Effects of Simulator Motion on Driver Steering Performance with Various Visual Degradations

Abstract

This paper investigates the effects of simulator motion on driver steering performance, and how this depends on the available visual information and external disturbances such as wind gusts. A human-in-the-loop driving experiment was performed in which twelve participants steered a fixedvelocity car to follow a winding road (target tracking, TT) while suppressing side-wind gusts (disturbance-rejection, DR). Driver performance with and without motion feedback is compared in six tasks: "regular" lane-keeping with optic flow, centerline tracking with optic flow, and centerline tracking without optic flow, all with both 5 and 100 m of preview. Performance is calculated in the frequency domain to separate TT and DR contributions. The results show that motion feedback always yields improved DR performance, but the actual improvement depends strongly on the available simulator visuals. TT performance is generally unaffected by motion feedback, except when preview is limited. We conclude that simulator motion is required to evoke realistic driver performance in tasks where substantial external disturbances are present, but not in disturbance-free tasks where a winding road is being followed.

A Simulation Study of Human Sensory Dynamics and Driver–Vehicle Response

“…There is also a small weighting on visuallyperceived lateral path error, which is necessary to prevent drift of the vehicle away from the target path. The small weighting on vestibular measurements for a target-only task is consistent with the experimental findings in [15].…”

“…Although the target path in this case (K ff = 0) is a straight path, it is still important for the driver to preview and follow the target path, and so the weighting on the target angles is still the largest of the measurements. The nonnegligible weighting on vestibular measurements for a disturbance rejection task is consistent with the experimental findings in [15].…”

“…A novel parameter, the forcing function gain, is introduced to quantify the proportion of the task that is path-following. The results complement and explain the experimental findings in [15]. In addition, the effect of motion scaling (typically present in a driving simulator) on the relative contributions of visual and vestibular sensory feedback is quantified, and the effect of the forcing function gain on the accuracies of the driver's perception of motion state and slip angle are calculated; these findings from the model provide new insight to the role of sensory feedback and state estimation in driver perception.…”

“…The visual system is found to be the dominant sensory system, with the influence of vestibular measurements increasing with the proportion of random disturbances on the vehicle. This result from the model explains the experimental finding in [15]. If the physical motion is scaled down (as might be experienced in a driving simulator) the weightings of the sensory measurements are not affected until the motion perception thresholds are reached.…”

“…Non-parametric frequency response models identified from driving simulator experiments have been used to obtain insight to a driver's use of sensory feedback. In [15] it was concluded that physical motion in the driving simulator is needed to obtain realistic driving behaviour when the vehicle is subjected to external disturbances. Physical motion was found not to be necessary when the driving task involved following a winding road in the absence of external disturbances.…”