Driver drowsiness and behavior detection in prolonged conditionally automated drives

Abstract

This paper reports a study that investigated driver behavior between manual and conditionally automated driving and behavioral progress in a long conditionally automated phase. The goal was to evaluate a novel framework of an assistant system for driver state monitoring during conditionally automated driving. The framework was based on the analysis of the drivers' eye closure and head movements provided by a driver observation camera. Furthermore, individual alertness requests to verify the take-over and reaction ability of the driver during the automated phase were included in the framework. The basis of the evaluation was a realistic driving simulator study with 18 participants and long monotonous drives (Mean: 2∶51 +/− 0∶18 h) with the majority of the drive being conditionally automated. The data showed a significant difference in the behavioral indicators (eye closure and head movement) between drivers driving manually or conditionally automated, independently of their current drowsiness state. These findings suggest that many of the common features used for drowsiness detection in manual driving phases are not applicable to the automated driving context without an adaption as presented within the framework provided.