisme inhérent à une solution robuste. Nous développons des algorithmes qui incorporent ces alternatives et sont personnalisés pour fonctionner avec des exemples de problèmes à moyenne ou grande échelle. En faisant cela, nous expérimentons une approche holistique du conservatisme en optimisation linéaire robuste et nous rassemblons les dernières avancées dans des domaines tels que l'optimisation robuste basée sur les données, optimisation robuste par distribution et optimisation robuste adaptative. Nous appliquons ces algorithmes dans des applications définies du problème de conception / chargement du réseau, problème de planification, problème combinatoire min-max-min et problème d'affectation de la flotte aérienne. Nous montrons comment les algorithmes développés améliorent les performances par rapport aux implémentations précédentes
The research scope of this thesis is two-stage robust linear optimization. We are interested in investigating algorithms that can explore its structure and also on adding alternatives to mitigate conservatism inherent to a robust solution. We develop algorithms that incorporate these alternatives and are customized to work with rather medium or large scale instances of problems. By doing this we experiment a holistic approach to conservatism in robust linear optimization and bring together the most recent advances in areas such as data-driven robust optimization, distributionally robust optimization and adaptive robust optimization. We apply these algorithms in defined applications of the network design/loading problem, the scheduling problem, a min-max-min combinatorial problem and the airline fleet assignment problem. We show how the algorithms developed improve performance when compared to previous implementations

This thesis presents a vision-based interface for human-robot interaction and control for autonomous robots in arbitrary environments. Vision has the advantage of being a low-power, unobtrusive sensing modality. The advent of robust algorithms and a significant increase in computational power are the two most significant reasons for such widespread integration. The research presented in this dissertation looks at visual sensing as an intuitive and uncomplicated method for a human operator to communicate in close-range with a mobile robot. The array of communication paradigms we investigate includes, but are not limited to, visual tracking and servoing, programming of robot behaviors with visual cues, visual feature recognition, mapping and identification of individuals through gait characteristics using spatio-temporal visual patterns and quantifying the performance of these human-robot interaction approaches. The proposed framework enables a human operator to control and program a robot without the need for any complicated input interface, and also enables the robot to learn about its environment and the operator using the visual interface. We investigate the applicability of machine learning methods – supervised learning in particular – to train the vision system using stored training data. A key aspect of our work is a system for human-robot dialog for safe and efficient task execution under uncertainty. We present extensive validation through a set of human-interface trials, and also demonstrate the applicability of this research in the field on the Aqua amphibious robot platform in the under water domain. While ourframework is not specific to robots operating in the under water domain, vision under water is affected by a number of issues, such as lighting variations and color degradation, among others. Evaluating the approach in such difficult operating conditions provides a definitive validation of our approach.
Cette thèse présentera une interface basée sur la vision qui permet l'intéraction entre humains et robots et aussi le control de robots autonomes parcourant des environments indéfinis. La vision à l'avantage d'être une modalité sensorielle discrète et à faible puissance. La probabilité d'algorithmes complexes et une hausse significative de puissance computationelle sont deux des raisons les plus importantes d'en faire une intégration si répandue. La recherche présentée dans cette dissertation évalue la détection visuelle comme méthode simple et intuitive pour un opérateur humain de communiquer à courte portée avec un robot mobil. L'ensemble des modèles communicationnels étudiés inclus, sans tous les nommés, la localisation et l'inspection visuelle, l'utilisation de signaux visuels pour la programmation comportemental de robots, la reconnaissance visuelle, la reconnaissance d'individus par leurs mouvements corporels caractéristiques utilisant des motifs visuels spatio-temporels tout en quantifiant la performance de cette approche à l'intéraction entre humains et robots. La structure proposée permet à l'opérateur humain de programmer et contôler un robot sans la nécessité d'une interface à entrée de données complexe. Cette structure permet aussi au robot de reconnaître des caractéristiques clés de son environment et de son opérateur humain par l'ulisation d'une interface visuelle. L'étude de l'appplication possible des méthodes d'apprentissage ulitilisées par certaines machines, toujours sous supervision, permet d'entraîner le système visuel à utiliser ses bases de données. Un aspect important de cette recherche est l'élaboration d'un système de dialogues entre humains et robots permettant l'exécution sécuritaire et efficace de tâches aux délimitations incertaines. On présente une ample validation à travers de nombreux essais utilisant notre interface avec l'aide de cobayes humains. On démontre aussi les applications possibles de cette recherche au sein des utilisations aquatiques du Aqua, robot amphibien à plateforme. Alors que notre structure de recherche ne se spécialise pas dans la robotique aquatique, la vision sous l'eau est toujours affectée par de nombreux facteurs, notamment la lumunosité variante et la dégradation de couleur. L'évaluation de l'approche nécessaire dans de telles conditions opérationnelles difficiles crée une validation définitive de notre recherche.

