DRF Lunchtime Clinic - Diana Davletshina & Percy Lam

We are always very proud to present speakers from our programme here on Digital Roads of the Future. Today, we have the pleasure of the first of the Digital Roads PhDs who are reaching the end of their time with us here. Diana Davletshina and Percy Lam are two exceptional PhD students whom we have watched grow in their talents and confidence over the last three years. They will be presenting work from their PhD theses: 

Automating Road Geometric Digital Twin Construction and Maintenance  and  Constructing and Maintaining Geometric Digital Twins of Road Conditions 

 Our speakers

Diana Davletshina Percy Lam Diana Davletshina is a PhD student within the EPSRC Industrial CASE studentship at the Department of Engineering in partnership with National Highways, Costain and the Digital Roads project.  Diana's research focuses on automating the process of generating digital twins of roads using existing large-scale visual and spatial datasets to reduce costs and enable performance optimisation, failure prediction and future scenario modelling. Diana holds a Master's degree in Data Science from the Ludwig Maximilian University of Munich and a Bachelor's degree in Computer Science from Innopolis University. Before joining PhD studies, Diana worked as a software engineer and an ML researcher Percy Lam is currently a PhD student at the University of Cambridge. His PhD research investigates the construction and maintenance of a digital twin for roads, specifically focusing on detecting road defects. He graduated from the University of Cambridge and began his career in geotechnical engineering, contributing to the expansion of the Hong Kong International Airport. He later became a Chartered Civil Engineer and a Member of the Institution of Civil Engineers. Currently pursuing a PhD, he has researched using different paradigms of artificial intelligence and computer vision in automating image annotation, generating synthetic data and building detectors for on- and off-pavement road defects. The goal is to develop efficient, explainable and accessible solutions to enhance the operation and maintenance of infrastructure.

Title of presentation:  

Automating Road Geometric Digital Twin Construction and Maintenance 

Title of presentation:  

Constructing and Maintaining Geometric Digital Twins of Road Conditions 

Abstract:  

The construction and transportation sectors are vital to economic activity and public safety, but they face challenges in cost efficiency, asset longevity and sustainability due to outdated, labour-intensive asset management practices. While Geometric Digital Twins (GDT) offer a promising avenue for decision-making, predictive maintenance and operational planning, large-scale adoption is limited by high costs and manual effort. Addressing this, my thesis proposes a framework to automate road GDT construction and maintenance using point clouds, achieving strong results in segmentation, meshing, relationship modelling and change detection. The presentation will focus on the key highlights of this work as presented in my thesis. 

Abstract:

Digitisation of road infrastructure remains a significant challenge, despite the availability of enabling technologies and visual data. Image data requires preparation and annotation, and may lack certain types of defects for training detectors. Road defect detection currently focuses on the pavement, with off-road assets and condition change over time remaining an open question. My project involved exploring solutions such as enhancing automation in image annotation and creating synthetic defects to supplement inadequate data. The work also improves defect detection by constructing a multi-defect classifier that combines traditional object detectors with vision language models configured to clarify contextual hallucinations, and advancing capabilities to update asset conditions. These solutions contribute to building the ultimate geometric digital twin of road conditions, empowering stakeholders to visualise, implement and review decisions across all phases of road maintenance and optimisation.