Search

Hoy en día, las pequeñas o medianas empresas se encuentran en una situación complicada a la hora de dar el salto hacia la robotización de sus procesos industriales, ya sea por motivos económicos o por desconocimiento de estas tecnologías. Es por eso que este sistema robotizado representa un avance significativo en la automatización industrial en un sector muy tradicional y manufacturero, el sector del juguete. Esta tecnología ofrecerá una solución eficiente, precisa y flexible al proceso de producción de juguetes, concretamente en el proceso de ensamblaje de la muñeca, mediante la optimización de recursos, lo que deriva en una reducción directa de costes y en una mayor adaptabilidad a los cambios en la demanda del mercado. Por último, el operario no será reemplazado en el proceso completo, únicamente en las tareas que requieran grandes esfuerzos o sean tediosas. La complejidad añadida a este proceso de ensamblaje es la manipulabilidad del producto, ya que tratamos con un plástico flexible y deformable., Nowadays, small and medium-sized companies find themselves in a complicated situation when it comes to making the leap towards the robotisation of their industrial processes, either for economic reasons or because they are unfamiliar with these technologies. That is why this robotic system represents a significant advance in industrial automation in a very traditional and manufacturing sector, the toy sector. This technology will offer an efficient, precise and flexible solution to the toy production process, specifically in the toy doll assembly process, by optimising resources, resulting in a direct reduction in costs and greater adaptability to changes in market demand. Finally, the operator will not be replaced in the entire process, only in the tasks that require great effort or are tedious. The added complexity to this assembly process is the manipulability of the product, because it is a flexible and deformable plastic.

In recent years, industrial automation has experienced a paradigm shift towards closer interaction and collaboration between humans and robots, driven by technological advances and the growing market for collaborative robots. As robots now operate in close proximity to humans, often without the constraints of spatial safety barriers, the demands on motion planning have increased substantially. Traditional motion planning approaches are reaching their limits in these uncertain and dynamically changing environments, emphasizing the need for novel reactive motion planning algorithms with a major focus on safety. Consequently, the primary objective of this thesis is to develop a comprehensive robotic motion planning framework that bridges the gap between global trajectory planning and reactive control. This framework aims to provide rigorous mathematical guarantees for collision avoidance and goal convergence, validated through extensive simulations and real-world experiments. To achieve this goal, we extend and enhance the circular fields motion planning approach to serve as a robust foundation for the locally reactive planning component. Inspired by the principles of electromagnetic fields, circular fields are inherently free of local minima and enable immediate reactions to dynamic obstacles and changing environmental conditions. We introduce several global planning components tightly integrated with the local control unit, which are augmented with virtual predictive agents for efficient global exploration, and leverage global information about promising avoidance directions. Our framework heavily exploits the inherent parallelizability of these combined approaches, ensuring that the reactive command calculation of the circular field-based local controller remains unaffected by potential delays in global planning processes. Furthermore, our rigorous mathematical analysis verifies the collision avoidance properties of the framework, providing formal guarantees for col

El uso de sensores táctiles está comenzando a ser una práctica común en tareas complejas de manipulación robótica. Este tipo de sensores proporcionan información extra sobre las propiedades físicas de objetos que están siendo agarrados y/o manipulados. En este trabajo, se ha implementado un sistema capaz de medir el deslizamiento rotacional que pueden sufrir objetos durante su manipulación. Nuestra propuesta emplea sensores táctiles ópticos DIGIT, a partir de los cuáles se capturan imágenes de contacto que luego se procesan e interpretan. En concreto, nuestro método hace uso de un modelo neuronal para la detección de la región de contacto. Y posteriormente, mediante extracción de características visuales de la región detectada, se estima el ángulo causado por movimientos de deslizamiento. Nuestro método estima correctamente la región de contacto obteniendo un 95% y 91% usando las métricas Dice e IoU. Y es capaz de obtener un error medio máximo de 3º en agarres de objetos nunca vistos anteriormente., Using tactile sensors is becoming a common practice to achieve complex manipulation in robotic tasks. These kinds of sensors provide extra information about the physical properties of the grasping and/or manipulation task. In this work, we have implemented a system that is able to measure the rotational slippage of objects in hand. Our proposal uses the vision-based tactile sensors known as DIGITs which allow us to capture contact images, which are then processed. In particular, our method is based on a neural network model applied to the detection of touch/contact regions. Afterwards, we extract visual features from detected contact regions and we then estimate the angle generated due to an unwanted slippage. Our method obtains results of 95% and 91% in Dice and IoU metrics for contact estimation. In addition, it is able to obtain a mean rotational error of 3 degrees in the worst case with previously unseen objects.