Récemment, le robot dirigeable a attiré l'attention de plus en plus des chercheurs grâce à ses avantages par rapport à d'autres aéronefs, tels que la capacité de VTOL, le vol stationnaire et à basse vitesse, une grande autonomie, et une interaction Homme-Robot sûre, etc. Ainsi c'est une plate-forme idéale pour diverses applications d'intérieur. Dans cette thèse, nous étudions la modélisation et le contrôle du mouvement d'un robot dirigeable d'intérieur et développons un prototype pour les opérations intérieures comme la surveillance. Le travail est composé de parties théoriques et pratiques. Concernant la partie théorique, d’abord, sous des hypothèses raisonnables, le modèle dynamique à 6-DOF est simplifié et divisé en deux parties indépendantes: le mouvement de l’altitude et le mouvement dans le plan horizontal. Ensuite, à fin d'assurer la précision de la modélisation et du contrôle, le modèle nominal est complété par des termes de perturbation qui sont estimés en temps réel et compensés dans les contrôleurs conçus. Des simulations sont effectuées pour vérifier les performances et la robustesse des contrôleurs. Pour la partie pratique du travail, basée sur l'analyse des fonctionnalités du robot afin de réaliser les applications intérieures souhaitées, le matériel du robot dirigeable est conçu et créé. Enfin, de vrais tests sont effectués sur la plate-forme de robot dirigeable pour la validation des lois de contrôle de mouvement conçues, et des résultats satisfaisants sont obtenus
Recently, the blimp robot has attracted more and more attentions of the researchers for its advantages compared to other aircrafts, such as ability for VTOL, stationary and low speed flight, long endurance in air and safe Human-Robot interaction, etc. Therefore it is an ideal platform for various indoor applications. In this thesis, we study the modeling and motion control of an indoor blimp robot, and develop a real robot for indoor operations such as the long-term surveillance. The work is composed of both theoretical and practical parts. For the theoretical part, first, under reasonable assumptions, the 6-DOF dynamic model is simplified and divided into two independent parts: the altitude motion and the horizontal plane movement. Then, to ensure the accuracy of modeling and control, the nominal model is complemented with disturbance terms which are estimated in real-time and compensated in the designed controllers. Simulations are carried out to verify the performance and robustness of the controllers. For the practical part of the work, based on the functionality analysis of the robot to achieve desired indoor applications, the hardware of the blimp robot is conceived and created. Finally, real tests are made on the blimp robot platform for the validation of the designed motion control laws, and satisfying results are obtained

One of the most widely used techniques for data clustering is agglomerative clustering. Such algorithms have been long used across any different fields ranging from computational biology to social sciences to computer vision in part because they are simple and their output is easy to interpret. However, many of these algorithms lack any performance guarantees when the data is noisy, incomplete or has outliers, which is the case for most real world data. It is well known that standard linkage algorithms perform extremely poorly in presence of noise. In this work we propose two new robust algorithms for bottom-up agglomerative clustering and give formal theoretical guarantees for their robustness. We show that our algorithms can be used to cluster accurately in cases where the data satisfies a number of natural properties and where the traditional agglomerative algorithms fail. We also extend our algorithms to an inductive setting with similar guarantees, in which we randomly choose a small subset of points from a much larger instance space and generate a hierarchy over this sample and then insert the rest of the points to it to generate a hierarchy over the entire instance space. We then do a systematic experimental analysis of various linkage algorithms and compare their performance on a variety of real world data sets and show that our algorithms do much better at handling various forms of noise as compared to other hierarchical algorithms in the presence of noise.

