Eyes on the road: A bicycle simulator study of cyclists’ gaze behavior while sharing the road with autonomous vehicles across infrastructure types

Integration of Autonomous Vehicles (AV) into urban traffic introduces new safety challenges for vulnerable road users, particularly cyclists. Unlike pedestrians, cyclists travel at higher speeds and frequently share lanes with motor vehicles, increasing the risk of conflict. Current research has mainly focused on driver–pedestrian interactions, leaving a gap in understanding how cyclists perceive and adapt to encounters with AVs. This study addresses this gap by analyzing cyclists’ gaze behavior across different infrastructure settings and AV traffic compositions, testing three undirected hypotheses on the effects of AV traffic penetration rate (TPR), shared- versus separated-lane infrastructure, and pavement color within separated lanes. The experiments were conducted using an advanced bicycle simulator that replicated a 200-meter urban road segment in Győr, Hungary. Fifty participants, equipped with wearable eye-tracking glasses, cycled through 11 scenarios representing variations in cycling infrastructure and AV traffic penetration rate (TPR) levels. The results show that infrastructure type was the dominant factor shaping gaze behavior in the tested simulator scenarios, while TPR did not produce a clear effect in this straight-road overtaking context. Statistically tested gaze metrics indicated broader visual scanning in physically separated lanes, particularly through larger saccade amplitudes, while the descriptive heatmap patterns were directionally consistent with this interpretation. By contrast, shared-lane scenarios were associated with reduced scanning spread and a narrower concentration of gaze on immediate threats. The color of the pavement (black vs. red lanes) showed only minor effects. These findings suggest that dedicated cycling infrastructure may support broader visual search strategies than shared-lane operation in this scenario family, providing evidence to support cyclist-centered urban road design in increasingly automated traffic environments.