Vehicle control and drowsiness

It is known that drivers\u27 drowsiness and fatigue is accompanied by deteriorated vehicle control. Albert Kircher, VTI, and Marcus Uddman and Jesper Sandin, Virtual Technology, have investigated which factors are possible to use for the prediction and detection of fatigue induced impaired driving. In this study focus was on variables directly related to lateral control and steering behaviour aiming at a method capable of detection and prediction of impaired driving performance in real time and in an automated manner. Psycho-physiological variables, subjective rating scales, etc. are only briefly discussed. The first part of the report is a literature study. It served to examine the available knowledge in the field and to identify the most promising indicators of impaired driving. A large number of publications describing various approaches have been reviewed. The survey clearly indicates that no single indicator can be used to reliably detect drowsy driving. A combination of different measures is recommended, e.g. analysis of lateral control performance and eye blink patterns. Furthermore, it should be noted that even though there are a small number of commercial drowsiness detection systems available, no system provides a sufficiently reliable and scientifically proven method to detect a drowsy driver. In the second part of the report experimental data from previous driving simulator experiments were analysed. Signal processing techniques, such as frequency analysis by means of Fourier transforms, and statistical analyses were used to disclose if data could be used to determine drivers\u27 drowsiness as rated on subjective rating scales (Karolinska Sleepiness Scale). Specific attention was paid to investigate the potential of lane control data (steering performance) as a mean to estimate driver drowsiness. Measures related to lateral vehicle position, such as time to line crossing and deviation of lateral position, were also analysed, but were not found useful. The analysis of experimental data did not reveal any clear answer to what driver behaviour indicators are the most prominent to detect drowsy driving behaviour. However, the data analysis conforms in large to the findings in the literature survey: a single variable is hardly usable as drowsiness predictor. More advanced signal processing techniques could be more proficient for the aspired goal. Further investigations and analyses of driving behaviour data are needed. Combination of different variables, such as eye blink patterns and lane control measures, are expected to be more successful