RoadMind: Investigating Road-User’s Behaviour in Immersive Experiences

Sammanfattning

Understanding how drivers perceive and respond to traffic hazards is critical for designing safer road systems and human-centred autonomous vehicles. This study investigates hazard perception and driver behaviour in virtual reality (VR) driving environments through a multimodal approach. A VR simulation was developed using Unreal Engine and a Meta Quest Pro headset with integrated eye tracking, enabling controlled exposure to urban traffic scenarios. Eye tracking, head movement, and vehicle control data were collected from 17 participants across three urban areas of varying complexity. The study examined how environmental complexity influences gaze behaviour, driving actions, and collision outcomes. Results indicate that hazard detection failure is more closely associated with the spatial distribution of gaze than with overall gaze volume. Broader visual engagement was associated with reduced collisions in complex scenarios, while concentrated gaze and delayed reactions were more prevalent in collision events. Road geometry also emerged as a key factor: curved roads constrained environmental scanning and were linked to higher collision frequencies. Based on these findings, a conceptual model is proposed describing the temporal relationship between environmental complexity, visual engagement, fixation timing, and collision probability. The study demonstrates the potential of integrating VR simulation and multimodal behavioural data for driving safety research, and identifies future directions including larger samples, enhanced simulation realism, and statistical validation.