ABSTRACT ROBUST CONTROL CHARTS Ç
etinyü
rek, Aysun M. Sc., Department of Statistics Supervisor: Dr. BariS Sü
rü
cü
Co-Supervisor: Assoc. Prof. Dr. Birdal Senoglu December 2006, 82 pages Control charts are one of the most commonly used tools in statistical process control. A prominent feature of the statistical process control is the Shewhart control chart that depends on the assumption of normality. However, violations of underlying normality assumption are common in practice. For this reason, control charts for symmetric distributions for both long- and short-tailed distributions are constructed by using least squares estimators and the robust estimators -modified maximum likelihood, trim, MAD and wave. In order to evaluate the performance of the charts under the assumed distribution and investigate robustness properties, the probability of plotting outside the control limits is calculated via Monte Carlo simulation technique.

This thesis describes discrete robust adaptive control methods based on using slow sampling and slow adaptation. For the stability analysis, we consider that the plant model order is not exactly known and assume that the estimation model order is lower than the plant model order. A stability condition is derived with a given upper limit for the adaptation gain which is related to a strictly positive real operator. Discussion of the relation between sampling and stability condition is then given. For the robust adaptive control design, we study slow adaptation and predictive control. For the slow adaptation, the main idea is to use only good estimates and use a compensation filter. Some frequency domain information on the plant is necessary for this method. For predictive control, we discuss the relationship between the extended control horizon and the critical sampling. For a simple case, it is shown that the larger extended control horizon brings more robust adaptive control. The purpose of this thesis is to provide robust discrete adaptive controller design guidelines, especially in such cases as using slow sampling frequency, slow adaptation rate. It is true that in practice, for various discrete adaptive control algorithms, slow sampling and slow adaptation rate will bring more robustness. The use of slow sampling and slow adaptation rate is simple and economic, thus a careful choice of sampling rate and adaptation rate is highly recommended. This thesis provides such guidelines for choosing proper sampling rate and adaptation rate for robust discrete adaptive control.
Applied Science, Faculty of
Electrical and Computer Engineering, Department of
Graduate

The prevalence of extreme outliers in many regression data sets has led to the development of robust methods that can handle these observations. While much attention has been placed on the problem of estimating regression coefficients in the presence of outliers, few methods address variable selection. We develop and study robust versions of the forward selection algorithm, one of the most popular standard variable selection techniques. Specifically we modify the VAMS procedure, a version of forward selection tuned to control the false selection rate, to simultaneously select variables and eliminate outliers. In an alternative approach, robust versions of the forward selection algorithm are developed using the robust forward addition sequence associated with the generalized score statistic. Combining the robust forward addition sequence with robust versions of BIC and the VAMS procedure, a final model is obtained. Monte Carlo simulation compares these robust methods to current robust methods like the LSA and LAD-LASSO. Further simulation investigates the relationship between the breakdown point of the estimation methods central to each procedure and the breakdown point of the final variable selection method.

