Implementation of the Drowsy Driving Prevention System using AI-Based Conversations to Maintain Driver Arousal

Noriki Uchida; Tomoyuki Ishida; Yoshitaka Shibata · 2024 · Crossref

DOI: 10.64799/rebicte.v10.4

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 persistent safety issue of drowsy driving, which accounts for approximately 33% of fatal accidents on Japanese highways. The authors argue that existing Intelligent Transport Systems (ITS) primarily rely on sensors to detect drowsiness after it has already begun, often resulting in delayed prevention due to high vehicle speeds. Furthermore, sensor-based methods require inconvenient or distracting wearable devices. To overcome these limitations, the study proposes a proactive drowsy driving prevention system that maintains driver arousal through AI-based conversations via a smart speaker, eliminating the need for physical sensors. The proposed system utilizes a smart speaker (Google Home Mini) connected to cloud-based AI modules on the Google Cloud Platform. Before driving, the system assesses the driver’s condition by asking about sleep duration, alcohol consumption, and work hours, categorizing the risk into four levels. Based on this assessment, the system determines the interval for subsequent interactions, ranging from every 10 minutes for high-risk drivers to every 60 minutes for low-risk drivers. The study implements two types of conversational interventions: "lightweight" conversations involving simple, selectable questions (e.g., preference between dogs or cats), and "heavyweight" conversations involving natural dialogue generated by the Google Small Talk AI agent. A prototype system was built using a driving simulator (PlayStation 4 with *Project Cars*), a NeuroSky MindWave Mobile2 electroencephalograph (EEG) to monitor brainwaves, and cloud services including DialogFlow and Firebase. Experiments were conducted with 12 participants who drove on a simulator course at 60 km/h under three conditions: lightweight conversation, heavyweight conversation, and no conversation. Brainwave analysis measured arousal levels, where beta and gamma waves indicate arousal, while alpha, theta, and delta waves indicate drowsiness. Results showed that both conversation types significantly increased beta and gamma wave activity (average ~5.0) and decreased drowsy waves (average ~3.0) compared to the no-conversation condition, where drowsy waves increased and arousal waves decreased. Survey results confirmed that participants felt less sleepy during both conversation conditions compared to the control. However, the heavyweight natural conversations induced higher levels of irritation, difficulty, and reduced concentration compared to the lightweight selectable questions. The study concludes that AI-based conversations via smart speakers are effective in maintaining driver arousal and preventing drowsiness. While both methods work, lightweight selectable questions are less likely to cause annoyance or distraction than natural conversations. The authors suggest that future work should focus on improving the accuracy of natural language processing to reduce driver irritation and expanding the participant pool for further evaluation.

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-nvidia on 2026-08-10; verification: verified.

Topics

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

• drowsiness detection algorithms
• generative ai voice assistants
• drowsiness

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
• Theoretical Contribution: computational model