Research projects

This project aims to investigate state-of-the-art learning and vision algorithms for integration into robotic systems to automatically inspect renewable energy systems, transmit the acquired images quickly while preserving their quality, and then generate detailed component assessments to determine a consistent repair and/or maintenance strategy. To do so, it first proposes segmenting the images to identify regions of interest, thus reducing the complexity of subsequent tasks. This allows the robotic system to efficiently track key parts during its autonomous exploration, allowing defect detection and compression tasks to focus solely on the area of interest. Thus, the energy system will first be autonomously inspected using real-time localization and 3D technology. For image compression, the project will develop an AI-based region-of-interest encoder that supports lossy and lossless coding. It will also explore how to build a robust detection model capable of identifying different types of defects at various scales. With all this information collected, the severity and energy loss caused by the defects will be examined to determine a predictive repair strategy for the renewable energy system. The use of this automated and intelligent maintenance and control system will directly improve energy efficiency by reducing downtime, optimizing action programs, and minimizing repair costs. The proposed neural algorithms aim to achieve energy systems that are more aware of the European ecosystem, while generating new opportunities, digitizing industry, reducing energy costs, and mitigating the effects of climate change.