[Resumen] Junto con el desarrollo de la industria 4.0 y la inclusión de procesos distribuidos y flexibles en diferentes arquitecturas como puede ser la de fabricación en enjambre, se han realizado avances en las funciones de los robots móviles que pueden aparecer en dichos escenarios tales como los AGV o los AMR, para poder trabajar en entornos con humanos de forma eficiente. Además de reducir el consumo de las actividades de flotas de estos dispositivos, proveyendo a la industria de soluciones para optimizar los procesos de fabricación (principalmente en aspecto de suministro de recursos mediante tareas de logística). En este artículo se realiza un estudio del estado de la materia en los distintos campos que engloban las funciones de las flotas de robots móviles: Asignación de Tareas, Planificación de Trayectorias y Planificación del Movimiento, centrándose en los últimos avances en la creación óptima de trayectorias y en el desempeño en entornos con humanos., [Abstract] Together with the development of Industry 4.0 and the introduction of flexible and distributed manufacturing processes into different architectures such as Swarm Manufacturing, there have been several research progresses about the main functionalities of mobile robot fleets. These groups of end devices are present in today’s main industry-based scenarios, in terms of AGVs and AMRs to work within human centered activities, such as intra-logistics. In this article, it is presented an overview about the startof-the-art in the main functionalities of industry mobile robot fleets: Task Allocation, Path Planning and Motion Planning. However, the principal topic where this paper is focused on is Path Planning. The most promising advances in this area are analyzed and discussed targeting the optimization of trajectories and the capacity to move alongside with humans.

[Resumen] El uso de sistemas de percepción que combinan sensores LiDAR y cámaras se está haciendo cada vez más frecuente, pues la fusión de sus datos proporciona una ayuda para la percepción del entorno. Además, las tareas de detección de objetos o la segmentación semántica del entorno precisan, frecuentemente, la combinación de sensores de diferente naturaleza con el objetivo de mejorar sus resultados. La calibración de estos sistemas es un paso fundamental para la correcta fusión de los datos. En este artículo se propone un método de calibración basado en un patrón para este tipo de sistemas, formados por una cámara fisheye y un LiDAR. El método evita transformar las imágenes a un modelo pin-hole, realizando una calibración a partir de la proyección esférica de la imagen, obtenida mediante los parámetros de calibración y la nube de puntos correspondiente., [Abstract] The usage of LiDAR and camera systems is increasing rapidly, since the fusion of their information may provide a better undestanding of the environment. In particular, the correct fusion of the data provided by both sensor types allows to improve the results in terms of object detection or semantic segmentation. The calibration of these systems is crucial for the correct fusion of data. This article proposes a target-based calibration method for such systems, formed by a LiDAR and a fisheye camera. The method avoids transforming the images to a pin-hole camera model, performing a calibration based on the spherical projection of the image, obtained with the calibration parameters and the corresponding point cloud.

