Pasar al contenido principal

Resultados relacionados

Resultados totales (Incluyendo duplicados): 3
Encontrada(s) 1 página(s)

e-Archivo. Repositorio Institucional de la Universidad Carlos III de Madrid
Publicación Tesis doctoral (DoctoralThesis)

Model-based motor control learning in robotics

  • González-Quijano Álvarez, Javier
Robot motor control learning is currently a very active research area in robotics. The challenge consists in designing an efficient heuristic to guide a search process which will attempt to find and appropriate motor control policy capable of accomplishing a certain task. Implicitly or explicitly, the search is an optimization process of some specific objective function. This means, in practise, that the robot will have to evaluate potential candidate solutions in the real environment. Each evaluation is a trial-and-error execution that will end up with useful information to feed the heuristic that will decide which potential candidate solution will be evaluated in the next execution. Many learning techniques have been developed during past decades. A very general classification divides these methods into two groups: global learning and local learning methods. The state of art in local learning techniques is very developed. There are hundreds of examples where complex learning tasks can be achieved using these methods. Nevertheless, they all constraint the search space to make the learning process feasible. The result is always a local optimum of the objective function. On the other hand, global learning techniques are oriented to find global optimum solutions. This is generally a much harder problem as the search space is larger. Research on global learning methods is also large. However, there is a lack of methods with direct applicability in complex robotics systems without assuming prior knowledge about the task. This is mainly due to two facts. First, they scale badly to problems with continuous and high dimensional state-action spaces. Second, too many real robot-environment interactions are needed. Whatever the performance of these methods is, it seems to be clear that global methods based on model-based learning have clear advantages with respect to model-free methods, especially those related to the minimization of the number of trial-error executions. A complete study of the most relevant work done in motor control learning is performed in Chapter 2, where the advantages and disadvantages of the different existing methods are discussed. Nevertheless, this thesis puts the focus on overcoming the limitations on current global learning methods, concretely those following a model-based learning philosophy. In our opinion, they have more advantages. This statement will be justified during the reading of this chapter. As already mentioned, every learning problem is indeed an optimization process. Learning methods based on an explicit optimization process have several advantages when the complete search space can be represented by control policy functions depending on a reduce set of parameters. This occurs always when taking advantage of the existing controllers in literature. For these cases, a new learning method consisting in an optimization process using surrogate models has been proposed in Chapter 3. Some learning problems are very complex because there is any controller which a priori can solve the objective task. In these cases is very hard to parametrize any control policy and therefore it is necessary to search directly for a sequence of low-level motor control commands. Accomplishing this task is very difficult if seeking for optimal solutions. If optimality is not crucial, it is possible to take advantage of kinodynamic planning algorithms to solve the problem. However, planners need models, and models are not known, otherwise it would not be a learning problem. Nevertheless, this research has found the way to combine kinodynamic planning with model learning to manage an efficient new solution for learning problems from scratch (without assuming prior knowledge). This finding is described in Chapter 4 and the first part of Chapter 5. Despite the effectiveness of such solution, it still lacks of a ideal behaviour for its application in real environments as the provided control policies are open-loop. Nonetheless, it is possible to combine the output of such algorithm with the potential of model predictive control. This is done in the second part of such chapter. The result is a complete architecture capable of providing robust close-loop policies capable of operating efficiently in real environments., I would like to acknowledge the Spanish Government for providing me with a FPU fellowship award from Spanish Ministry of Education (MEC), which certainly has helped on the purpose of achieving the results of this research work. Furthermore, it is fair to specially mention the European project HANDLE funded within the Seventh Framework Programme (FP7)., Programa Oficial de Doctorado en Ingeniería Eléctrica, Electrónica y Automática, Presidente: Antonio Barrientos Cruz.- Secretario: Fares Jawad Moh D Abu-Dakka.- Vocal: Claudio Rossi
e-Archivo. Repositorio Institucional de la Universidad Carlos III de Madrid
Publicación Tesis doctoral (DoctoralThesis)

