APRÈN. Portal de la producció docent del professorat de la UPC

In an increasingly globalised and competitive world, companies in the production sector are forced to make quantitative leaps in terms of production in order to be able to meet the growing demand. At the same time, coinciding with a critical climatic context in which the focus is on responsible production focused on minimising the impact on the environment. This thesis studies automation as the main driving force of industry towards the optimisation of production processes, reviewing its historical trajectory and at the same time looking to the future. At the same time, it also focuses on inverse logistics, which is the main protagonist in the advance of the productive fabric towards sustainable production in order to be able to maintain production growth while at the same time being environmentally responsible. All of this is highlighted in this work through the simulation of various models of production processes that incorporate these techniques, highlighting the benefits that they bring for both companies and the environment.

Dual-arm robotic manipulation using visual guidance poses many challenges in order to actu- ally perform manipulation tasks with enough efficiency. The motion coordination and synch- ronization of arms or working in real-world environments that presents uncertainties, are some of the challenges that this thesis will focuses to solve. To cope with these challenges, this final master thesis describes the implementation of a sys- tem for motion coordination and synchronization of a mobile anthropomorphic dual-arm robot (MADAR) that focuses on the capability for performing manipulation actions with precision by means of a visual guidance provided by a camera mounted on an articulated head. The proposal uses the second version of Robot Operating System (ROS 2), where two different control modes are implemented, using C++, to develop a solution. Implementation has been developed for MADAR and uses the UR5 model from Universal Robots as arms, but it has been developed generically to be used with any robot that uses ros_control standards. Evaluated with real experiments, that consist on executing a given trajectory to the goal detected by the perception system and perform dynamic movements based on the task. The implemen- ted proposal brings the necessary capabilities to the robot to achieve the objectives. Finally, a discussion of the main future research lines is explained to improve the implementation.

En aquest treball de fi de grau (TFG) es tracta de establir les bases teòriques i pràctiques d'un cas particular d'implementació d'un sistema de Machine Learning en una màquina formadora de caixes de l'empresa Tavil Ind. SAU. L'objectiu del treball per tant és afegir l'estructura en el funcionament normal de la màquina per tal de realitzar-ne el manteniment predictiu i poder anticipar-se a una possible fallada imminent o no prevista. Es pretén que es capturi un registre de dades d'un cicle i el classifiqui en un règim de funcionament, depenent de l'estat dels mecanismes associats a l'element d'estudi. Cal destacar que el treball es desenvolupa en Python, essent un llenguatge de programació bàsicament destinat a la realització de tasques d'aprenentatge automàtic, gràcies al gran abast de biblioteques disponibles. Com a IDE (entorn de desenvolupament de software), s'utilitza el Jupyter Notebook degut a la seva facilitat de visualització de totes les operacions realitzades. Tot i que es presenta una solució concreta i rígida, es demostra que el model d'aprenentatge automàtic proposat és adient per treballar amb un conjunt de dades com el generat pels sensors de la màquina formadora, cosa que pot ser útil per a futures solucions més flexibles i generals. Per últim, en el present document s'observa una aplicació d'un fenomen molt recurrent actualment com és la Intel·ligència Artificial (IA), on cada cop més els automatismes van adquirint la capacitat de realitzar tasques autònomes i aprendre d'aquestes.

The aim of this project is the modelling and simulation of the STÄUBLI TX90 industrial robot from Stäubli, located in the robotic cell located in the robotics laboratory of the IOC (Institut Obert de Catalunya) of Escola Tècnica Superior d'Enginyeria Industrial de Barcelona (ETSEIB). This industrial robot corresponds to a robotic arm with IP65 protection (suitable for hostile environments) with a total mass of 111 kg, which has a total of six degrees of freedom, these being rotating joints. Characterised by high precision and optimum use of its working area. This thesis will conduct a comprehensive analysis of the necessary tools for the execution of various programs. The assimilation of key concepts will be crucial in establishing a guiding framework for the project's development. The combination of these concepts will facilitate the development of Python scripts designed to execute standard robotic movements and other functionalities exposed below: • Subscription to the position nodes of the joints of our industrial robot. • Publication of new positions of the joints as well as their simulation • Reproduction of the MoveL movement • Reproduction of the MoveJ movement All this will be implemented in order to make use of the Python scripts through the terminal of the operating system thanks to the implementation of the ROS 2 middleware and its available libraries.

