Inattentive driving is the cause of many devastating accidents on the roads each year and he cost of this increases annually; both in money and in lives. Many researchers around the world are currently working on new methods to reduce the effects of inattentive driving and hopefully these systems will be available on the market soon. Inattentive driving usually results in lane departure crashes, and to warn these drivers before the accident occurs, Lane Departure Warning Systems (LDWS) have been developed and are currently used in the automobile industry. This system sends out a warning every time a lane marking is crossed, which can result in redundant warnings that are annoying to the driver. his project tries to investigate if it is possible to create a warning system for inattentive drivers by using only in-vehicle signals, which means signals that are already available in the truck. The system should also use this detection of inattention to reduce the number of redundant warnings from the LDWS and create a better warning system. To solve this, the inattention detection was divided into two different subcategories: drowsiness and distraction, each detected separately. Drowsiness works with long timespans and distraction works with short time-spans. This project has been conducted at Scania in collaboration with Pauline Deram. In this report the drowsiness detection algorithm is described. To be able to investigate how the drowsiness detection should be carried out, an extensive literature study was conducted. Experiments with professional drivers were carried out and the signals were processed in an attempt to find a reliable drowsiness detection algorithm. This algorithm was created, based on the literature studies and the experiments and it showed good results. Both the drowsiness detection and the distraction detection (created by Pauline Deram) were combined into an inattention grade and put into a new warning algorithm the “Adaptive Lane Departure Warnings System” (ALDWS). This ALDWS works with a “Virtual Lane Boundary” (VLB) instead of the real lane boundary that is used in the LDWS. The inattention grade moves the VLB according to the state of the driver and therefore reduces the number of redundant warnings. The ALDWS works satisfactory in real-time and during simulations it was able to suppress up to 70% of the redundant warnings while keeping the vast majority of the good warnings. (Author/publisher) This thesis is available at: http://www.s3.kth.se/signal/reports/exjobb/04/IR-SB-EX-0413.pdf
Abstract