Driver Takeover Performance and Monitoring Behavior with Driving Automation at System-Limit versus System-Malfunction Failures. Issue 4 (April 2020)

• Record Type: Journal Article
• Authors: DeGuzman, Chelsea A.; Hopkins, Samantha A.; Donmez, Birsen
• Is Part Of: Transportation research record. Volume 2674:Issue 4(2020)
• Page Start: 140
• Page End: 151
• Publication Date: 2020-04
• Subjects: Transportation -- Periodicals; Roads; Transport -- Périodiques; Routes -- Périodiques; Routes -- Conception et construction -- Périodiques; Roads; Transportation; 388.05
• DOI: 10.1177/0361198120912228
• Languages: English
• ISSNs: 0361-1981

Abstract

Today's vehicles are becoming highly automated, however, if the automation fails, drivers must take over control of the vehicle. Automation may fail as a result of known system limits (system-limit failure) or of malfunctions that are unforeseen by system designers (system-malfunction failure). The aim of this research was to quantify the differences between how these two failure types influence driver takeover performance and monitoring behaviors. In a simulator with SAE Level 2 driving automation, 18 drivers experienced both a system-limit and system-malfunction failure while engaging in a secondary task. Results show that drivers put their hands on the wheel 0.62 s sooner and took over 0.51 s faster for the system-limit failure compared with the system-malfunction failure. Eye tracking data revealed that the percent of time looking at the secondary task display was 12.7% lower and the percent of time looking at the roadway was 11.2% higher before the system-limit failure compared with before the system-malfunction failure. Given that takeover performance and monitoring behavior differ significantly based on failure type, a distinction should be made in the literature between system-limit and system-malfunction failures, and comparisons between previous studies using these failures should not be done without considering this distinction. Furthermore, as SAE Level 2 vehicles are currently available to consumers, efforts should be focused on supporting drivers' mental models of automated systems, so that drivers are able to successfully predict system-limit failures.