Hazard perception in manual and hands-off Level 2 driving under daytime and after dark lighting conditions: A driving simulator study

Lee, Yee Thung; Gonçalves, Rafael C.; Öztürk, İbrahim; Solernou, Albert; Proctor, Karl; Merat, Natasha · 2026 · Crossref

DOI: 10.1007/s10111-026-00870-9

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 lighting conditions (daytime vs. after dark) and driving automation levels (manual vs. hands-off SAE Level 2) influence drivers’ hazard perception (HP) in urban environments. Motivated by higher crash rates at night and the emerging prevalence of hands-off automation, the research addresses a gap in understanding how these factors interact to affect hazard detection and reaction. The authors hypothesized that HP would be worse after dark and that hands-off automation might improve detection due to reduced operational demands, though manual driving might yield faster reactions. The experiment utilized a high-fidelity driving simulator with 48 participants in a 2x2 within-subjects design. Participants completed four drives, encountering six hazardous events (three materialized, three non-materialized) involving pedestrians and oncoming cars. Lighting was manipulated via global illumination settings for daytime (12:00 PM) and after dark (12:00 AM with streetlights). In manual conditions, drivers controlled the vehicle; in Level 2 conditions, an Advanced Driver Assistance System (ADAS) managed lateral and longitudinal control, allowing hands-off operation. Data collection included eye-tracking metrics (time to first fixation) and vehicle metrics (time to first relevant manoeuvre, such as braking or steering). Statistical analysis employed Generalized Linear Mixed Models (GLMM) to assess the effects of lighting, automation, and their interaction on HP performance. Results indicated that lighting significantly affected hazard detection but not reaction time. Drivers detected hazards significantly earlier in daytime conditions compared to after dark, particularly for pedestrians approaching from the left. However, there was no significant difference in response times between lighting conditions. Regarding automation, drivers exhibited proactive responses during manual driving, reacting before potential hazards materialized. In contrast, responses during hands-off Level 2 automation were more reactive, occurring after hazards had developed. The study found no significant interaction between lighting and automation levels for the primary HP metrics, suggesting the effects of lighting and automation operated independently in this context. The findings highlight that while hands-off automation may alter the timing and nature of driver responses—shifting them from proactive to reactive—lighting conditions remain a critical factor in early hazard detection. The lack of difference in reaction times between lighting conditions suggests that once a hazard is detected, drivers respond similarly regardless of visibility, but the delay in detection at night poses a risk. These results imply that systems designed to support hazard perception must account for both environmental context, such as low-light conditions, and the specific demands of automated driving modes. The study underscores the need for tailored protocols to ensure timely driver response, particularly in after-dark scenarios where detection delays are most pronounced.

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-nvidia on 2026-08-10; verification: verified.

Topics

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

• hazard perception
• situational awareness
• dark adaptation mesopic
• peripheral attention
• looked but failed to see
• automation surprise

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: tool software
• Theoretical Contribution: computational model