Driver Cognitive Distraction Classification While Using Eco-driving Applications

Lin, Rui; Wang, Pei · 2024 · Crossref

DOI: 10.54941/ahfe1005226

archive: archived pipeline: cataloged

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the safety concerns associated with onboard eco-driving systems, which provide real-time speed guidance to reduce fuel consumption but may induce driver distraction. The research investigates the accuracy of detecting driver cognitive distraction while interacting with such a system in both acceleration and deceleration scenarios. The primary objective was to evaluate machine learning models for classifying cognitive attentional states, specifically examining the relative importance of driving performance features versus eye-glance features. The experiment involved 20 valid participants (aged 18–48) who completed a driving simulator study using a Force Dynamics 401cr simulator and a Pupil-Labs eye tracker. The experimental design included four drives: a baseline familiarization drive, a training drive, and two test drives counterbalanced across participants. In the test drives, participants followed eco-driving speed guidance derived from a velocity planning algorithm. One condition required participants to perform an N-back cognitive task simultaneously with driving (distracting condition), while the other required only following the speed guidance (attentive condition). Data were segmented using overlapping sliding windows, and features were extracted using the TSFRESH package. A random forest algorithm was trained on 16 participants’ data and tested on the remaining 4, with hyperparameters optimized via random search and 10-fold cross-validation. Results indicated that eye-glance features were the most effective predictors of cognitive distraction. In the acceleration scenario, the glance-based classifier achieved 90.8% accuracy, a performance level not significantly different from the combined feature set (glance + driving). In the deceleration scenario, the combined feature set yielded significantly higher accuracy than the glance-only set. Permutation importance analysis revealed that normalized gaze position (x and y axes) and fixation duration were among the most critical features for classification in both scenarios. While driving features such as steering control and acceleration contributed to the model, they were less effective than glance features when used in isolation. The findings suggest that eye-movement data is superior to driving performance metrics for detecting cognitive distraction in the context of eco-driving. This implies that future eco-driving systems should integrate driver monitoring capabilities that prioritize visual attention data to ensure information is presented only when drivers have sufficient cognitive resources. The study highlights the potential for machine learning to enhance the safety of connected vehicle technologies by preventing distraction-induced performance degradation, although the authors note limitations regarding sample size and the ecological validity of the simulated urban scenarios.

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 qwen3.8-27b-gittensor on 2026-08-23; verification: pending re-verification.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• distraction detection algorithms
• visual
• cognitive
• temporal
• external distraction
• gaze based attention detection

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: tool software
• Theoretical Contribution: conceptual framework, computational model