This thesis presents the current automotive macrotrends and the growth of CUPRA in the actual scenario. Its main objective is to identify the key aspects that make the commercialisation of CUPRA an unprecedented success story in the european market. The study validates the initial hypothesis that suggests this brand will play a very relevant role in the automotive sector in 2035, being this date the one set by the European Union to achieve zero emissions in new registrations. Electrification, digitalization, artificial inteligence, multimodal and on-demand mobility and new distribution models are presented as key points. Also, the growth of CUPRA and SEAT S.A.'s strategy to position itself as an expertised manufacturer in the development of electric cars in the near future

La teleoperación remota en robótica ofrece numerosas ventajas, como la ejecución a distancia de tareas sin desplazamiento físico, la multitarea simultánea en determinados escenarios y la supervisión centralizada. Concretamente, en el contexto de la limpieza autónoma de playas, la teleoperación permite la solución remota de ciertos problemas, la generación de nuevas rutas de limpieza y la ejecución de pruebas e inspecciones. Este proyecto de final de Máster se centra en el desarrollo de un sistema de teleoperación para un robot de limpieza de playas. La tesis explora el estado actual de la robótica en la limpieza ambiental, así como la teleoperación remota en robótica. Proporciona una descripción detallada del sistema y una visión general de las diversas herramientas empleadas. El diseño de la solución se ha realizado desde cero, involucrando el desarrollo tanto de los programas cliente que se ejecutan en el ordenador de teleoperación como de la arquitectura completa del robot. La arquitectura del robot se basa en un diseño modular utilizando el frame- work ROS2, con el objetivo de lograr simplicidad, facilidad de depuración y escalabilidad para futuros avances hacia la navegación autónoma. Para la teleoperación, se hace uso de un joystick como interfaz de control principal, con funcionalidades automatizadas en el lado del robot para garantizar el funcionamiento inmediato de las medidas de seguridad y facilitar la gestión de la iluminación, reduciendo así la carga de trabajo del operador. Además, en el proyecto se ha incorporado la creación de una simulación que ofrece un entorno seguro y flexible para pruebas y desarrollo. Este enfoque permite realizar pruebas más rápidas al minimizar la necesidad de utilizar un robot físico. Finalmente, el proyecto presenta pruebas experimentales y resultados, comparando el rendimiento del sistema tanto en simulación como en entornos reales. Las pruebas abarcan di- versos aspectos, como la funcionalidad de comandos de velocidad, las medidas de seguridad, los sensores, la conexión remota general entre el teleoperador y el robot, y los efectos de latencia. Los resultados de la experimentación validan la efectividad del sistema de teleoperación

Con la acelerada transformación digital llevada a cabo en los últimos años, para aprovechar las nuevas tecnologías han surgido automatizaciones de software para optimizar los procesos de una empresa. En este proyecto se estudia de qué trata las automatizaciones, las diferentes tecnologías que existen y su aplicación en las empresas, así como todas sus ventajas y desventajas. Pero se centrará en el análisis más detallado de la automatización robótica de procesos que es la principal y más usada automatización por las empresas dentro del mundo de la automatización de procesos digitales. Se detallará cómo se lleva a cabo una automatización RPA en todas sus fases, las herramientas que existen y las posibles integraciones con otras tecnologías de automatización, así como también se desarrollará un caso práctico simulando una automatización. Para su correcto entendimiento se facilita un video para comprobar su correcto funcionamiento y entender sus numerosos beneficios. Para acabar, se estudiará la continua evolución de las automatizaciones de software y sus próximos pasos dentro de la Inteligencia Artificial. Se verá las actuales tecnologías que ofrecen automatizaciones, su funcionamiento y sus posibles aplicaciones en los procesos empresariales presentes y futuros

Aquest document recull la memòria del projecte fet a Gestamp Abrera. L'objectiu del projecte ha estat millorar el subministrament de contenidors de dos processos a partir de vehicles de guiat automàtic. A partir de la descripció dels processos s'ha pogut fer una idea de la situació prèvia a la millora. Més endavant s'ha definit els robots i s'ha fet la caracterització dels sistemes que l'envolten. Un cop fet l'anàlisi d'alternatives s'ha escollit l'opció que s'ajusta més als criteris de l'empresa. La modelització del robot s'ha dut a terme a partir d'una empresa proveïdora, en què s'ha supervisat en tot moment per l'equip de projectes. Malauradament degut a imprevistos només s'ha pogut fer una descripció dels passos que s'han dut a terme durant aquesta part del projecte. Per acabar s'ha fet un estudi econòmic per veure la seva rendibilitat. S'ha pogut comprovar que aquesta implementació suposarà un estalvi de 80.000 euros durant els cinc anys. També s'ha estudiat l'impacte ambiental i social del projecte, en què respectivament s'ha vist un descens en les emissions de CO2 i els diferents canvis que afectaran els treballadors de la planta, precisament als carreters