[Resumen] La robótica ha mejorado nuestra calidad de vida al interactuar con el entorno de formas innovadoras, siendo la microrrobótica capaz de realizar tareas específicas y delicadas en áreas inaccesibles para dispositivos convencionales, como la medicina. El objetivo de este trabajo es diseñar, fabricar y controlar un robot movido por cabeceo con la ayuda de un flagelo pasivo utilizando simulaciones computacionales, modelos matemáticos y controlador previos. Este robot servirá como prototipo para estudios iniciales sobre microrrobótica. El robot se fabricó utilizando impresión 3D y se utilizó electrónica y motores específicos para generar la propulsión. Se realizaron experimentos de natación con y sin flagelo. El movimiento del prototipo sin flagelo durante la natación tiende a descontrolarse rápidamente, mientras que el que posee el flagelo, se mantiene recto. Estos resultados concuerdan con las simulaciones realizadas en trabajos previos., [Abstract] Robotics has improved our quality of life by interacting with the environment in innovative ways, with microrobotics being able to perform specific and delicate tasks in areas inaccessible to conventional devices, such as medicine. The aim of this work is to design, manufacture and control a robot moved by pitching with the help of a passive flagellum using computational simulations, mathematical models and a previous controller. This robot will serve as a prototype for initial studies on microrobotics. The robot was fabricated using 3D printing and specific electronics and motors were used to generate propulsion. Swimming experiments were conducted with and without a flagellum. The movement of the prototype without the flagellum during swimming tends to rapidly get out of control, while the one with the flagellum stays straight. These results are in agreement with simulations carried out in previous work.

[Resumen] Hoy en día se están investigando robots de exploración terrestre y espacial más rápidos en respuesta a la creciente demanda de capacidades de exploración e investigación más rápidas, eficaces y rentables. Para estas plataformas móviles rápidas la identificación y evasión de obstáculos lejanos son críticas, ya que su alta velocidad implica la necesidad de tener en cuenta el mayor número posible de obstáculos cercanos y lejanos para el cálculo de la trayectoria global, evitando cualquier posible accidente debido a su velocidad y al tiempo de cálculo de los algoritmos de replanificación. Debido a su distancia, los obstáculos lejanos no se incluyen en los mapas locales, que están limitados por el alcance de las cámaras de profundidad. Por estas razones, este artículo propone el uso de técnicas de Inteligencia Artificial para detectarlos a partir de imágenes y estimar sus tamaños y posiciones con un cierto grado de incertidumbre. Los obstáculos detectados se incluirán posteriormente en los mapas globales, corrigiendo la trayectoria global en caso de colisionar con ellos., [Abstract] Nowadays faster terrestrial and space exploration robots are being investigated, in response to the growing demand for faster, more efficient, and cost-effective exploration and research apabilities. For these rapid mobile platforms, the identification and avoidance of far obstacles are critical, since their high speed implies the need to take into account as many near and far obstacles as possible for the global path computation, avoiding any possible accident due to their speed and the computation time of the replanning algorithms. Due to their distance, the far obstacles are not included within the local maps, which are limited by the range of the depth cameras. For these reasons, this paper proposes the use of Artificial Intelligence techniques to detect them from images and estimate their sizes and positions with a certain degree of uncertainty. The detected obstacles will be later included in the global maps, correcting the global path in case it collides with them.

[EN] Computer vision is gaining more and more importance in the world of industrial robotics, since it is necessary to carry out increasingly precise and autonomous tasks, which is why a more exact positioning of the robot is needed. This requires the support of a vision system that is the one that gives the robot precision in its pose, calibrating said system with respect to the robot. This work presents a simple methodology to approach this form of calibration, called hand-eye, using a structured light 3D camera that obtains information from the real world and a six-axis industrial robotic arm. The method uses the RANSAC algorithm for the determination of the planes, which represents a notable reduction in errors, since the coordinates of the points sought come from planes adjusted to thousands of points. This allows the system to always have the ability to obtain a transformation matrix from the coordinates of the camera to the base of the robot. In addition, the proposed method is ideal for making a precision comparison between cameras, due to its simplicity and speed of use. In this study, the resulting error analysis was performed using two dfferent 3D cameras: a basic one (Kinect 360) and an industrial one (Zivid ONE + M)., [ES] La visión artificial está cobrando cada día más auge en el mundo de la robótica industrial, ya que es necesario realizar tareas cada vez más precisas y autónomas, por lo que se necesita un posicionamiento del robot más exacto. Para ello se precisa del apoyo de un sistema de visión que sea el que preste al robot precisión en su pose, calibrando dicho sistema con respecto al robot. Este trabajo presenta una metodología sencilla para abordar esta forma de calibración, llamada ojo a mano, empleando una cámara 3D de luz estructurada que obtiene la información del mundo real y un brazo robótico industrial de seis ejes. Esto permite utilizar el algoritmo RANSAC para la determinación de los planos, cuya intersección nos da las coordenadas de los puntos,lo que supone una reducción notable de los errores, ya que las coordenadas proceden de planos ajustados a miles de puntos, lo cual hace que el sistema sea más robusto y capaz de obtener una matriz de transformación de las coordenadas de la cámara a la base del robot, que le permitirá abordar cualquier tarea que precise con una precisión eficiente. Se ha realizado el análisis de errores resultante utilizando dos cámaras 3D diferentes: una básica (Kinect 360) y otra industrial (Zivid ONE+ M).

