Autor: Marco Ojer De Andrés

Tesis: Robotic Advancements for Smart Automation in Industry 4.0

Dirección: Elena Lazkano / Xiao Lin

Día: 2025eko irailaren 22an
Hora: 11:00h
Lugar: Ada Lovelace aretoa (Informatikako fakultatea)

Abstract:

"In the context of Industry 4.0, the growing demand for customized products has imposed the need for rapid flexibility and adaptability in manufacturing processes, requiring automation systems capable of dynamically adjusting to changing production requirements while maintaining high efficiency and reliability.

In this sense, the integration of technologies such as robotics, artificial intelligence, and digital twins enables smart automation, allowing systems to perceive their surroundings and make intelligent decisions to adapt to changing conditions. However, multiple challenges remain, such as enabling robots to perform complex tasks autonomously, simplifying the training of artificial intelligence models, meeting the stringent requirements of industrial environments for robustness, reliability and safety, and ensuring ease of use and maintenance for the existing workforce, among others. In addition to these challenges, hardware limitations in robotics, such as sensor inaccuracies, calibration challenges, and real-world physical constraints, further complicate the development of robust and reliable smart automation systems.

This thesis focuses on the advancement of robotics within smart automation, specifically addressing the challenges and opportunities presented by Industry 4.0 manufacturing environments.

A first research line tackles the challenge of improving the accuracy of robotic manipulators, which is crucial for applications requiring precise movements in dynamic environments.

It introduces a novel hybrid kinematic modelling approach, which combines classical kinematic modelling with neural networks which are employed for error compensation.

The second research line focuses on automating bin picking, a complex task that involves the integration of various technologies like computer vision, robotics, and Artificial Intelligence. This line aims to simplify the development and deployment of bin picking applications, addressing the complexities involved and making them more accessible and efficient.

It presents a plugin-based framework that provides modular components for building bin picking systems. Additionally, it introduces a synthetic data generator that reduces the need for real-world data collection to train AI models and a workflow that optimizes task execution for faster cycle times.

The last line addresses the broader challenges of reliability, safety, and efficiency in smart automation. It proposes a novel self-improvement method that enables continuous performance enhancement without interrupting ongoing operations. Moreover, it introduces a novel operational paradigm that leverages remote human operators to effectively handle errors and unexpected situations in flexible automation lines."