Dr Yuandong Pan
- Marie Sklodowska-Curie Future Roads Fellow
- Postdoc Researcher at Civil and Environmental Engineering, School of Engineering, Stanford Universit
About
Dr Yuandong Pan currently holds the position of Marie Skłodowska-Curie Future Roads Fellow at the University of Cambridge. His research interests encompass the creation of geometric digital twins from multi-modal data sources, including point clouds, images, thermal images, and ground penetrating radar data. Dr Pan's expertise lies in developing automated methods for generating digital twins for different facilities, significantly reducing human labour. He earned his joint Ph.D. from the Technical University of Munich, Germany, where he specialized in creating information-rich digital twins of indoor environments. His research is primarily focused on automatic methods for constructing digital twins from point cloud data and images.
Fellowship period: 15 June 2023 – 31 August 2025
Research
Strategic Themes
Creating virtual replicas of physical infrastructure, using advanced technologies like sensors, AI, and IoT.
Decision making in Infrastructure Management Supported by Digital Twins.
Investigate the use of multi-modal data in creating digital twins.
Using advanced data processing techniques to analyse and evaluate the condition and performance of various physical infrastructure.
Research Project
Project Title: Construct and maintain a highway digital twin from multi-modal data
Theme: Digital Twins
Abstract:
Highways serve as fundamental infrastructure within the transportation network, with road freight accounting for 79% of domestic movement in the UK in 2018 and cars, vans, and taxis representing 88% of passenger kilometres in 2021. Emerging from the manufacturing sector, digital twins have gained prominence in the infrastructure domain, particularly within the Architecture, Engineering, and Construction (AEC) sector. A digital twin is universally acknowledged as a representation linking an asset or system, offering transformative potential for enhancing lifecycle processes of transportation infrastructure, such as highways, railways, and bridges. Despite the promise, challenges persist in efficiently constructing useful digital twins from diverse data sources due to manual reconstruction efforts, disjointed data linkage, and insufficient data structure design.
This project aims to automate digital twin construction and maintenance using Artificial Intelligence (AI) methods applied to multi-modal data, encompassing laser-scanned point clouds, RGB images, thermal images, and ground-penetrating radar (GPR) data. Leveraging the capabilities of deep learning in processing point cloud and image data, the project has amassed a substantial annotated dataset suitable for training models, necessitating high-performance computing resources.
Based on the thermal data GPR data in CamHighway dataset, I generated thermal point cloud data for a road section and prepared the GPR data samples for training machine learning models.
The anticipated outcomes of this research encompass: a) contributing a registered and annotated multi-modal dataset and trained deep learning models to the research community; b) introducing an automated approach to developing highway digital twins, mitigating costs and human involvement; and c) facilitating industry advancements through the integration of digital twins in real-world applications, enhancing maintenance procedures, optimising planning and construction processes, and monitoring asset conditions.
Alignment with SDG: This research contributes to SDG 9: Industry, Innovation, and Infrastructure and SDG 11: Sustainable Cities and Communities.
Project TRL: The current research is positioned at a TRL 4-5, focusing on validating key findings through controlled environments and stakeholder engagement.
Dr Yuandong Pan is a distinguished researcher specializing in developing geometric digital twins from diverse sources, such as point clouds and images. His primary focus is on devising automated methods to create digital twins for various facilities, thereby minimizing manual intervention.