Machine learning agent to recommend the best modality for takeover during conditionally automated driving
DOI: 10.54941/ahfe100845
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 addresses the safety challenges associated with takeover requests (TOR) in conditionally automated vehicles, specifically focusing on optimizing the modality used to alert drivers. In Level 3 automation, drivers may be disengaged from the driving loop, making the transition of control critical. The authors propose a machine learning-based smart agent that recommends the most effective TOR modality combination—haptic-visual, auditory-visual, or haptic-auditory-visual—based on the driver’s physiological state and external weather conditions. The goal is to predict and maximize takeover quality, defined by reaction time and maximum steering wheel angle, thereby reducing accident risks during handover. The methodology involved a driving simulator study with 15 participants. Each driver completed a 50-minute session involving nine takeover scenarios triggered by fixed obstacles. Drivers performed one of three tasks: a visual 2-back task, an auditory 2-back task, or a monitoring task, under either sunny or adverse weather conditions. Physiological signals, including electrodermal activity (EDA), electrocardiogram (ECG), and respiration, were recorded for the 90 seconds preceding each TOR. The researchers processed 175 raw physiological features using Neurokit, applying redundancy filtering, Pearson correlation analysis, and Mutual Information analysis to reduce the dataset to 22 relevant features. After outlier removal, 80 valid TOR instances remained for modeling. The team trained and compared four regression models—KNeighbors, Support Vector, Random Forest, and Neural Networks—using grid search and cross-validation. Data augmentation techniques, including SMOGN and random noise, were also tested to enhance the training set. The results demonstrated that the Random Forest Regressor outperformed the other models and the baseline (constant mean prediction). The Random Forest achieved a Mean Squared Error (MSE) of 0.0261, improving upon the baseline MSE of 0.0600 by 56.5%. The agent successfully varied its recommendations based on situational inputs, with a distribution of 62.5% for auditory-visual, 25% for auditory-haptic-visual, and 12.5% for haptic-visual modalities. The study introduced a novel metric for modality impact, finding a mean difference of 4.95% (standard deviation: 2.7%) in predicted takeover quality between the best and worst modality choices. While data augmentation provided a slight performance boost, its significance was deemed debatable. The significance of this work lies in the development of a dynamic Human-Machine Interface (HMI) capable of adapting TOR modalities in real-time. By leveraging driver physiology and environmental context, the agent offers a more personalized and potentially safer approach to automated driving handovers compared to static TOR designs. The authors conclude that this smart HMI could improve takeover quality and safety, though further research is needed to evaluate its impact on user experience.
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 | Crossref | — | — | 1 | 2026-08-09 |
| archive | success | canonical_url | — | — | 1 | 2026-08-09 |
| extract | success | pdftotext | — | — | 4 | 2026-08-10 |
| clean | success | clean | — | — | 2 | 2026-08-10 |
| chunk | success | chunk | — | — | 2 | 2026-08-10 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 2 | 2026-08-10 |
| promote | success | — | — | — | 1 | 2026-08-09 |
| summarize | success | llm | qwen3.6-27b-nvidia | summ-v5 | 2 | 2026-08-10 |
| tag | success | vector_similarity | — | — | 17 | 2026-08-11 |
| verify | partial | — | — | — | 2 | 2026-08-10 |
Summary generated by qwen3.6-27b-nvidia on 2026-08-10; verification: verified_with_issues.
Topics
Ranked by relevance to this paper. Hover a topic for its definition.
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: measurement protocol
- Theoretical Contribution: computational model, conceptual framework