Towards a Context-Dependent Multi-Buffer Driver Distraction Detection Algorithm

Ahlstrom, Christer; Georgoulas, George; Kircher, Katja · 2022 · Crossref

DOI: 10.1109/tits.2021.3060168

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 introduces AttenD2.0, a context-dependent algorithm for detecting driver distraction that extends the original AttenD algorithm by integrating elements from the Minimum Required Attention (MiRA) theory. The research addresses the limitations of existing gaze-based distraction detection systems, which often rely on fixed thresholds for looking-away time and fail to account for situational complexity. Current methods frequently generate false positives by treating all glances away from the forward roadway as equally distracting, regardless of whether the glance is necessary (e.g., checking mirrors) or irrelevant. AttenD2.0 aims to provide a more human-centered and situation-aware assessment of driver attention by incorporating both static environmental requirements and dynamic driving conditions. The algorithm operates using a multi-buffer system where each relevant glance target (e.g., forward roadway, mirrors, side roads) is linked to its own attention buffer. These buffers deplete when the driver looks away and refill when looking back; if any buffer empties, the driver is classified as distracted. AttenD2.0 enhances this mechanism by making buffer dynamics context-dependent. Static requirements, such as the need to check side roads at intersections, are handled through intermittent buffers activated by digital map data. Dynamic requirements, such as driving speed relative to the speed limit, are managed via a weight function that adjusts buffer increment and decrement rates, allowing drivers to "buy time" by slowing down. Additionally, the algorithm incorporates neural delay latencies and exponential buffer updates to better reflect information density and cognitive processing times. The authors demonstrate the scalability and functionality of AttenD2.0 through a driving simulator experiment involving 16 bus drivers. The study focused on an automated bus stop docking and depot procedure, using AttenD2.0 to ensure drivers were attentive before taking back control from the automation. While the experiment was primarily designed to evaluate the docking functionality rather than the algorithm itself, it served as a practical showcase for real-time implementation. The results illustrated how AttenD2.0 correctly identifies inattention in complex scenarios, such as failing to check a side road within a required zone or engaging in non-driving-related activities that deplete the forward attention buffer, while appropriately acknowledging necessary glances to mirrors or speedometers. The significance of this work lies in its move toward more sophisticated, theory-driven attention monitoring systems. By integrating environmental sensing and digital maps, AttenD2.0 offers a scalable solution adaptable to different levels of vehicle automation. The paper concludes that while the current implementation relies on foveal vision and excludes peripheral information, the framework provides a robust foundation for future real-world applications. Future work aims to automatically integrate situational information from vehicle sensors and digital maps to further refine the context-dependent detection capabilities, ultimately improving the accuracy and reliability of driver state estimation in intelligent transportation systems.

Key finding

The AttenD2.0 algorithm successfully integrates context-dependent multiple buffers to detect driver distraction by dynamically adjusting attention requirements based on situational complexity and driving behavior.

Methodology

simulator

Sample size: 16

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.

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.

• gaze based attention detection
• distraction detection algorithms
• visual
• temporal
• attention allocation
• situational awareness

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: theory or model, conceptual framework