Takeover performance according to the level of disengagement during automated driving

Gallouin, Evan; Wang, Xuguang; Beillas, Philippe; Bellet, Thierry · 2022 · Crossref

DOI: 10.17077/dhm.31754

archive: archived pipeline: cataloged verified

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

Summary

This study investigates how a driver’s level of engagement or disengagement during automated driving affects their performance when resuming manual control. As automation levels increase from Level 2 (partial automation, requiring monitoring) to Level 3 (conditional automation, allowing non-driving related tasks), drivers may become less vigilant. The research addresses two primary questions: whether a TakeOver Request (TOR) improves performance for drivers monitoring the road, and whether the state of vigilance (hyper- vs. hypo-vigilant) during non-driving tasks impacts takeover safety. The experiment utilized a static driving simulator with forty participants (aged 20–43) who completed sixteen critical takeover scenarios. Each scenario involved a highway situation ending in a critical event requiring manual intervention within a short time budget (mean 4.2 seconds). Four conditions were tested: (C1) monitoring without a TOR, (C2) monitoring with a TOR, (C3) performing a cognitively demanding secondary task (hyper-vigilant) with a TOR, and (C4) resting with eyes closed (hypo-vigilant) with a TOR. Performance was measured by TakeOver Time (TOT), defined as the interval from the TOR to the first control input, and by the number of collisions. Results indicated that monitoring with a TOR (C2) yielded the shortest reaction times (mean 1.01 s) and the fewest collisions (24). In contrast, monitoring without a TOR (C1) resulted in the highest number of collisions (48) and significantly longer reaction times during collision events, suggesting that continuous supervision without alerts leads to neglect of the monitoring task. Regarding disengaged states, drivers in the relaxed, eyes-closed position (C4) took significantly longer to react than in other conditions and experienced 40 collisions. Drivers performing the cognitively demanding task (C3) had 27 collisions. While there was no significant difference in reaction times between the hyper- and hypo-vigilant states, the hypo-vigilant state resulted in more accidents, attributed to the physical distance from controls in a reclined posture. The findings conclude that TORs are critical for maintaining safety in partial automation, as drivers often fail to effectively monitor the system without alerts. Furthermore, disengagement from driving, particularly in a relaxed posture, degrades takeover performance. The study highlights the limitations of relying on human supervision in Level 2 automation and suggests that postural and vigilance states significantly influence the safety of Level 3 takeovers. These results inform the design of automation systems, emphasizing the need for effective warning systems and consideration of driver posture in safety protocols.

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 qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

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

• takeover transitions
• automation
• automation complacency bias
• manual
• automation surprise
• 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