Driver Takeover Performance and Monitoring Behavior with Driving Automation at System-Limit versus System-Malfunction Failures

DeGuzman, Chelsea A.; Hopkins, Samantha A.; Donmez, Birsen · 2020 · Crossref

DOI: 10.1177/0361198120912228

archive: archived pipeline: cataloged

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study investigates how the type of driving automation failure—specifically system-limit versus system-malfunction failures—affects driver takeover performance and monitoring behavior in SAE Level 2 automated vehicles. While current Level 2 systems (e.g., Tesla Autopilot) require drivers to monitor the environment and resume control when the automation fails, prior research has largely treated these failure types interchangeably or studied them in isolation. The authors argue that drivers can predict system-limit failures (e.g., poor lane markings) based on environmental cues, whereas system-malfunction failures (e.g., algorithmic errors) are unforeseeable. This distinction is critical because it implies drivers should exhibit different preparatory behaviors and reaction times depending on the failure type. The study aims to quantify these differences to inform the literature and improve driver support systems. The research employed a driving simulator study with 18 participants (ages 25–30) who experienced both failure types in a counterbalanced within-subjects design. Participants engaged in a visual-manual secondary task on a tablet while driving a simulated rural highway with adaptive cruise control and lane keeping engaged. System-limit failures involved the vehicle deviating into a right-turn lane at an intersection, a scenario participants were trained to anticipate. System-malfunction failures involved the vehicle veering right without prior environmental indicators. Data collected included hands-on-wheel time, takeover time, steering wheel metrics, and eye-tracking data (glance rate, duration, and percent of time) for the roadway and secondary task. Mixed linear models were used to analyze the data, controlling for participant variability. Results indicated that drivers responded significantly faster to system-limit failures. Specifically, participants placed their hands on the wheel 0.62 seconds sooner and disengaged the lane-keeping system 0.51 seconds sooner for system-limit failures compared to system-malfunction failures. At the study’s speed limit of 50 mph, this difference equates to an additional 37.4 feet of reaction distance. Monitoring behavior also differed: before system-limit failures, participants spent 12.7% less time looking at the secondary task and 11.2% more time looking at the roadway. Eye-tracking analysis revealed that this increased roadway attention was driven by longer average glance durations rather than a higher frequency of glances, suggesting drivers sustained their focus on the road when they could predict the failure. No significant differences were found in steering wheel angle variability or range. The findings demonstrate that failure type significantly influences driver behavior, challenging previous studies that did not distinguish between predictable and unpredictable failures. The results suggest that when drivers must independently decide when to take over, the ability to predict a system-limit failure leads to earlier and more sustained monitoring. Consequently, the authors conclude that the research community should explicitly differentiate between system-limit and system-malfunction failures in experimental designs and literature reviews. Furthermore, as Level 2 vehicles become more prevalent, efforts should focus on enhancing drivers’ mental models of automation limits to improve their ability to predict and prepare for system-limit failures, thereby improving safety outcomes.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.8-27b-gittensor on 2026-08-23; verification: pending re-verification.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• automation surprise
• takeover transitions
• automation
• automation complacency bias
• mode awareness
• situational awareness

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: behavioral performance data
• Methodological Resource: measurement protocol
• Theoretical Contribution: conceptual framework