Robust network design takes the very successful framework of robust optimization and applies it to the area of network design, motivated by applications in communication networks. The main premise is that demands across the network are not fixed, but are variable or uncertain. However, they are known to fall within a prescribed uncertainty set. Our solution must have sufficient capacity to route any demand in this set; moreover, the routing must be oblivious, meaning it must be fixed up front, and not depend on the particular choice of demand from within the uncertainty set. A particular choice of uncertainty set within this framework yields the "hose model", which has received particular attention due to applications to virtual private networks. A 2-approximation was known for the problem, using a solution template in the form of a tree. It was conjectured that this tree solution is actually always optimal; this became known as the "VPN Conjecture". As one of the central results of this thesis, we prove this conjecture in full generality. In addition, we demonstrate a counterexample to a stronger multipath (fractional routing) version of the conjecture which had also been proposed. We initiate a study of the robust network design problem more generally, with a focus on approximability. In the general model, where the uncertainty set is given by an arbitrary separable polyhedron, we give a strong inapproximability result. We then consider a new and natural model generalizing the symmetric hose model, based on demands routable on a given tree, and provide a constant factor approximation algorithm. Lastly, we compare oblivious routing with the much more flexible (but also less practical) dynamic routing scheme where the routing may vary depending on the demand pattern. We show that in the worst case, the cost of an optimal oblivious routing solution can be much more expensive than the dynamic optimum, by up to a logarithmic factor.
Motivé par les applications concernantes les réseaux de communication, le dessein des réseaux robustes applique les méthodes très réusies provenant de l'optimisation robuste. La prémisse principale est que les demandes sur le réseau ne sont pas fixes, mais variables ou incertaines. Cependant, nous savons qu'elles sont tirées d'un ensemble d'incertitude prescrit. Il faut que la solution ait une capacité suffisante pour pouvoir router toute demande appartenant à cet ensemble. En outre, il faut que le routage soit oublieux, ce qui signifie qu'il peut être fixé à l'avance, et ne dépends pas du choix particulier de la demande appartenant de l'ensemble d'incertitude. Dans ce cadre, il existe un choix particulier d'ensemble d'incertitude qui mène au « modèle de tuyau ». Ce modèle a reçu une attention particulière à causede ses applications aux réseaux privés virtuels. On connaissait un 2-rapprochement utilisant une solution en forme d'arbre. La « Conjecture de VPN » énonce que cette solution en forme d'arbre est toujours optimale. L'un des résultats principaux de cette thèse démontre cette conjecture en toute généralité. En outre, nous donnons un contre-exemple à une version plus forte de la conjecture concernant les chemins multiples (le routage étant fractionnel) qui avait également été proposée. Nous initions l'étude du problème de la conception de réseaux robustes dans une plus grande généralité, en insistant sur l'approximabilité. Dans le modèle général, où l'ensemble d'incertitude est un polyèdre séparable arbitraire, nous donnons un résultat fort d'inapproximabilité. Nous considérons ensuite un nouveau modèle naturel généralisant le modèle de tuyau symétrique, qui est basé sur des demandes qui peuvent être routées sur un arbre donné, et nous fournissons un algorithme ayant un facteur de rapprochement constant. Finalement, nous comparons le routage oublieux avec le schéma beaucoup plus flexible (

The problem of locating collisions between computer modeled moving rigid bodies is considered. Each rigid body is modeled by the union of polytopes. Safe and reliable collision detection algorithms are constructed.
Since the computer representation of real numbers is finite, an interval projective point is used to encompass all localization errors of a modeled point. An interval point is the elementary geometrical form from which all others are constructed. The Euclidean convex set spanned by an interval point is also a polytope.
The construction of a polytope relies on a robust convex hull algorithm. The computed hull is guaranteed to contain all interval projective points.
An extrusion based collision detection algorithm builds an AND-OR decision tree. Each leaf is a univariate function that expresses the relation between a moving point and a moving plane. Interval zero finding methods are applied to find the overlap and non-overlap portions of the trajectories of moving polytopes.
The swept-volume based collision detection algorithm depends on the construction of a convex approximation that comprises the real swept volume. To obtain a convex approximation, the convex hull algorithm is applied to the bounding volumes of the vertices of a moving polytope. Each bounding volume is an interval projective point.
Finally, both collision detection algorithms are tested in the context of generate-and-test path planning.

It is a challenging problem to make a general hand gesture recognition system work in a practical operation environment. In this study, it is mainly focused on recognizing English letters and digits performed near the steering wheel of a car and captured by a video camera. Like most human computer interaction (HCI) scenarios, the in-car gesture recognition suffers from various robustness issues, including multiple human factors and highly varying lighting conditions. It therefore brings up quite a few research issues to be addressed. First, multiple gesturing alternatives may share the same meaning, which is not typical in most previous systems. Next, gestures may not be the same as expected because users cannot see what exactly has been written, which increases the gesture diversity significantly.In addition, varying illumination conditions will make hand detection trivial and thus result in noisy hand gestures. And most severely, users will tend to perform letters at a fast pace, which may result in lack of frames for well-describing gestures. Since users are allowed to perform gestures in free-style, multiple alternatives and variations should be considered while modeling gestures. The main contribution of this work is to analyze and address these challenging issues step-by-step such that eventually the robustness of the whole system can be effectively improved. By choosing color-space representation and performing the compensation techniques for varying recording conditions, the hand detection performance for multiple illumination conditions is first enhanced. Furthermore, the issues of low frame rate and different gesturing tempo will be separately resolved via the cubic B-spline interpolation and i-vector method for feature extraction. Finally, remaining issues will be handled by other modeling techniques such as sub-letter stroke modeling. According to experimental results based on the above strategies, the proposed framework clearly improved the system robustness and thus encouraged the future research direction on exploring more discriminative features and modeling techniques.

