Predicting Lane Keeping Behavior of Visually Distracted Drivers Using Inverse Suboptimal Control
archive: archived pipeline: cataloged verified
Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)
Summary
This paper addresses the challenge of predicting lane-keeping behavior in visually distracted drivers to enable situation-dependent risk assessment for driver assistance systems. Current approaches often detect distraction based solely on gaze direction or environmental sensors, neglecting the interplay between driving conditions, driver impairment, and compensation strategies. The authors propose an inverse suboptimal control framework that models distracted driving as a partially observable Markov decision process (POMDP). This approach allows for the prediction of both the driver’s future gaze state and vehicle trajectory, enabling a more accurate assessment of crash risk than binary distraction detection. The method models distracted lane keeping by combining a linear kinematic vehicle dynamics model with a reward function that penalizes lateral deviation, steering effort, and gaze-switching costs. Crucially, it incorporates an observation model where the driver’s perception of vehicle states degrades when their gaze is off-road. To handle the stochastic and suboptimal nature of human behavior, the authors employ Maximum Causal Entropy (MCE) inverse optimal control. This technique infers the driver’s underlying reward preferences from observed data, allowing the model to predict behavior in new situations by computing the optimal policy for those inferred preferences. The approach was evaluated using real-world traffic data collected from seven drivers on a public highway. Drivers performed a visually distracting secondary task (reading and typing random numbers) at four fixed speeds (80–110 km/h) controlled by Adaptive Cruise Control. Data included lane position, curvature, steering angle, and eye-tracking metrics. The proposed MCE-IOC method was compared against a baseline combining Salvucci’s two-point steering model and Johnson’s gaze-allocation model. Evaluation metrics included the expected squared error of lane position and the Kullback-Leibler divergence of off-road gaze duration. Results demonstrate that the MCE-IOC approach achieves significantly lower prediction errors than the baseline across all conditions. Specifically, the proposed method showed superior generalization to driving speeds unseen during the training phase, confirming that inferred driver preferences are more transferable than directly estimated policies. The authors conclude that this framework provides a robust foundation for situation-specific visual distraction assessment, allowing assistance systems to predict the probability of critical incidents like lane departure. Future work aims to extend the model to include longitudinal control and peripheral vision effects.
Key finding
Inverse suboptimal control yields lower lane-keeping prediction error and better generalization across speeds than baseline models, supporting situation-aware distraction risk assessment.
Methodology
other
Sample size: 7
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 (6 acquisition events logged).
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | author_sweep | — | — | 4 | 2026-05-28 |
| archive | success | — | — | — | 1 | 2026-05-04 |
| 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 | 2026-05-04 |
| promote | success | — | — | — | 1 | 2026-05-04 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 2 | 2026-06-10 |
| tag | success | vector_similarity | — | — | 18 | 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.
- distraction detection algorithms
- visual
- gaze based attention detection
- visual manual
- manual
- lane positioning
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, theory or model