HSI

XR and eye tracking in construction sites for accident prevention

Initial situation

A better understanding of how people perceive and respond to hazards could provide useful insight into hazard recognition. Gaze behavior can help identify patterns in which hazards are noticed or overlooked and could provide insight into future approaches to construction-site safety.

Project goal

The objective of this project is to develop and evaluate a gaze-adaptive warning system for hazard recognition in a VR construction-safety simulation. By comparing hazard-identification performance with and without the warning system, the project investigates whether gaze-adaptive feedback affects the number of correctly identified hazards and the time required to complete the hazard-identification task.

Top-down view of the first scenario, in which the participants need to identify hazards
User testing
Collection of hazards for the project
Grabbed drill, required for the second scenario
Top-down view of the second scenario, in which participants also need to complete other minor tasks
Slippery hazard next to warning sign, which denotes the hazard having been found.
Interactable keypad, used in the second scenario.

Solution developed and its benefits

For this project we investigated how gaze-adaptive attention guidance in a VR construction-site simulation affects hazard identification, task completion time and subjective workload. For this purpose, we developed a prototype that uses eye tracking on the Meta Quest Pro to adapt visual warnings based on participants' attention. It contained the following elements:

  • A simulated VR enviroment
  • Two scenarios to conduct tests in
  • A gaze-adaptive warning system
  • Hazards for participants to identify
  • Tasks for users to complete to test the warning system while distracted

Our experiment showed, that participants with the warning sytsem were more successful at finding hazards in the simulation, but also took longer to complete the task and reported increased amount of focus necessary to do so, when compared to the group without warnings.

While our findings show promising tendencies, our simulation still has room for improvemnts and a larger testing sample to improve the accurarcy of our findings.

Key terms

  • VRConstruction site simulation in full VR
  • eye-trackingUser's eye gaze tracked with Meta Quest Pro
  • safetyIncreased safety due to better awareness on construction sites

Customer

Logos

Stanford University
Dr. Iro Armeni
450 Jane Stanford Way
CA 94305, Stanford, United States
https://www.stanford.edu/

Team

Joshua Wyss

Advisors
Prof. Arzu Cöltekin, Dr. Elif Gurcinar Andreas Leu