Eyes on the Curve: Visual Constraints and Steering Accuracy in a VR Driving Simulator

Abstract

Human vision is critical to safe driving, enabling drivers to track the curvature of the road, maintain lane position, and anticipate hazards. Although traditional driving studies have identified key gaze behaviors, the mechanisms by which visual field manipulations affect steering remain difficult to isolate in real-world settings. This study uses immersive VR to manipulate visual input and investigate its effects on steering and gaze behavior. We introduce a depth-aligned aperture method allowing controlled manipulation of the visible field relative to road geometry. Twenty-eight participants drove through curved road segments in a VR simulator under seven conditions: normal vision, tangent point-aligned tunnel vision, misaligned offsets (±0.75°, ±1.5°), and an occluded central reference. Integrated eye tracking and vehicle telemetry assessed lateral control, gaze distribution, variability, and pupil dilation as a proxy for mental workload. The results show that restricting vision to a foveal region significantly impairs steering accuracy. Even when aligned with the tangent point, foveal information alone is insufficient to obtain accurate control, whereas misalignment induces systematic performance degradation. These findings provide causal evidence that visuomotor control in curve driving depends on the integration of multiple visual cues, including peripheral information. More broadly, this work demonstrates how immersive VR can be used to causally investigate the visual information supporting complex visuomotor behavior.