CONSELHO NACIONAL DE DESENVOLVIMENTO CIENTÍFICO E TECNOLÓGICO
Os mercados de ações dos países emergentes tem como principal característica a presença de dias atípicos, os quais fazem impossível o uso, com sucesso, dos modelos de equilíbrio. O principal objetivo desta pesquisa é a de desenvolver uma nova teoria, baseada na estatísica robusta, que possa ser aplicada a estes mercados sem ser seriamente afetada por observações extremas, e que ao mesmo tempo, ofereça resultados eficientes e precisos.
Emergent stock markets are characterized by the presence of atypical days, which make impossible the use of equilibrium models with sucess. The main aim of this research is to define a new theory, based on the robust statistical theory, which could be applied to those markets without being affected by extreme observations and, at the same time, offer efficient and accurate results.
Los mercados de acciones de los países emergentes tienen como principal característica la presencia de días atípicos, lo que hace imposible el uso, con suceso, de los modelos de equilibrio. EL objetivo principal de esta investigación es desarrollar una nueva teoría, basada en la estadística robusta, que pueda ser aplicada a estos mercados sin ser seriamente afectada por observaciones extremas, y que al mismo tiempo, obtenga resultados eficientes y precisos.

Digital image watermarking is used to protect the copyright of digital images. In this thesis, a novel blind logo image watermarking technique for RGB images is proposed. The proposed technique exploits the error correction capabilities of the Human Visual System (HVS). It embeds two different watermarks in the wavelet/multiwavelet domains. The two watermarks are embedded in different sub-bands, are orthogonal, and serve different purposes. One is a high capacity multi-bit watermark used to embed the logo, and the other is a 1-bit watermark which is used for the detection and reversal of geometrical attacks. The two watermarks are both embedded using a spread spectrum approach, based on a pseudo-random noise (PN) sequence and a unique secret key. Robustness against geometric attacks such as Rotation, Scaling, and Translation (RST) is achieved by embedding the 1-bit watermark in the Wavelet Transform Modulus Maxima (WTMM) coefficients of the wavelet transform. Unlike normal wavelet coefficients, WTMM coefficients are shift invariant, and this important property is used to facilitate the detection and reversal of RST attacks. The experimental results show that the proposed watermarking technique has better distortion parameter detection capabilities, and compares favourably against existing techniques in terms of robustness against geometrical attacks such as rotation, scaling, and translation.

In this thesis, our aim is to improve deep auto-encoders, an important topic in the deep learning area, which has shown connections to latent feature discovery models in the literature. Our model is inspired by robust principal component analysis, and we build an outlier filter on the top of basic deep auto-encoders. By adding this filter, we can split the input data X into two parts X=L+S, where the L could be better reconstructed by a deep auto-encoder and the S contains the anomalous parts of the original data X. Filtering out the anomalies increases the robustness of the standard auto-encoder, and thus we name our model ``Robust Auto-encoder'. We also propose a novel solver for the robust auto-encoder which alternatively optimizes the reconstruction cost of the deep auto-encoder and the sparsity of outlier filter in pursuit of finding the optimal solution. This solver is inspired by the Alternating Direction Method of Multipliers, Back-propagation and the Alternating Projection method, and we demonstrate the convergence properties of this algorithm and its superior performance in standard image recognition tasks. Last but not least, we apply our model to multiple domains, especially, the cyber-data analysis, where deep models are seldom currently used.

