Kontessa Ioanna Zorpala¹, Joan Lopez-Moliner¹; ¹University of Barcelona (UB)

Although there are many studies addressing optic flow in the past, less is known about how gaze behaviour and optic-flow dynamics operate in natural and high-velocity scenarios, like cycling, where rapid changes in heading, curvature and environmental conditions create different visuomotor demands. Our study aims to characterise the retinal information available during cycling and to determine how flow-based measures relate to the future path allowing for steering control. Participants (n=4) rode a paved, winding road (3 km, 6.5% average negative slope), while wearing a mobile eye-tracker (Neon pupil labs, 200Hz) with a world camera (30 Hz). Maximum speeds were between 50 and 70 km/h. Gaze, world-camera and GPS data were synchronised, enabling the reconstruction of the visual input associated with each location point. Optic-flow was computed in two ways: (1) directly from the world-camera and (2) from a reconstructed retinal-flow field, created by cropping each frame around the gaze location (Matthis et al 2022) resulting in head and retinal frame of reference respectively. Curl and Focus of Expansion (FOE) were extracted for both references, while road curvature was obtained from GPS data. Gaze in world coordinates was computed using eye-tracker and IMU data and was aligned with road coordinates. We used cross-correlation functions (CCFs) and obtained peak correlations and their temporal lags, between gaze direction, curl, FOE with respect to road curvature. Gaze showed a consistent anticipatory relationship to upcoming curves, indicating that cyclists orientated their gaze toward the turn-tangent before entering it. Both optic-flow measures correlated with road curvature, but retinal-flow curl displayed stronger and more temporally structured coupling, particularly for faster subjects. Retinal FOE-route correlations showed a similar pattern. Overall, the retinal frame of reference supplies robust predictive information for future-path estimation and steering, expressed through FOE, curl, or both, highlighting its contribution to high-speed cycling.

Acknowledgements: The research was supported by Grant PID2023-150081NB-I00 funded by MICIU/AEI/10.13039/501100011. KIZ was supported by fellowship PREP2023-001890.