Assessing the effects of road geometry and fidelity on driver behaviour in driving simulator experiments

The paper presents a set of programs created specifically to investigate the driver's behaviour along a road circuit when the driving simulation scenario changes in the quantity and quality of the displayed Level of Details (LOD) (fidelity), as well as when the road curvature ratio (geometry) varies. In the experiments, carried out on a sample of 29 drivers who drove on the same circuit twice, data on vehicle features (speed, trajectory, and gas pedal usage), physiological factors of the drivers (SC, skin conductance), and information from an eye-tracking equipment (gaze) were collected. The programs used in this study are designed to study the complex and position-based information present in the data that cannot be analysed by classical statistical methods. The results are twofold. On one side, they show the applicability and usefulness of the proposed functions, which are capable of extracting information and referring it to a specific point in the circuit. On the other side, they give effective insights on driving behaviour. When driving is more demanding (i.e., along curvilinear stretches), variations between different scenarios are nearly non-existent, and the effect of geometry changes greatly depending on the signal being investigated (speed, trajectory, gas pedal, gaze, skin conductance). The effects of vehicle speed are essentially the same for all drivers and are closely tied to the geometry of the road. Trajectory and gas-pedal usage signals exhibit diverse effects on drivers, indicating that their choice of trajectory and frequency of gas-pedal usage is undoubtedly influenced by their driving style as well as by geometry. Skin conductance appears to be significantly altered where driving is more challenging, but it varies greatly from driver to driver. The results of gaze analysis also demonstrate that the driver's behaviour strongly depends on geometry and much less on LOD and loop sequence.