La visión artificial está cobrando cada día más auge en el mundo de la robótica industrial, ya que es necesario realizar tareas cada vez más precisas y autónomas, por lo que se necesita un posicionamiento del robot más exacto. Para ello se precisa del apoyo de un sistema de visión que sea el que preste al robot precisión en su pose, calibrando dicho sistema con respecto al robot. Este trabajo presenta una metodología sencilla para abordar esta forma de calibración, llamada ojo a mano, empleando una cámara 3D de luz estructurada que obtiene la información del mundo real y un brazo robótico industrial de seis ejes. Esto permite utilizar el algoritmo RANSAC para la determinación de los planos, cuya intersección nos da las coordenadas de los puntos, lo que supone una reducción notable de los errores, ya que las coordenadas proceden de planos ajustados a miles de puntos, lo cual hace que el sistema sea más robusto y capaz de obtener una matriz de transformación de las coordenadas de la cámara a la base del robot, que le permitirá abordar cualquier tarea que precise con una precisión eficiente. Se ha realizado el análisis de errores resultante utilizando dos cámaras 3D diferentes: una básica (Kinect 360) y otra industrial (Zivid ONE+ M)., Computer vision is gaining more and more importance in the world of industrial robotics, since it is necessary to carry out increasingly precise and autonomous tasks, which is why a more exact positioning of the robot is needed. This requires the support of a vision system that is the one that gives the robot precision in its pose, calibrating said system with respect to the robot. This work presents a simple methodology to approach this form of calibration, called hand-eye, using a structured light 3D camera that obtains information from the real world and a six-axis industrial robotic arm. The method uses the RANSAC algorithm for the determination of the planes, which represents a notable reduction in errors, since the coordinates of the points sought come from planes adjusted to thousands of points. This allows the system to always have the ability to obtain a transformation matrix from the coordinates of the camera to the base of the robot. In addition, the proposed method is ideal for making a precision comparison between cameras, due to its simplicity and speed of use. In this study, the resulting error analysis was performed using two different 3D cameras: a basic one (Kinect 360) and an industrial one (Zivid ONE + M).

Diese Dissertation behandelt die Aufklärung und Überwachung von Umweltfeldern mit kleinen autonomen Unterwasserroboter (µAUVs). Trotz jüngster Fortschritte sind recheneffiziente Methoden für Planung, Lokalisierung und Regelung von µAUVs weitgehend unerforscht. Die im Rahmen dieser Arbeit entwickelte µAUV-Plattform HippoCampus und seine Regelarchitektur erlauben agiles Manövrieren auf engstem Raum. Zu diesem Zweck wird ein Selbstlokalisierungssystem entwickelt, das visuelle, elektromagnetische und akustische Signale zur räumlichen Lokalisierung verwendet. Basierend auf Methoden der informationstheoretischen Regelung wird abschließend wird eine recheneffiziente Architektur zur autonomen Feldaufklärung mit mehreren Robotern vorgeschlagen. Diese kombiniert Deep-Reinforcement Learning mit einer stochastischen Feldmodellierung., Exploration and monitoring of hazardous environmental fields are among the most promising tasks to be performed by micro autonomous underwater vehicles (µAUVs). Despite recent progress, computationally efficient solutions for guidance, navigation, and control are largely unsolved for agile µAUVs. First, the hydrobatic micro robot platform HippoCampus is presented along with a control system that allows agile maneuvering in confined spaces. Furthermore, an embedded self-localization system is developed which consists of modules using visual, electromagnetic, and acoustic ranging. Finally, an informative path planning framework for autonomous field exploration with multiple robot agents is proposed. It combines a deep reinforcement learning planner with a stochastic representation of the environmental field.

