Research lines and projects
AI applied to AEC: 3D capture, digital twins and energy optimization.
My research at TECNALIA applies data science and artificial intelligence to the AEC sector (architecture, engineering and construction): from real-time 3D capture for BIM to digital twins and energy optimization.
Real-time scan-to-BIM and 3D capture
Scan-to-BIM methods based on laser scanning or photogrammetry require specialized hardware and offline processing. With Fast-Asset we developed a real-time, on-site framework that combines mixed-reality capture, learning-based detection and incremental 3D reconstruction to generate BIM-ready point clouds of small MEP elements.
Contributions and highlights:
- Mixed-reality capture with HoloLens 2, deep-learning object detection (YOLOv7) and incremental 3D reconstruction (Open3D).
- Capture-to-model time 3 to 10 times shorter than with laser scanning or photogrammetry, with comparable surface roughness.
- Immediate visual feedback to validate and recapture on site.
- Web evolution on mobile with WebXR.
Digital twins and AI in construction (AEC)
Application of artificial intelligence to the building life cycle, with BIM/IFC and digital twins as the foundation. My work starts from data science and analytics for architecture, engineering and construction.
Contributions and highlights:
- Keeping small assets up to date while maintaining digital twins (Fast-Asset).
- Talk Transforming Construction with Digital Twins and AI at AI in AEC 2024 (March 2024), with Sarah Noyé.
- Day-to-day at TECNALIA (Digital Construction): VR/AR, real-time object detection, evolutionary algorithms, time series and team-wide MLOps practices.
Energy optimization and control
Model predictive control (MPC) with evolutionary algorithms for combined electrical and thermal energy storage systems, on a foundation of operations research.
Contributions and highlights:
- MPC on a generalized mathematical model for electrical and thermal storage, with a method to incorporate physical constraints.
- Comparison of three evolutionary algorithms: the Genetic Algorithm achieves the largest saving on the energy bill, with an average daily saving of 38.7%.
- Published in the EC³ 2024 proceedings (pp. 703–710).