Researchers from University College London (UCL), the University of Tokyo and University of the West of England, are developing an AI-enabled VR safety tool designed to monitor how construction workers visually search for and recognise risks and warning signs.
The EyeISENS project combines immersive VR, eye tracking and AI to assess visual attention during construction tasks. The team has developed the immersive simulation environment and is now undertaking experiments to develop AI-based visual-attention measurement models. Balfour Beatty and Japanese civil engineer Yokogawa Bridge Corporation are involved in the research.
Preliminary findings indicate that temporal gaze sequences can predict construction hazard identification and safety behaviour, opening the possibility of using visual-attention data to support more personalised and responsive safety systems.
The technology is particularly promising for supporting adaptive safety training, real-time feedback and intelligent safety systems that respond to workers’ distinct visual attention and behavioural profiles, according to the research team.
Professor Abdul-Majeed Mahamadu MCIOB MRICS of the Bartlett School of Sustainable Construction at UCL leads the UK team. He said: “This project is helping us move beyond simply observing where workers look. By combining cutting-edge tools and AI, we can begin to understand how visual attention influences workers’ safety behaviour and cognitive state in real-time. This could be used to provide more personalised training and the development of intelligent human-aware construction safety tools.”
Individual, site and cultural practice
Professor Pang-jo Chun of the Department of Civil Engineering at the University of Tokyo leads the team in Japan. He added: “Construction safety cannot be achieved by rules and inspection checklists alone. What matters is capturing, through temporal gaze sequences, how workers search for hazards and at which moment a miss occurs.
“The way hazards are found is shaped not only by individual differences, but also by looking habits rooted in site practice and culture. In this project, we are working with the UK team to compare the same tasks in near-site environments in Japan and the UK, so that shared cognitive mechanisms can be distinguished from country-specific search tendencies.
“The aim is not to transfer one country’s way of directing attention unchanged to the other, but to design training and support with a clear understanding of each other’s habits; that is the more direct path to improving safety in both countries. We hope to use the findings for more personalised safety training that takes culture into account, and for human-aware construction support.”
The project is supported by the British Council’s Research Collaboration Programme, as part of the International Science Partnerships Fund, and funded by the UK’s Department for Science, Innovation and Technology.