When mobile robots are employed in transportation tasks involving contact with humans, their control software shall guarantee that in every possible circumstance safety and, in general, task requirements are guaranteed. When control models are manually translated into an executable implementation, it becomes cumbersome to provide such guarantees. Model-driven engineering approaches provide an answer to such a problem. Domain specific models are automatically translated into an executable implementation. Some model-driven engineering approaches exist that are specific to robotics. However, formal guarantees on correctness of the model and the generated implementation with respect to the requirements are, often, not provided. This paper investigates how a general purpose modelling language for supervisory controller synthesis can be used to formally model plants and requirements for a robotic navigation task and can generate an executable implementation that can be integrated into a leading middleware for robotic applications. The starting point is the modelling of the interface provided by existing navigation components available in the targeted middleware. We demonstrate, with simulations and real-life experiments, that the generated supervisory controller is suitable for real-time deployment and guarantees correctness of the model with respect to the requirements of the navigation task at hand. Results on the reaction time of the supervisory controller show that such reaction time is about twenty times smaller than the one of the same supervisory controller implemented with a conventional framework.

In dieser Arbeit werden neue Mission Control Strategien für Multiagentensysteme vorgeschlagen, die auf Szenarien wir Source Seeking und Level Curve Tracking angewandt werden können. Zunächst wird ein gradientenbasiertes Verfahren für Source Seeking und anschließendes Level Curve Tracking vorgestellt, in dem Agenten diskreten Konsens über den Übergang von der ersten zur zweiten Task erreichen. Weiterhin wird eine gradientenfreie, flocking-basierte Methode für Source Seeking entwickelt, die auf Glowworm Swarm Optimization beruht. Das Multiple-Source-Seeking Problem unter zeitveränderlicher Kommunikationstopologie wird gelöst durch eine Kombination von bio-inspirierter Attraktion mit geschätztem Gradienten und Formationsregelung, wobei der Wechsel zur nächsten Task durch Regeln der linearen temporalen Logik bestimmt wird. Schließlich wird ein Ansatz vorgestellt, der Flocking und Formationsregelung kombiniert, wobei auf einer unteren Ebene die Agenten Formationen bilden, während auf der oberen Ebene die Leader durch Flocking ein unbekanntes Extremum lokalisieren. Neben einer Konvergenzanalyse für alle Verfahren wird deren Leistungsfähigkeit durch Simulationen und experimentelle Ergebnisse illustriert., In this work novel mission control strategies are proposed for multiagent systems, that can be applied to scenarios such as source seeking and level curve tracking. A gradient-based method for source seeking and subsequent level curve tracking is presented, where the agents reach discrete consensus on whether they should switch from Task 1 to Task 2. Moreover, a gradient-free flocking-based approach to the source seeking problem is developed that is based on glowworm swarm optimization. The multiple-source-seeking problem for time-varying communication topologies is solved using a combination of bio-inspired attraction with estimated gradient and formation control, where task-switching is performed under a linear temporal logic rule. Finally an approach that combines flocking and formation control techniques is presented, where low-level agents attain a specified formation, while the formation leaders at the upper level flock to locate an unknown extremum. Convergence analysis for all methods is provided, together with simulation results and experimental validation.