Thesis: M.B.A., Massachusetts Institute of Technology, Sloan School of Management, 2015. In conjunction with the Leaders for Global Operations Program at MIT.
Thesis: S.M., Massachusetts Institute of Technology, Department of Civil and Environmental Engineering, 2015. In conjunction with the Leaders for Global Operations Program at MIT.
Cataloged from PDF version of thesis.
Includes bibliographical references (pages 61-63).
As an equipment manufacturer for the mining and construction industries, Caterpillar's business is heavily impacted by the cyclicality of these two industries. To be competitive in the market while investing prudently to ensure shareholders' value, Caterpillar needs to make capacity decisions with peaks and troughs of the cycles in mind. Through the peaks and troughs of business cycles, the perceptions change from not enough capacity to under-utilization of asset constrained capacity respectively. The purpose of this thesis is to establish a framework that enhances robustness in investment decision-making within the 2-6 year capacity planning horizon. This entails the ability to balance risk and return for the company and account for the cyclical nature of business and other strategic considerations. The project also aims to enhance consistency across stakeholders and introduce an aggregate risk perspective under corporate capital constraint.
by Shui Ying Chua.
M.B.A.
S.M.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Sloan School of Management, Operations Research Center, 2007.
This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections.
Includes bibliographical references (p. 133-138).
We propose a novel robust optimization technique, which is applicable to nonconvex and simulation-based problems. Robust optimization finds decisions with the best worst-case performance under uncertainty. If constraints are present, decisions should also be feasible under perturbations. In the real-world, many problems are nonconvex and involve computer-based simulations. In these applications, the relationship between decision and outcome is not defined through algebraic functions. Instead, that relationship is embedded within complex numerical models. Since current robust optimization methods are limited to explicitly given convex problems, they cannot be applied to many practical problems. Our proposed method, however, operates on arbitrary objective functions. Thus, it is generic and applicable to most real-world problems. It iteratively moves along descent directions for the robust problem, and terminates at a robust local minimum. Because the concepts of descent directions and local minima form the building blocks of powerful optimization techniques, our proposed framework shares the same potential, but for the richer, and more realistic, robust problem.
(cont.) To admit additional considerations including parameter uncertainties and nonconvex constraints, we generalized the basic robust local search. In each case, only minor modifications are required - a testimony to the generic nature of the method, and its potential to be a component of future robust optimization techniques. We demonstrated the practicability of the robust local search technique in two realworld applications: nanophotonic design and Intensity Modulated Radiation Therapy (IMRT) for cancer treatment. In both cases, the numerical models are verified by actual experiments. The method significantly improved the robustness for both designs, showcasing the relevance of robust optimization to real-world problems.
by Kwong Meng Teo.
Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2009.
Cataloged from PDF version of thesis.
Includes bibliographical references (p. 135-138).
This thesis presents an approach for grasping objects robustly under significant positional uncertainty. In the field of robot manipulation there has been a great deal of work on how to grasp objects stably, and in the field of robot motion planning there has been a great deal of work on how to find collision-free paths to those grasp positions. However, most of this work assumes exact knowledge of the shapes and positions of both the object and the robot; little work has been done on how to grasp objects robustly in the presence of position uncertainty. To reason explicitly about uncertainty while grasping, we model the problem as a partially observable Markov decision process (POMDP). We derive a closed-loop strategy that maintains a belief state (a probability distribution over world states), and select actions with a receding horizon using forward search through the belief space. Our actions are world-relative trajectories (WRT): fixed trajectories expressed relative to the most-likely state of the world. We localize the object, ensure its reachability, and robustly grasp it at a specified position by using information-gathering, reorientation, and goal-seeking WRT actions. This framework is used to grasp objects (including a power drill and a Brita pitcher) despite significant uncertainty, using a 7-DOF Barrett Arm and attached 4-DOF Barrett Hand equipped with force and contact sensors. Our approach is generalizable to almost any sensor type, as well as wide ranges of sensor error and pose uncertainty.
by Kaijen Hsiao.
Ph.D.

Thesis (M. Eng.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2005.
Includes bibliographical references (p. 91-98).
In this thesis, we present various models of distributed computation and algorithms for these models. The underlying theme is to come up with fast algorithms that can tolerate faults in the underlying network. We begin with the classical message-passing model of computation, surveying many known results. We give a new, universally optimal, edge-biconnectivity algorithm for the classical model. We also give a near-optimal sub-linear algorithm for identifying bridges, when all nodes are activated simultaneously. After discussing some ways in which the classical model is unrealistic, we survey known techniques for adapting the classical model to the real world. We describe a new balancing model of computation. The intent is that algorithms in this model should be automatically fault-tolerant. Existing algorithms that can be expressed in this model are discussed, including ones for clustering, maximum flow, and synchronization. We discuss the use of agents in our model, and give new agent-based algorithms for census and biconnectivity. Inspired by the balancing model, we look at two problems in more depth.
(cont.) First, we give matching upper and lower bounds on the time complexity of the census algorithm, and we show how the census algorithm can be used to name nodes uniquely in a faulty network. Second, we consider using discrete harmonic functions as a computational tool. These functions are a natural exemplar of the balancing model. We prove new results concerning the stability and convergence of discrete harmonic functions, and describe a method which we call Eulerization for speeding up convergence.
by David Pritchard.
M.Eng.