Robots performing manipulation tasks require the accurate location and orientation of an object in space. Previously, at the Robotics Laboratory of IOC-UPC this data has been generated artificially. In order to automate the process, a perception module has been developed for providing task and motion planners with the localization and pose estimation of objects used in robot manipulation tasks. The Robot Operating System provided a great framework for incorporating vision provided by Microsoft Kinect V2 sensors and the presentation of obtained data to be used in the generation of Planning Domain Definition Language files, which define a robots environment. Localization and pose estimation was done using fiducial markers along with studying possible enhancements using deep learning methods. Perfectly calibrating hardware and setting up a system play a big role in enhancing perception accuracy and while fiducial markers provide a simple and robust solution in laboratory conditions, real world applications with varying lighting, viewing angles and partial occlusions should rely on AI vision

The Final Degree Project consists of the design of an online platform, or web, that allows managing more efficiently and simply the cycle of the ETSEIB Forum for the different users: organizers, companies and students. This platform should reduce the management tasks by the organizers, optimize the monitoring and control of the status of participation and localize the existing databases in one place allowing a better updating and checking. On the other hand, it will permit companies to reduce the paperwork and make them able to manage all the documentation from a single space without the need to exchange a large number of files via e-mail. Finally, it will facilitate the process of enrolling students in the activities and their confirmation. To be able to design this platform, the historical context has been previously defined and the methodology currently used has been analyzed, which has made it possible to detect the weak points on which action should be taken. Based on these weaknesses and the needs expressed by the organizers, the functional and non-functional requirements have been defined from which a design of the platform has been made after analyzing the different existing alternatives. Based on this design, a person or company specialized in the development of web pages may be hired for its creation and subsequent implementation if the ETSEIB direction's committee decides so

The objective of this project is to develop an integrated framework for the automatic execution of complex Task and Motion planning problems in a multi-robot environment. In dynamic environments, such as the introduction of a human in the loop, the system cannot rely solely on an original plan. Possible disruptions occurring in the environment mid-execution, like an object being placed where it should not be by the human, or an object falling on its side, require a way to react to them and obtain a new plan in order to achieve the goal. The chosen tool to center the framework around is Behavior Trees (BTs). BTs have been cho- sen due to their ease of use, their modularity, and their portability. Each of the steps that are executed in the plan is represented as an Action Node. Through the use of Control Nodes, the order in which the different branches of the tree are executed can be dictated. The use of Reac- tive Control nodes allows for the system to react to the changes in the environment as soon as they occur. The framework begins by executing an original plan which has been planned offline based on the initial state of the environment. Once the BT executor is run, the system will check that the environment has not deviated from the original. If it has, a replan is triggered, and the BT will be rewritten mid-execution with the updated Task and Motion plan. Good monitoring of the system state is required in order to react to disruptions at either the geometric or the symbolic level. This led to the development of the State Module. Two different States are defined: the Expected State collects all the information from the original plan and represents the state the system should be in, while the Observed State uses the sensors in the perception module to sense the state the system is in. A mismatch between the two signifies that the original plan is no longer valid, and a replan is required. The robot chosen to perform most of the testing is a dual-armed 7 Degree of Freedom ABB YuMi. The YuMi robot does not offer a ROS interface, which would make it impossible to integrate the robot in the rest of the framework, which is based on ROS. For this reason, a ROS interface for the YuMi has been developed, which serves as a module for other robots and their ROS interfaces if they have to be integrated in the framework. Part of the framework involved converting the output of the motion planners, which are a se- ries of waypoints in joint space, to a trajectory with the temporal component for the robot to perform. A new methodology named Waypint, short for Waypoint Interpolation, has been developed, with the goal of generating efficient and robust trajectories considering the mechan- ical constraints of robotic manipulators. The experiments done to evaluate the framework proved its efficacy when purposely gener- ating changes in the environment that required the system to perform a replan. Finally, the shortcomings of the implementation and potential avenues for improvement which remain as future work are discussed.

