Predicting Driver Takeover Time in Conditionally Automated Driving
DOI: 10.1109/tits.2022.3154329
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 predicting driver takeover time in conditionally automated driving (SAE Level 3). While previous research has identified individual factors influencing takeover performance—such as takeover lead time, non-driving tasks, and request modality—there is a lack of computational models that integrate these variables simultaneously to estimate exact takeover durations. Accurate prediction is essential for designing adaptive in-vehicle alert systems that provide timely situation awareness. To bridge this gap, the authors developed an explainable machine learning model using eXtreme Gradient Boosting (XGBoost) and SHapley Additive exPlanations (SHAP). The methodology utilized a dataset derived from a meta-analysis of 129 prior studies, comprising 519 takeover events. The dataset included 18 predictor variables categorized into driver characteristics (e.g., age, cognitive load), automated vehicle systems (e.g., automation level, simulator fidelity), takeover request (TOR) modalities, non-driving tasks (NDTs), and scenario urgency. The XGBoost regressor was trained using a 10-fold cross-validation strategy repeated 100 times to ensure stability. To address the "black-box" nature of XGBoost, SHAP was employed to provide both global explanations (variable importance and interaction effects) and local explanations (individual instance predictions). Feature selection was performed by iteratively adding variables based on importance ranking until the Root Mean Square Error (RMSE) stabilized. The results demonstrated that the XGBoost model achieved superior predictive performance compared to previous linear and non-linear regression models. The optimal model utilized seven critical predictors: urgency, time budget to collision/boundaries, driver age, hand occupation, visual TOR presence, simulator fidelity, and interaction with other road users. This configuration yielded an RMSE of 0.806 seconds, a Mean Absolute Error (MAE) of 0.505 seconds, an Adjusted R² of 0.573, and a correlation coefficient of 0.883. SHAP analysis revealed specific relationships: higher urgency and shorter time budgets significantly reduced takeover time, while the presence of visual TORs and handheld devices increased it. Additionally, driver age showed a non-linear effect, with takeover times increasing until age 45, after which they began to decrease. The significance of this work lies in its provision of a robust, explainable framework for predicting takeover time. By identifying the most influential variables and their interactions, the study offers actionable insights for the design of personalized, adaptive alert systems in automated vehicles. The integration of XGBoost for high-accuracy prediction and SHAP for interpretability ensures that the model is not only effective but also trustworthy for human-automation interaction design, facilitating safer transitions from automated to manual control.
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 | semantic_scholar | — | — | 6 | 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 |
| enrich | success | semantic_scholar | — | — | 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 | — | — | 10 | 2026-08-11 |
| verify | success | — | — | — | 2 | 2026-08-10 |
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.
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
- Theoretical Contribution: computational model, conceptual framework