Doctor of Philosophy
Department of Statistics
Weixin Yao and Kun Chen
Ordinary least-squares (OLS) estimators for a linear model are very sensitive to unusual values in the design space or outliers among y values. Even one single atypical value may have a large effect on the parameter estimates. In this proposal, we first review and describe some available and popular robust techniques, including some recent developed ones, and compare them in terms of breakdown point and efficiency. In addition, we also use a simulation study and a real data application to compare the performance of existing robust methods under different scenarios. Finite mixture models are widely applied in a variety of random phenomena. However, inference of mixture models is a challenging work when the outliers exist in the data. The traditional maximum likelihood estimator (MLE) is sensitive to outliers. In this proposal, we propose a Robust Mixture via Mean shift penalization (RMM) in mixture models and Robust Mixture Regression via Mean shift penalization (RMRM) in mixture regression, to achieve simultaneous outlier detection and parameter estimation. A mean shift parameter is added to the mixture models, and penalized by a nonconvex penalty function. With this model setting, we develop an iterative thresholding embedded EM algorithm to maximize the penalized objective function. Comparing with other existing robust methods, the proposed methods show outstanding performance in both identifying outliers and estimating the parameters.

Master of Science
Department of Statistics
Weixin Yao
In practice, when applying a statistical method it often occurs that some observations deviate from the usual model assumptions. Least-squares (LS) estimators are very sensitive to outliers. Even one single atypical value may have a large effect on the regression parameter estimates. The goal of robust regression is to develop methods that are resistant to the possibility that one or several unknown outliers may occur anywhere in the data. In this paper, we review various robust regression methods including: M-estimate, LMS estimate, LTS estimate, S-estimate, [tau]-estimate, MM-estimate, GM-estimate, and REWLS estimate. Finally, we compare these robust estimates based on their robustness and efficiency through a simulation study. A real data set application is also provided to compare the robust estimates with traditional least squares estimator.

Mestrado em Finanças
Este estudo empírico tem como objectivo avaliar o impacto da estimação robusta nos portefólios de média variância. Isto foi conseguido fazendo uma simulação do comportamento de 15 acções do SP500. Esta simulação inclui dois cenários: um com amostras que seguem uma distribuição normal e outro com amostras contaminadas não normais. Cada cenário inclui 200 reamostragens. O performance dos portefólios estimados usando a máxima verosimilhança (clássicos) e dos portefólios estimados de forma robusta são comparados, resultando em algumas conclusões: Em amostras normais, portefólios robustos são marginalmente menos eficientes que os portefólios clássicos. Contudo, em amostras não normais, os portefólios robustos apresentam um performance muito superior que os portefólios clássicos. Este acréscimo de performance está positivamente correlacionado com o nível de contaminação da amostra. Em suma, assumindo que os retornos financeiros têm uma distribuição não normal, podemos afirmar que os estimadores robustos resultam em portefólios de média variância mais estáveis.
This empirical study's objective is to evaluate the impact of robust estimation on mean variance portfolios. This was accomplished by doing a simulation on the behavior of 15 SP500 stocks. This simulation includes two scenarios: One with normally distributed samples and another with contaminated non-normal samples. Each scenario includes 200 resamples. The performance of maximum likelihood (classical) estimated portfolios and robustly estimated portfolios are compared, resulting in some conclusions: On normally distributed samples, robust portfolios are marginally less efficient than classical portfolios. However, on non-normal samples, robust portfolios present a much higher performance than classical portfolios. This increase in performance is positively correlated with the level of contamination present on the sample. In summary, assuming that financial returns do not present a normal distribution, we can state that robust estimators result in more stable mean variance portfolios.