Neurophysiological Mechanisms and Mitigation Strategies of Driver Fatigue in Human-Artificial Intelligence Cooperative Driving Using EEG

Lu, Huoping; Tang, Bangbei; Li, Yan; Wu, Yingzhang; Tang, Qiuyang · 2026 · Crossref

DOI: 10.21203/rs.3.rs-9777533/v1

archive: archived pipeline: cataloged

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

Summary

This study investigates the distinct neurophysiological mechanisms of driver fatigue in manual versus autonomous driving modes within a human–AI cooperative framework, addressing a critical safety gap where existing research lacks comparative analysis of these two states. The motivation stems from the differing nature of fatigue: manual driving induces active fatigue from sustained operational demands, while autonomous driving induces passive fatigue due to monotony and reduced cognitive load. The research aims to elucidate these differences using electroencephalography (EEG) and to evaluate multimodal mitigation strategies. The experimental design involved 15 participants (10 male, 5 female, aged 20–25) who underwent simulated driving tests using the Forza Horizon 5 game and a PXN V99 simulator. EEG data were collected via a 31-channel cap following the international 10–10 system. Fatigue was induced through sleep restriction and a three-hour monotonous driving task, with fatigue levels monitored via the Stanford Sleepiness Scale (SSS). Participants experienced seven different alertness interventions (olfactory, auditory, visual, and combinations thereof) in both manual and autonomous modes. Data preprocessing included artifact removal and bandpass filtering into delta, theta, alpha, beta, and gamma bands. Feature extraction focused on time-domain, frequency-domain, and spatial-domain features, specifically using Common Spatial Patterns (CSP) for spatial analysis. Feature selection was performed using correlation analysis (p < 0.05), and fatigue classification was modeled using Support Vector Machines (SVM). Key findings reveal that fatigue in both modes primarily involves the frontal and occipital lobes, but the central region is activated only during autonomous driving. Neurophysiologically, manual driving fatigue is characterized by synchronization between theta and alpha waves, whereas autonomous driving fatigue manifests as abnormalities in theta and beta waves, suggesting a state of "ineffective focus" or "rigid readiness." Spatial features demonstrated superior discriminatory power compared to time or frequency domains. Classification accuracy reached 82.14% for manual driving and 80.36% for autonomous driving after feature selection, representing improvements of 8.81% and 7.15%, respectively. Notably, models using only spatial features outperformed hybrid models, with accuracy improvements of 7.52% (manual) and 8.93% (autonomous). Regarding mitigation, the combined olfactory-auditory-visual stimulus was the most effective intervention, reducing subjective fatigue scores by 2.93 points in manual mode and 3.27 points in autonomous mode, outperforming single or dual-modal stimuli. The significance of this work lies in providing a theoretical foundation for context-aware, real-time fatigue monitoring systems that distinguish between driving modes. The identification of CSP as a robust biomarker and the demonstration that spatial features alone can outperform hybrid feature sets offer practical insights for developing efficient, high-accuracy driver monitoring systems. Furthermore, the validation of multimodal sensory stimulation as the optimal mitigation strategy provides actionable guidelines for enhancing driver alertness in intelligent transportation systems, thereby improving safety in human-machine cooperative driving 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.

• drowsiness detection algorithms
• drowsiness
• truck driver fatigue

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: physiological data
• Methodological Resource: tool software, validation psychometrics