Transforming Infrastructure Through Digital Twins
Infrastructure underpins modern society, yet much of it is ageing, increasingly complex, and expensive to maintain. Bridges, buildings, transportation networks, and construction projects require continuous monitoring and informed decision-making throughout their lifecycle. However, inspections are often labour-intensive, costly, and based on incomplete information. At the same time, the construction sector faces growing pressure to improve efficiency, sustainability, and resilience while reducing costs and risks.
Digital twins offer a promising solution by creating virtual representations of physical assets that can be continuously updated with real-world observations. Yet building reliable digital twins at scale remains a major challenge. Existing methods often struggle to capture the complexity of real-world infrastructure, particularly when observations are incomplete, noisy, or collected from multiple sensing platforms. As digital twins increasingly support critical decisions, there is also a growing need to understand the reliability and uncertainty of the information they provide.
Creating Digital Representations of the Built Environment
The research aims to develop the next generation of digital twin technologies for infrastructure and construction (here the CV4DT link?). Our goal is to enable the automatic creation, updating, and analysis of detailed digital models that accurately represent the physical world and support decision-making across the entire asset lifecycle.
The work focuses on understanding the built environment from sensor data, ranging from individual building components to large-scale urban infrastructure. Key research areas include semantic scene understanding, structured 3D reconstruction, neural scene representations, robot manipulation, and uncertainty-aware modelling. By combining geometric, semantic, and physical information, we seek to transform raw observations into intelligent digital twins capable of supporting inspection, monitoring, simulation, and planning.
Combining AI, 3D Vision, Robotics, and Physical AI
Our research integrates computer vision, photogrammetry, machine learning, robotics, and physical AI. We develop algorithms that learn from diverse sensing technologies, including laser scanning, imagery, mobile mapping systems, drones, and emerging neural representations such as Gaussian Splatting.
A major focus is enabling machines to understand infrastructure beyond geometry alone. Our methods identify semantic components, reconstruct structured building models, estimate confidence in predictions, and fuse information from multiple sources into coherent digital representations. We also develop benchmark datasets and evaluation frameworks that help measure progress and ensure that emerging technologies can be reliably deployed in real-world infrastructure and construction scenarios.
Enabling Smarter, Safer, and More Sustainable Infrastructure
The research has produced new methods for semantic object reconstruction, uncertainty-aware 3D modelling, infrastructure perception, robotics, electromagnetic wave sensing, and large-scale benchmark datasets for the built environment. These advances contribute to more accurate and trustworthy digital twins while reducing the effort required to generate and maintain them.
In the long term, this work supports a future in which infrastructure can be monitored continuously, construction projects can be managed more efficiently, and maintenance decisions can be informed by up-to-date digital representations of the physical world. By combining artificial intelligence with advanced 3D sensing technologies, the research contributes to safer, more sustainable, and more resilient infrastructure systems while laying the foundations for intelligent digital twins that bridge the gap between the physical and digital worlds.
Browse the many Construction Engineering Masters dissertations on
