A drowsy driver detection system for heavy vehicles

Abstract

Driver drowsiness/fatigue is an important cause of combination-unit truck crashes. Drowsy driver detection methods can form the basis of a system to potentially reduce accidents related to drowsy driving. We report on efforts performed at the Carnegie Mellon Driving Research Center to develop such in vehicle driver monitoring systems. Commercial motor vehicle truck drivers were studied in actual fleet operations. The drivers operated vehicles that were equipped to measure vehicle performance and driver psychophysiological data. Based on this work, two drowsiness detection methods are being considered. The first is a video-based system that measures PERCLOS, a scientifically supported measure of drowsiness associated with slow eye closure. The second detection method is based on a model to estimate PERCLOS based on vehicle performance data. A non-parametric (neural network) model was used to estimate PERCLOS using measures associated with lane keeping, steering wheel movements and lateral acceleration of the vehicle.