Predicting Takeover Quality in Conditionally Autonomous Vehicles based on Takeover Request Modalities, Driver Physiological State and the Environment
DOI: 10.54941/ahfe100989
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 critical safety challenge of takeover requests (TORs) in conditionally autonomous vehicles, where the system must transfer control back to the driver. Current systems often issue TORs without considering the driver’s state or environmental context, potentially leading to poor takeover quality. The research aims to determine if machine learning models can predict takeover quality—defined as an aggregation of reaction time and maximum steering wheel angle—by analyzing the interaction between driver physiological signals, environmental conditions, and TOR modalities. The methodology involved a driving simulator experiment with 15 participants. Each participant completed a 50-minute rural driving session containing nine takeover scenarios triggered by fixed obstacles with a seven-second time-to-collision. Physiological data, including Electrodermal Activity (EDA), Electrocardiogram (ECG), and respiration, were recorded for the 90 seconds preceding each TOR. The study varied three factors: TOR modalities (visual-haptic, visual-auditory, and visual-auditory-haptic), non-driving-related tasks (visual 2-back, auditory 2-back, or monitoring), and weather conditions (sunny vs. adverse weather with rain and low luminosity). After processing physiological features using the Neurokit library and applying feature selection, 80 takeover instances were used for model training. Data augmentation via random noise was employed to mitigate overfitting. Four machine learning algorithms—K-Neighbors, Support Vector Regressor, Random Forest, and Neural Network—were trained and evaluated using Mean Squared Error (MSE) and Mean Absolute Error (MAE). The results indicate that predicting takeover quality is feasible. The Random Forest model achieved the best performance, with an MSE of 0.0261 and an MAE of 0.1274. This represented a 56.5% improvement in MSE and a 38.64% improvement in MAE compared to a baseline model that predicted the mean takeover quality. The feature selection process retained 20 physiological features, including specific EDA, ECG, heart rate variability, and respiration metrics, alongside weather and TOR modality variables. Although the abstract mentions a Neural Network as the best model with an MSE of 0.0538, the detailed results section identifies the Random Forest as superior with an MSE of 0.0261. The significance of this work lies in demonstrating that takeover quality can be predicted in real-time using a short 90-second window of physiological and contextual data. This capability allows for the development of adaptive human-machine interfaces that can select optimal TOR modalities based on predicted driver performance, thereby enhancing safety in conditionally autonomous driving. The authors suggest that future work should implement and evaluate real-time versions of this model to assess the practical implications of dynamic TOR selection.
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 | cached | — | — | 3 | 2026-08-10 |
| clean | success | clean | — | — | 1 | 2026-08-09 |
| chunk | success | chunk | — | — | 1 | 2026-08-09 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-08-09 |
| promote | success | — | — | — | 1 | 2026-08-09 |
| summarize | success | llm | qwen3.6-27b-nvidia | summ-v5 | 2 | 2026-08-10 |
| tag | success | vector_similarity | — | — | 11 | 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