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

This project explores the enhancement of motor controllers in the mobile robots developed by the CDEI, simplifying the control system by reducing the number of elements and, consequently, the weight, while improving the robots' overall efficiency. To achieve this, the study evaluates the implementation and use of Ingenia EtherCAT servo drives, specifically the EVSNET-E and EVE-XCR-E models. There are two primary objectives for the project: first, to validate the proper functioning of a printed circuit board (PCB) that integrates the EVS-NET-E drive, ensuring it enables real-time control of a brushless motor; and second, to develop a control program for installing this device in the robots. Additionally, the project establishes a methodology for integrating these drives into robots, including configuration guidelines and programming examples. To validate the PCB, the MotionLab3 software was employed, allowing the configuration of the attached drive. During this process, the various functionalities of the drive and its associated circuits were analyzed, making the necessary modifications. For the design of the control program, the fucntions of the Python Ingeniamotion library were used. The development of the code followed an iterative approach to ensure proper functionality at all stages. The control program was written in Python and subsequently implemented in ROS. Validation of the code was carried out through tests with the actual motor and simulations of the robot's movement. The analysis of the PCB revealed that while some components, such as communication and feedback circuits, were successfully verified, achieving a fully functional board requires a more detailed analysis to identify and address the causes of motor control issues. The design of the control program resulted in the creation of a final modular and versatile control program that allows for easy modifications and is compatible with any of the Ingenia controllers. Additionally, example codes were created to demonstrate the various functionalities of the drives. Therefore, despite some challenges encountered during the PCB analysis, the project demonstrated the feasibility of integrating Ingenia EtherCAT drives into the robots, obtaining the benefits that this involves, such as improved performance and reduced overall programming complexity.

Statement of the Issue The robotic manipulation of deformable materials represents a significant challenge due to its dynamic complexity and the difficulty of modeling and controlling the imprecisely. Unlike rigid objects, deformables, such as fabrics and ropes, exhibit variable configurations that demand innovative solutions. In this context, reinforcement learning emerges as a key tool for robots to learn manipulation strategies through autonomous interaction with their environment. Project objectives This project focuses on the robotic manipulation of deformable objects using reinforcement learning. The main objective is to develop a neural network that enables a robotic arm to autonomously learn how to manipulate deformable materials. This project addresses the implementation of advanced simulation environments and the analysis of reinforcement learning algorithms. Methodology The document details the theoretical elements, simulation tools, and control strategies used. Several frameworks have been introduced to structure the simulated environments and train the agents. Additionally, various simulation platforms have been explored, evaluating their compatibility and suitability. Main Results The experimental results obtained in increasingly complex environments highlight the ability of agents to manipulate linear deformable objects with a certain degree of precision and stability. The tools and methodological approaches have proven robust, providing a solid foundation for future research and industrial applications. Key Conclusions The key conclusions of this work show that the progressive development of simulations, combined with advanced reinforcement learning algorithms, significantly facilitates learning in complex tasks such as the manipulation of deformable objects. This presents a great advantage compared to methods that rely on control algorithms or pre-programmed behavioral sequences. Future Research The developed methodology can be applied to other types of deformable materials, such as fabrics with greater dynamic complexity. Furthermore, the results obtained could lead to the optimization of control strategies based on real data from advanced tactile sensors, opening new opportunities for implementation in industrial and service environments.

This project document presents the development and evaluation of an autonomous robotic system designed to enhance irrigation management in agricultural environments, specially the region of Penedès. The report is structured to guide the reader through the various stages of the project, from its conception to its practical implementation and future prospects. The document begins by setting the context for the project, explaining its origins and the motivations behind its development. These aspects are important to understand the need for innovative solutions to optimize resource use and improve efficiency. These sections provide insight into the challenges faced in modern agriculture. Following this, the project objectives are outlined, detailing the goals that the system aims to achieve. These objectives include the creation of a simulated environment using ROS2, the development of route planning capabilities, and the exploration of potential improvements in real-world scenarios. The implementation phase is discussed in detail, where the system's components are developed and tested within the simulation environment. Although the JetBot robot is proposed for future integration, the current project does not include its physical implementation. Instead, the document explains how communication with the JetBot would be established, providing a clear framework for its potential use. The actual implementation and testing were conducted using a different robot, where the practical application of the system is demonstrated.

Power electronic components are becoming essential for the present and future grids, giving useful services apart from integrating electrical vehicles (EV), renewable energies or batteries. Voltage Source Converters (VSCs) are a good option to ensure the stability of the grid when there is a fault, which is known as Fault-Ride Through (FRT). In most cases the fault is unbal- anced, which affects directly the control of the power electronic device, having to add a positive and negative sequence control. Despite the wide use of the symmetrical components in the lit- erature, their explanation is not clear in the literature. This thesis aims to explain what are these components and how they can be calculated. Furthermore, the analysis of the symmetrical se- quences is used to try to obtain a small-signal model of a VSC with positive and negative control directly from the sequences. The objective is to use the linear model to analyse the stability and design the controls taking into account the effect of the asymmetry. This thesis also explains the limitations of the presented methodology due to the coupling between positive and negative se- quence components at 0 Hz and 100 Hz. Finally, the obtained model is analysed to identify the differences in the dynamics with respect to a model with positive sequence control.

