Towards Context-Aware Modeling of Situation Awareness in Conditionally Automated Driving

Avetisyan, Lilit; Yang, X. Jessie; Zhou, Feng · 2024 · arXiv

archive: archived pipeline: cataloged verified

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

Summary

This study addresses the critical safety challenge of maintaining driver Situation Awareness (SA) in conditionally automated vehicles (SAE Level 3), where drivers must be prepared to retake control during takeover requests (TORs). Traditional SA measurement methods are often intrusive or rely on self-reports that fail to capture dynamic fluctuations. To bridge this gap, the authors developed a context-aware, real-time predictive model for SA using multimodal data, aiming to enable unobtrusive monitoring that accounts for individual differences and varying driving conditions. The researchers conducted a driving simulator experiment with 67 participants (44 retained after data cleaning) who experienced automated driving scenarios involving TORs. The experimental design was a 2x2 mixed design manipulating risk perception (high vs. low) and automation reliability (error vs. no error). Data collection included physiological signals (Galvanic Skin Response, Heart Rate, Heart Rate Variability), eye-tracking metrics (fixations, dispersion, duration across road and task areas), and demographic information. Self-reported SA ratings served as the ground truth. The study employed a Light Gradient Boosting Machine (LightGBM) for regression and SHapley Additive exPlanations (SHAP) to interpret feature contributions and identify the most significant predictors. The LightGBM model achieved promising performance in predicting SA, with a Root Mean Square Error (RMSE) of 0.89, Mean Absolute Error (MAE) of 0.71, and a correlation coefficient of 0.78 when trained on the top 12 features identified by SHAP. Key predictors included demographic factors (age, gender, AV knowledge) and physiological/behavioral markers. Specifically, higher Galvanic Skin Response and Heart Rate were positively correlated with SA, suggesting links to arousal and alertness. Eye-tracking analysis revealed that increased fixations on the secondary task (game) correlated with higher SA, while greater dispersion on the game screen correlated with lower SA. Additionally, the study found that automation errors significantly impacted physiological responses, such as increasing heart rate and altering fixation patterns, and that high-risk conditions elicited higher self-reported SA compared to low-risk conditions. The findings demonstrate the feasibility of using machine learning to integrate multimodal sensor data for real-time SA assessment in automated driving. By identifying specific physiological and behavioral markers that correlate with SA, this work provides a foundation for developing context-aware monitoring systems. These systems could enhance driver-AV interactions by providing timely interventions, thereby improving safety and trust in conditionally automated vehicles. The study highlights the importance of considering individual characteristics and environmental context in designing effective SA prediction models.

Key finding

A LightGBM model trained on multimodal physiological, eye-tracking, and demographic predictors estimated continuous driver SA in conditionally automated driving with RMSE = 0.89, MAE = 0.71, and r = 0.78 against self-report. Risk perception and automation errors significantly elevated reported SA, with eye-tracking showing more road-center and traffic-checking gaze under risk and error conditions.

Methodology

simulator

Sample size: 67 enrolled (30 F mean age 28.3 SD 11.5; 37 M mean age 25.9 SD 12.3); 44 analyzed after sensor/simulator exclusions

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. Discovered via discover_arxiv on 2026-05-04 (3 acquisition events logged).

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified_with_issues.

Topics

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

• situational awareness
• mode awareness
• situation awareness theory
• automation surprise
• automation
• anticipation

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
• Theoretical Contribution: theory or model, computational model