Modeling and control of an anthropomorphic robotic hand

  • Bensalah, Choukri
Mención Europea en el título de doctor, This thesis presents methods and tools for enabling the successful use of robotic hands. For highly dexterous and/or anthropomorphic robotic hands, these methods have to share some common goals, such as overcoming the potential complexity of the mechanical design and the ability of performing accurate tasks with low and efficient computational cost. A prerequisite for dexterity is to increase the workspace of the robotic hand. For this purpose, the robotic hand must be considered as a single multibody system. Solving the inverse kinematics problem of the whole robotic hand is an arduous task due to the high number of degrees of freedom involved and the possible mechanical limitations, singularities and other possible constraints. The redundancy has proven to be of a great usefulness for dealing with potential constraints. To be able to exploit the redundancy for dealing with constraints, the adopted method for solving the inverse kinematics must be robust and extendable. Obviously, addressing such complex problem, the method will certainly be computationally heavy. Thus, one of the aims of this thesis is to resolve the inverse kinematics problem of the whole robotic hand under constraints, taking into account the computational cost. To this end, this thesis extends and reduces the most recent Selectively Damped Least Squares method which is based on the computation of all singular values, to deal with constraints with a minimum computational cost. New estimation algorithm of singular values and their corresponding singular vectors is proposed to reduce the computational cost. The reduced extended selectively damped least squares method is simulated and experimentally evaluated using an anthropomorphic robotic hand as a test bed. On the other hand, dexterity depends not only on the accuracy of the position control, but also on the exerted forces. The tendon driven modern robotic hands, like the one used in this work, are strongly nonlinear dynamic systems, where motions and forces are transmitted remotely to the finger joints. The problem of modeling and control of position and force simultaneously at low level control is then considered. A new hybrid control structure based on the succession of two sliding mode controllers is proposed. The force is controlled by its own controller which does not need a contact model. The performance of the proposed controller is evaluated by performing the force control directly using the force sensor information of the fingertip, and indirectly using the torque control of the actuator. Finally, we expect that the applications of the methods presented in this thesis can be extended to cover different issues and research fields and in particular they can be used in a variety of algorithm that require the estimation of singular values., This work was partially supported by the European project HANDLE, FP7-231640, and by the Spanish ministry MICINN through FPI scholarship within the project DPI-2005-04302., Programa Oficial de Doctorado en Ingeniería Eléctrica, Electrónica y Automática, Presidente: Anis Sahbani.- Secretario: Fares Jawad Moh D Abu-Dakka.- Vocal: Claudio Rossi
e-Archivo. Repositorio Institucional de la Universidad Carlos III de Madrid
Publicación Tesis doctoral (DoctoralThesis)

Desarrollo y evaluación de estrategias para aplicaciones de percepción 3D usando cámaras de rango

  • Rodríguez Jiménez, Silvia
La evolución de sensores de profundidad de bajo coste en los últimos años ha permitido un uso más generalizado de las cámaras de rango. Su aplicación, gracias a que proporcionan información de profundidad en tiempo real, se ha ampliado a diversos ámbitos de percepción de objetos donde es necesario el conocimiento del entorno en tres dimensiones (3D) para poder interaccionar con él. Por este motivo, existe una demanda creciente de sistemas de percepción visual capaces de explotar la información procedente de los dispositivos de rango. Sin embargo, la mayoría de los enfoques existentes presentan limitaciones en entornos poco estructurados, comunes en el mundo real. El objetivo de esta tesis es contribuir al progreso de sistemas eficaces de percepción 3D de objetos en este tipo de entornos, abordando diversos retos aún sin resolver como la falta de información a priori o ruido en las imágenes. Para ello se ha diseñado una metodología que permite desarrollar estrategias para entornos complejos con el objetivo de dar solución a los desafíos de percepción planteados, explotando al máximo las características de la escena. Para validar la estrategia planteada se ha seguido una metodología guiada por escenarios, por lo que para su evaluación se han seleccionado tres entornos con diferente complejidad, y lo suficientemente significativos a la vez que complementarios entre sí. Cada contexto se enmarca dentro de un grupo de escenarios semiestructurados de percepción visual, clasificados según condiciones ambientales así como grado de información conocida a priori. Para cada escenario se propone una estrategia, la cual se desarrolla y se evalúa exhaustivamente en un contexto experimental, seleccionando el hardware así como el software a implementar más adecuado a los requisitos y necesidades del entorno. El enfoque descrito en esta tesis permite reducir la incertidumbre centrándose en un determinado contexto para plantear la estrategia a seguir, reduciendo el espacio de estados sin afectar a la tarea a desempeñar., The evolution of low cost depth sensors in recent years has allowed a more widespread use of range cameras. They provide depth information in real time so their application has been extended to various areas of object perception where three-dimensional (3D) knowledge of the environment is mandatory to interact with it. For this reason, there is a growing demand for visual perception systems to exploit the information from range devices. However, most existing approaches are limited in unstructured environments, which are common in the real world. The objective of this thesis is to contribute to the progress of 3D object perception in these environments, addressing open challenges like lack of prior information or picture noise. A methodology to develop strategies for complex environment has been designed to solve the challenges, exploiting the characteristics of the scene. The strategy is validated following a methodology based on scenarios, so three environments have been selected with different complexity. These are sufficiently significant and complementary. Each context is part of a group of perception unstructured scenarios, classified by environmental conditions and the degree of information known a priori. For each scenario, a strategy is proposed, which is developed and evaluated thoroughly in an experimental context. This has also taken into account that hardware and software should be appropriated to the requirements and environmental needs. The approach described in this thesis decreases uncertainty by focusing on a particular context. Furthermore the state space is reduced without affecting the task to be performed., Esta tesis ha sido parcialmente financiada por el proyecto europeo HANDLE, dentro del Séptimo Programa Marco de la Comunidad Europea en virtud del acuerdo de subvención ICT 231640. También ha sido parcialmente financiada por una beca de Formación de Personal Investigador de la Universidad Carlos III de Madrid (PIF-UC3M ref. 03-1213)., Programa Oficial de Doctorado en Ingeniería Eléctrica, Electrónica y Automática, Presidente: José María Sebastián y Zúñiga.- Secretario: Juan Antonio Escalera Piña.- Vocal: Óscar Reinoso García