Modeling Fatigue Level by Driver's Lane-Keeping Indicators

Abstract

Fatigue while driving leads to significant deterioration of a driver's lane-keeping ability and increases the risk of crashes. Also, the lane-keeping indicator may be a good predictors of fatigue. A study using the Tongji high-fidelity driving simulator investigated the relationship between fatigue and lane-keeping indicators. During about 1 hour monotonous highway driving for each driver, driving data of 30 participants including lateral position, speed, and steering wheel movement data were collected. Meanwhile, Karolinska Sleepiness Scale (KSS) was recorded to measure driver's fatigue scale. To estimate driver's lane keep ability appropriately, standard deviation of lateral position (SDLP), lane crossing time-space area (LCTSA), standard deviation of steering wheel speed (SDSWS), and steering wheel reversal rate (SWRR) were measured. While controlling for other contribution factors including driving speed and road alignment, a multilevel ordered logistic model was established using Winbugs software. The research found that experienced drivers have a higher threshold for KSS. SDLP, SDSWS, and SWRR show significant positive relation to fatigue level. Another finding is that the differences between two adjacent KSS thresholds increase with KSS value.