El trabajo que a continuación se presenta, trata sobre el estudio sobre la capacidad que ofrece la cámara OAK-D, cámara característica por su integración de inteligencia artificial. Este modelo de cámara de la empresa Luxonis Holding Corporation, dispone de una plataforma espacial de inteligencia artificial y visión por computador, el cual se conoce como DepthAI. Gracias al uso de esta plataforma, se desarrolla una interfaz de Python para su integración en una plataforma móvil, la cual deberán actuar en conjunto mediante el uso de la interfaz Robot Operating System (ROS). Destacar que en este mismo proyecto, debido a la necesidad de colaboración entre los elementos cámara – plataforma móvil, se implementa este middleware en el código de Python desarrollado.

The objective of this project is to develop an integrated framework for the automatic execution of complex Task and Motion planning problems in a multi-robot environment. In dynamic environments, such as the introduction of a human in the loop, the system cannot rely solely on an original plan. Possible disruptions occurring in the environment mid-execution, like an object being placed where it should not be by the human, or an object falling on its side, require a way to react to them and obtain a new plan in order to achieve the goal. The chosen tool to center the framework around is Behavior Trees (BTs). BTs have been chosen due to their ease of use, their modularity, and their portability. Each of the steps that are executed in the plan is represented as an Action Node. Through the use of Control Nodes, the order in which the different branches of the tree are executed can be dictated. The use of Reactive Control nodes allows for the system to react to the changes in the environment as soon as they occur. The framework begins by executing an original plan which has been planned offline based on the initial state of the environment. Once the BT executor is run, the system will check that the environment has not deviated from the original. If it has, a replan is triggered, and the BT will be rewritten mid-execution with the updated Task and Motion plan. Good monitoring of the system state is required in order to react to disruptions at either the geometric or the symbolic level. This led to the development of the State Module. Two different States are defined: the Expected State collects all the information from the original plan and represents the state the system should be in, while the Observed State uses the sensors in the perception module to sense the state the system is in. A mismatch between the two signifies that the original plan is no longer valid, and a replan is required. The robot chosen to perform most of the testing is a dual-armed 7 Degree of Freedom ABB YuMi. The YuMi robot does not offer a ROS interface, which would make it impossible to integrate the robot in the rest of the framework, which is based on ROS. For this reason, a ROS interface for the YuMi has been developed, which serves as a module for other robots and their ROS interfaces if they have to be integrated in the framework. Part of the framework involved converting the output of the motion planners, which are a series of waypoints in joint space, to a trajectory with the temporal component for the robot to perform. A new methodology named Waypint, short for Waypoint Interpolation, has been developed, with the goal of generating efficient and robust trajectories considering the mechanical constraints of robotic manipulators. The experiments done to evaluate the framework proved its efficacy when purposely generating changes in the environment that required the system to perform a replan. Finally, the shortcomings of the implementation and potential avenues for improvement which remain as future work are discussed.

El trabajo que se presenta a continuación expone el desarrollo de un módulo de detección de objetos mediante el uso de tecnología RFID (Identificación por Radiofrecuencia). Este desarrollo está orientado a robots manipuladores de tal manera que éstos puedan conocer la localización de un objeto, el cual contiene una etiqueta RFID, que se conoce como Tag (Circuito con información y antena emisor-receptor), en un sector o área aproximada en un ambiente interior. Además, este desarrollo involucra la lectura y escritura en la memoria interna de la etiqueta RFID. Para lograr este objetivo, y contando con todo el equipamiento e infraestructura proporcionado por el laboratorio de robótica, se han realizado pruebas de localización colocando diversos objetos, con sus respectivos tags RFID, en 4 zonas diferentes del laboratorio onsiderando que cada zona cuenta con una antena de detección. De esta forma, y a través de algoritmos, se podrá conocer qué antena es la que detecta el objeto requerido, y por tanto se podrá conocer la localización de este

