Model-based estimation of the state of vehicle automation as derived from the driver's spontaneous visual strategies
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 whether the state of vehicle automation (manual versus automated driving) can be estimated based on drivers’ spontaneous visual strategies. The research is motivated by the shift in driver roles during automated driving, where the removal of active steering and braking tasks alters visuo-motor coordination and gaze behavior. While previous studies have documented that automated driving leads to less gaze directed at the road center, this study reverses the approach: it aims to predict the automation state from gaze data to identify the most critical oculometric indicators. This methodology is intended to support driver monitoring systems in autonomous vehicles, particularly for assessing whether a driver remains engaged or has disengaged from the driving task. The experiment involved 12 participants who completed 18-minute highway drives in a fixed-base simulator under two conditions: manual driving (SAE Level 1) and conditional automation (SAE Level 3). Eye-tracking data were recorded and analyzed by dividing the visual field into 13 Areas of Interest (AOIs), including the road center, mirrors, dashboard, and peripheral screens. The researchers computed two types of indicators: static indicators (percentage of time spent gazing at each AOI) and dynamic indicators (transition probabilities between AOIs). Partial Least Squares (PLS) regression models were developed to predict the automation state using four input sets: road center percentage only, static indicators only, dynamic indicators only, and a combination of both. The results demonstrated that gaze dynamics played the most significant role in distinguishing between manual and automated driving. Models using only static indicators achieved zero classification errors but had higher prediction error rates compared to dynamic models. The model incorporating dynamic indicators alone yielded a mean square error of prediction (MSEP) of 0.04, while the combined static and dynamic model achieved the lowest MSEP of 0.02. Analysis of the regression coefficients revealed that manual driving was characterized by frequent transitions into the road center, reflecting active visuo-motor control. In contrast, automated driving was associated with gaze transitions away from the road center toward peripheral areas, the dashboard, and non-driving-related areas, indicating a dispersed visual strategy and partial disengagement. The study concludes that while static gaze metrics provide a baseline for discrimination, the dynamics of gaze transitions are essential for accurately estimating the automation state. These findings suggest that driver monitoring systems should prioritize gaze dynamics over simple fixation durations to detect changes in driver engagement. The proposed PLS regression approach offers a interpretable method for identifying key visual behaviors associated with automation, which could be adapted to monitor driver readiness for takeover requests or detect out-of-the-loop phenomena in other automated systems, such as aviation.
Key finding
Best classification of manual vs. automated driving state was achieved using both static AOI indicators and dynamic gaze transition indicators together. Gaze dynamics (AOI transitions) contributed more to classification quality than static indicators alone.
Methodology
simulator
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.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | — | — | — | 1 | 2026-05-07 |
| archive | success | canonical_url | — | — | 2 | 2026-06-03 |
| extract | success | cached | — | — | 2 | 2026-06-10 |
| clean | success | — | — | — | 1 | 2026-06-01 |
| chunk | success | — | — | — | 1 | 2026-06-01 |
| embed | success | — | — | — | 1 | 2026-06-02 |
| enrich | success | — | — | — | 1 | 2026-05-07 |
| promote | success | — | — | — | 2 | 2026-06-06 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 2 | 2026-06-10 |
| tag | success | vector_similarity | — | — | 15 | 2026-06-11 |
| verify | success | — | — | — | 2 | 2026-06-10 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.
Topics
Ranked by relevance to this paper. Hover a topic for its definition.
- situational awareness
- gaze based attention detection
- attention allocation
- eye movements scanning
- automation
- mode 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).
- Methodological Resource: tool software, measurement protocol
- Theoretical Contribution: conceptual framework