In this paper, we present an experimental test bench to implement various cooperative control strategies for multi-agent systems, and illustrate its use with experimental results for a source-seeking problem, where a group of small wheeled robots termed as Zooids should locate a source of a given spatial scalar field. This algorithm is implemented as a validation to demonstrate the capabilities of the test bench. We propose to achieve this by utilising an internal target-based position controller, under the assumptions of convexity of the scalar, continuous/discrete field and availability of local measurements of the field, so that agents can calculate its gradient and its Hessian. We then show in experiments, that using estimated gradients and Hessians (with data communicated from neighbours) in the presence of noisy measurements of the field strength provides satisfactory results for convex fields, under various algorithms such as Steepest Descent, Gauss-Newton, Levenberg Marquardt. These algorithms are analysed, and experimental results are discussed.

Control Lyapunov Function - Dynamic Movement is a recently proposed imitation learning approach for controlling point-to-point movements. In this work some practical issues arising with this approach have been investigated by studying a particular application: robotic handwriting of German letters. One difficulty in this context is the handling of crossings. Here a novel segmentation scheme is proposed to eliminate crossings. Using a segmentation framework for splitting up the movement into a number of partial movements has additional advantages: It reduces the computational time for calculating the dynamic system that is required in this approach, as well as the time for calculating a Lyapunov function because of the simpler parametrization. This proposed segmentation scheme has been validated experimentally on a CRS A465 robotic manipulator.

This paper presents a novel approach to the source seeking problem, where a group of mobile agents tries to locate the maximum of a scalar field defined on the space in which they are moving. The agents know their position and the local value of the field, and by communicating with their neighbors estimate the gradient direction of the field. A distributed cooperative control scheme is then designed that drives the group towards the maximum while maintaining a specified formation. Previously proposed control schemes that are based on a combination of H∞-optimal formation control and local gradient estimation suffer from premature convergence to local maxima. To overcome this problem, here the use of particle swarm optimization for locating the global maximum is proposed. Agents take the role of particles and an information flow filter approach is employed to separate the consensus dynamics from the local feedback loops governing the agent dynamics. Stability of the overall scheme is established based on the small gain theorem, and by decomposing the synthesis problem for the distributed information flow filter the problem size is reduced to that of a single agent. Simulation results with multiple maxima and quadrocopter models as agents illustrate the practicality of the approach.

This study is concerned with the shared object manipulation problem in a physical Human-Robot Interaction (pHRI) setting. In such setups, the operator manipulates the object with the help of a robot. In this paper, the operator is assigned with the lead role, and the robot is passively following the forces/torques exerted by the operator. We propose a controller that is free from the well-known translation/rotation problem and enhances the operator's ability to move the object by reducing the human effort. The key point in our study is that the controller is defined based on the instantaneous center of rotation. The passivity of the system including the object and the manipulator has been evaluated. Simulation results validate the theoretical findings on different scenarios of subsequent rotations and translations of the object., QC 20180314

Reinforcement learning techniques enable robots to deal with their own dynamics and with unknown environments without using explicit models or preprogrammed behaviors. However, reinforcement learning relies on intrinsically risky exploration, which is often damaging for physical systems. In the case of the bipedal walking robot Leo, which is studied in this paper, two sources of damage can be identified: fatigue of gearboxes due to backlash re-engagements, and the overall system damage due to falls of the robot. We investigate several exploration techniques and compare them in terms of gearbox fatigue, cumulative number of falls and undiscounted return. The results show that exploration with the Ornstein-Uhlenbeck (OU) process noise leads to the highest return, but at the same time it causes the largest number of falls. The Previous Action-Dependent Action (PADA) method results in drastically reduced fatigue, but also a large number of falls. The results reveal a previously unknown trade-off between the two sources of damage. Inspired by the OU and PADA methods, we propose four new action-selection methods in a systematic way. One of the proposed methods with a time-correlated noise outperforms the well-known e-greedy method in all three benchmarks. We provide guidance towards the choice of exploration strategy for reinforcement learning applications on real physical systems., Biomechatronics & Human-Machine Control, Learning & Autonomous Control