This thesis is integrated in a research project about collaborative autonomous mobile robots and focusses on energy-aware strategies for exploring unknown environments. To this aim, we have reviewed the data in the literature by focusing on the energy efficiency, as well as the major energy consumer elements in autonomous mobile robots. Then, an analysis of coordination strategies in collaborative mobile robots is realised. The evaluation of different communication and decision-making methods is carried out. This review allows us to select the optimal model for this thesis purpose. Accordingly, orchestration of robots has been selected as a method of robot coordination that allows the implementation of energy-efficient exploration strategies. As a result, an analysis of a specific orchestration method, OROS, and its exploration optimization is studied. Our hypothesis was that implementing and simulating the coordination of multiple robots allows the optimization of energetic parameters to be applied in real applications. For this purpose, robot orchestration has been implemented using a ROS network and Gazebo for robot simulation and Matlab and Simulink for the orchestrator. In robot simulation, an evaluation of the robot navigation and localization has been performed. Additionally, the use of a Gazebo battery simulator is used, and the battery parametrization is defined. Moreover, the robot exploration behaviour has been developed by implementing a state machine whose functionality has been explained. In the orchestration, the optimization problem is implemented in Matlab, which allows the evaluation of several scenarios depending on the exploration area and robot characteristics. For the path planning and ROS network communication with the robots, a Simulink model is used. The main elements of the orchestrator are evaluated and defined. Then, exploration simulations have been done with the OROS implementation. Firstly, simulations with different battery capacities have been performed using one robot in an obstacle-free scenario. Lastly, simulations with one and two robots have been done using four different scenarios. Then, the results of all simulations are evaluated. Furthermore, the comparison between one and two robots OROS estimations and simulation is realised. The contribution of collaborative robots is evaluated and discussed. Finally, we conclude that orchestration optimizes the battery energy savings while providing a coordinated exploration strategy for collaborative mobile robots. However, this method proves to be dependent on the orchestrator energy consumption estimations. Therefore, some differences between mathematical prediction and simulations results might be observed.

This thesis proposes a framework for configuring Task Planning Problems flexibly in an automatic manner using two main modules which are the Perception Module and the Reasoning Module. In order to automatize the overall process, initially, a knowledge layer is generated manually, in which the information regarding the environment is stored using ontologies, whereas the environmental state where the task is taking place is observed with the help of the Perception Module. The knowledge layer is then reasoned within the Reasoner Module in order to automatically configure task planning problems by filling Planning Domain Definition Language (PDDL) [1] files. During this reasoning process, the information retrieved from the Perception Module is used. In this paper, both of these modules mentioned above are explained in detail before providing the results separately for each module. Then, in addition to individual results, a scenario is created within a lab environment to test the overall system including both modules. Furthermore, alternative areas where the Reasoning Module implementation can be benefited from is also discussed

Todavía hoy, en la industria, hay algunos procesos de manipulación de piezas que se realizan manualmente, como la tarea de random bin picking. ́Esta consiste en extraer piezas que están dispuestas aleatoriamente dentro de un contenedor y colocarlas en otro lugar, como, por ejemplo, una cinta transportadora. Para resolver esta tarea de manera automatizada, en los últimos años se han desarrollado soluciones industriales, como el software InPicker. Este programa integra una solución de visión artificial que estima la posición y orientación de las piezas del contenedor. A partir de esta estimación, calcula donde debe colocarse la garra del robot para coger la pieza. En general, para mover el robot hasta la posición calculada, se manda la posición on de agarre al robot y éste calcula la trayectoria para llegar. Como las piezas están dispuestas aleatoriamente, la trayectoria que tiene que hacer el robot es distinta para cada pieza. Por esta razón, aunque el espacio esté bien diseñado, no es posible prever todas las trayectorias que hará el robot en el proceso de vaciado. Por lo tanto, pueden aparecer (i) problemas de colisiones con objetos del entorno, (ii) problemas de singularidades del robot, (iii) de límites articulares o (iv) de alcanzabilidad. De aquí se desprende el objetivo principal de este proyecto: que un brazo robótico sea capaz de coger todas las piezas alcanzables que haya localizado una cámara, sin interrumpir en ningún momento la producción. Para conseguirlo, se debe usar un planificador de trayectorias que tenga en cuenta el modelo cinemático del robot y los objetos del entorno. El primer aspecto abordado en este proyecto es qué algoritmo de planificación de trayectorias y con qué parámetros se obtienen los mejores resultados en un escenario típico de random bin picking con un robot manipulador. Para ello, se ha diseñado y realizado un análisis comparativo entre 22 planificadores de trayectorias. Para escoger el planificador más adecuado, se han valorado aspectos como el porcentaje de resolución de trayectorias y el tiempo empleado en ello. Con el planificador obtenido, se ha ejecutado un segundo análisis variando sus parámetros para encontrar con cuáles se obtienen trayectorias de mayor calidad –cortas y suaves– y empleando el menor tiempo posible en calcularlas. Una vez se ha encontrado el planificador más adecuado para el caso de estudio, se ha implementado en el InPicker. Se han observado las mejoras producidas respecto a los resultados obtenidos con el planificador que el InPicker utilizaba previamente. Además, se han desarrollado dos mejoras adicionales en el algoritmo ya existente: (i) poder añadir puntos de pre-agarre y (ii) añadir a la planificación la nube de puntos generada por la cámara. Con la primera, el algoritmo permite planificar la trayectoria para coger la pieza pasando por los puntos de pre-agarre definidos por el usuario. Con la segunda mejora, se pueden evitar las colisiones entre la garra del robot y las otras piezas del contenedor. Llegados a este punto, con el planificador ejecutándose en el InPicker, se debe enviar el resultado de la planificación al robot. Además, el robot tiene que mandar señales al InPicker para, por ejemplo, indicar cuándo tiene que buscar una pieza. Para establecer estas comunicaciones robot-InPicker y crear el programa del robot para que siga la trayectoria planificada, se ha desarrollado una URCap para los robots de Universal Robots. Esta URCap contiene una interfaz gráfica a través de la cual el usuario solo debe definir los paŕámetros mínimos para ejecutar el vaciado del contenedor. Con los datos introducidos por el usuario, la URCap internamente establece conexión con el InPicker y mueve el robot siguiendo la trayectoria calculada. Con esto, se consigue que el usuario final no tenga que programar nada para utilizar la planificación de trayectorias, sino que solo deberá definir sus necesidades a través una interfaz interactiva