Power electronic components are becoming essential for the present and future grids, giving useful services apart from integrating electrical vehicles (EV), renewable energies or batteries. Voltage Source Converters (VSCs) are a good option to ensure the stability of the grid when there is a fault, which is known as Fault-Ride Through (FRT). In most cases the fault is unbal- anced, which affects directly the control of the power electronic device, having to add a positive and negative sequence control. Despite the wide use of the symmetrical components in the lit- erature, their explanation is not clear in the literature. This thesis aims to explain what are these components and how they can be calculated. Furthermore, the analysis of the symmetrical se- quences is used to try to obtain a small-signal model of a VSC with positive and negative control directly from the sequences. The objective is to use the linear model to analyse the stability and design the controls taking into account the effect of the asymmetry. This thesis also explains the limitations of the presented methodology due to the coupling between positive and negative se- quence components at 0 Hz and 100 Hz. Finally, the obtained model is analysed to identify the differences in the dynamics with respect to a model with positive sequence control.

In recent years, there has been considerable attention given to business practices and policies aimed at moving towards a more environmentally sustainable model. While this reduction in environmental impact is undeniable, it casts shadows on the other significant advantage of these policies: the improvement of economic profitability. The objective of this project is to test the hypothesis that Certain reverse logistics policies can substantially improve the economic performance of a company, alongside their impact on the environment, through simulations. These simulations will be conducted using the Arena and Simul8 programs, employed in the MUAR Master's program, and using a Catalan company dedicated to the production of power electronics as the model.

Aquesta Tesi de Màster presenta el disseny i implementació d'un enregistrador de dades CAN FD d'alt rendiment amb un potent software d'anàlisi. El registrador de dades és capaç de capturar i emmagatzemar dades de múltiples busos CAN2.0 i CAN FD, amb altes taxes de mostratge i sincronització precisa del temps gràcies a un emmagatzematge en binari. Aquesta solució també proporciona una aplicació basada en MQTT per a transmetre dades en directe a servidors arreu del món, convertint-la en un producte complet no només per a equips de motorsport, sinó per a totes les indústries que utilitzen protocols CAN per a comunicació IoT. El software d'anàlisi ITS permet la visualització i anàlisi de les dades capturades, amb funcions com filtratge, escalat i representació gràfica. Més dades no significa més resultats, per la qual cosa desenvolupar tècniques per comprendre la informació derivada de centenars de canals esdevé cada vegada més important en temps com els nostres, on els llacs de dades poden arribar a tenir múltiples gigabytes d'informació. El sistema ha sigut provat i avaluat pel seu rendiment i experiència d'usuari, amb resultats positius. Aquesta solució té el potencial de beneficiar diverses indústries, com automoció i l'aeroespacial, en l'anàlisi de rendiment, diagnòstic de fallades i control de qualitat

Aquesta Tesi de Màster presenta el disseny i implementació d'un enregistrador de dades CAN FD d'alt rendiment amb un potent software d'anàlisi. El registrador de dades és capaç de capturar i emmagatzemar dades de múltiples busos CAN2.0 i CAN FD, amb altes taxes de mostratge i sincronització precisa del temps gràcies a un emmagatzematge en binari. Aquesta solució també proporciona una aplicació basada en MQTT per a transmetre dades en directe a servidors arreu del món, convertint-la en un producte complet no només per a equips de motorsport, sinó per a totes les indústries que utilitzen protocols CAN per a comunicació IoT. El software d'anàlisi ITS permet la visualització i anàlisi de les dades capturades, amb funcions com filtratge, escalat i representació gràfica. Més dades no significa més resultats, per la qual cosa desenvolupar tècniques per comprendre la informació derivada de centenars de canals esdevé cada vegada més important en temps com els nostres, on els llacs de dades poden arribar a tenir múltiples gigabytes d'informació. El sistema ha sigut provat i avaluat pel seu rendiment i experiència d'usuari, amb resultats positius. Aquesta solució té el potencial de beneficiar diverses indústries, com automoció i l'aeroespacial, en l'anàlisi de rendiment, diagnòstic de fallades i control de qualitat

This project involves the design and implementation of an unmanned aerial surveillance vehicle capable of navigating autonomously and at the same time recognising its environment. It combines computer science, programming and artificial intelligence. In terms of computing, a Linux server is implemented using an embedded computer which is connected to the network via an Ethernet connection. At the same time, since it is equipped with a Wi-Fi card, it is also connected to the UAV, which acts as a master controller and is accessed via SSH. On the other hand, also a Homebridge server is also installed. This allows the system to be integrated with Apple's Voice Assistant. In terms of programming, the UAV SDK is used to control the vehicle with the other system blocks such as the autonomous flight algorithm, environmental processing or remote control with the computer keyboard. All these blocks are used to build the algorithm to perform the tasks. Finally, an API for the system is implemented to facilitate its integration with external systems such as the voice assistant. In terms of artificial intelligence, neural networks are used to estimate the depth of the image in real time

This project involves the design and implementation of an unmanned aerial surveillance vehicle capable of navigating autonomously and at the same time recognising its environment. It combines computer science, programming and artificial intelligence. In terms of computing, a Linux server is implemented using an embedded computer which is connected to the network via an Ethernet connection. At the same time, since it is equipped with a Wi-Fi card, it is also connected to the UAV, which acts as a master controller and is accessed via SSH. On the other hand, also a Homebridge server is also installed. This allows the system to be integrated with Apple's Voice Assistant. In terms of programming, the UAV SDK is used to control the vehicle with the other system blocks such as the autonomous flight algorithm, environmental processing or remote control with the computer keyboard. All these blocks are used to build the algorithm to perform the tasks. Finally, an API for the system is implemented to facilitate its integration with external systems such as the voice assistant. In terms of artificial intelligence, neural networks are used to estimate the depth of the image in real time