The system described in this paper warns the driver, if there is a risk of leaving the lane caused by drowsiness or inattention. As a criterion for lane departure, the Time-to-Line-Crossing (TLC) is used based on measurements of vehicle speed and lateral position. By means of image processing, the lane markings are extracted from the video image and the position of the vehicle inside the lane is determined. The very effective warning signal - the „rumble strip noise“ - guarantees a fast and correct reaction of the driver. This was verified during several experiments at the driving simulator. To get an additional, more predictive criterion for drowsiness, data records during drowsy driver experiments at the simulator were studied. As a result, it seems to be possible to predict drowsiness by an algorithm only based on vehicle parameters.