Robotics is one of the fastest growing fields in recent decades. It is increasingly common to find robotic systems in our day to day. In the industry, these systems are replacing the traditional methodology and processes used until now, with many processes that have become automated by robots. Most of these tasks now in the hands of machines are repetitive and can become impossible to perform or even dangerous for people. One of the fields in which robotics is making its way is agriculture, in which rudimentary techniques and tools are used for most processes, and there are more and more robots that automate these processes avoiding that they must be carried out by humans. Earth Rover Europe is a company dedicated to robotics applied to precision agriculture. Its mission is to help feed the growing world population by reducing food waste in its growth stage. Using automated systems, Earth Rover robots carry out agricultural activities currently performed by humans in highly inefficient ways. In addition, the company is concerned at all times to carry out these activities in the most ecological way possible in order to reduce the CO2 emissions produced in agricultural processes. The present work is part of one of these robotic agriculture projects: the weeding system. This project is based on the development of a robot capable of automatic weeding through the use of concentrated light. Therefore, the robot to be developed will be able to automatically detect weeds, point them with the concentrated beam of light at the precise focal length, and emit lightpulses of the right power to transmit all the heat energy necessary to kill the weeds. This document presents the fundamentals for the system's electronics design, the low-level control programming, and ROS integration. Also, some field test results are presented in order todemonstrate the robot behaviour in the real world

The objective of the thesis is to create a framework where complex Task and Motion planning problems can be executed in an easy manner in a real robot. As the robot carries out the tasks, a change in environment could possibly cause failure of a pre-planned task and motion plan. This report tackles two problems, first, creation of a task and motion plan and second, the ability to update the plan according to the change in the environment. The methodology uses Behaviour Trees as a tool to automate the offline planning and the online planning of tasks and motion. The framework enables the user to create behaviour trees and execute them in a real robot. This report also includes a scenario where Tiago robot is employed as a waiter to serve drinks. As future work, a scenario where Tiago robot is employed as a warehouse picker will be considered. To conclude, using the advantages that Behaviour trees offer, the framework is able to complete a complex task with maximum autonomy

This thesis develops and implements a visual odometry algorithm within the ROS software framework that is to be embedded into an autonomous rover manufactured by Earth Rover Europe, S.L., a company dedicated to the implementation of automated and robotic solutions in the agriculture industry. It also develops and implements the overall localization system that the rover is equipped with, which is based on a sensor fusion by an Extended Kalman Filter. Visual Odometry is explored as a powerful tool due to the efficiency of the algorithms currently developed and the low cost cameras that make its implementation possible. A proposition is made that the implementation of a visual odometry algorithm into the overall localization system of a mobile robot will make it more robust by adding visual odometry as a sensor source into a Kalman Filter-based localization system. This thesis explores this proposition by developing the and integrating the ROS packages necessary for a complete localization system, as well as providing a simulated environment created in Gazebo, a ROS compatible simulator. The work done in this thesis provides evidence that the addition of visual odometry to the rover an its localization process makes for a more robust system, and finds visual odometry a valuable addition to any mobile or fixed robotic platform.