Given the individual differences in driving behavior, it would be ideal to provide the drivers with a self-customizing driver-support system that adapts itself to different driving styles. Toward this goal, the present paper proposes the idea of
RFind (Risk Feeling of Individuals), which attunes itself to individual drivers, by extending the notion of risk feeling (
RF), the idea we proposed in the previous studies to estimate, on average, the drivers' risk feeling when they are closing on the lead car. The proposal is followed by the description of an experimental test we conducted to compare the performance of
RFind and
RF.
To obtain
RFind from
RF, we normalized the two terms of the latter, namely, 1/
THW (=the velocity of the host vehicle / the distance between the host and lead vehicles) and 1/
TTC (=the relative velocity of the two vehicles / the distance between the two vehicles), respectively, by individual characteristics, where they were defined to be the averages of 1/
THW and 1/
TTC for individual drivers at the brake onset when closing on the lead car.
To put the idea on a test, we compared the extent to which the timing of “gas off” and “brake on” can be separated in terms of
RFind and
RF, using a driving simulator experiment data when closing on the lead vehicle. The results showed that the true positive rate of the brake on was 98.6% and the false positive rate of the gas off was 0.9% when
RFind was used for the estimation, demonstrating remarkable improvements in order of 12.9% and 89.2%, respectively, when the estimation is based on the
RF.
View full abstract