DriveIntel: An Intelligent Driver Behaviour Detection System for Road Safety

Abstract

Road crashes remain a persistent public-health challenge, with fatigue, alcohol consumption, and smoking-related distraction responsible for a significant share of preventable incidents. Many low-cost vehicles and older fleets in developing regions lack continuous driver monitoring and reliable escalation pathways. This paper presents DriveIntel, an intelligent, privacypreserving driver behaviour detection system that integrates lowcost embedded hardware and a modern edge-to-cloud software pipeline. The prototype uses an ESP32 Dev Kit V1 with an MQ3 gas sensor, vibration motor, LED indicator, and buzzer, coupled with a Firebase-backed web dashboard (Next.js + React) for live visibility. The firmware employs moving averages, adaptive thresholds, and hysteresis to minimize false positives, while the cloud layer ensures durable logging and contact notification. Across controlled trials, alcohol alerts averaged near one second from stimulus to actuation; drowsiness haptic nudges were similarly rapid; and event-to-UI visibility remained under five seconds. The contributions include a reproducible hardware–software blueprint, a methodology and development process suitable for student and field teams, and a comparative analysis of performance and deployment trade-offs relative to camera-first and single-modality systems.

Introduction

Road traffic injuries remain a major global problem, often caused by distracted, drowsy, or intoxicated drivers. Most vehicles—especially older and low-cost ones—lack built-in telematics or driver monitoring systems. Although high-end cars feature advanced ADAS and camera-based monitoring, these systems are expensive, require calibration, and raise privacy concerns. To address these barriers, the text proposes DriveIntel, an affordable, camera-free driver-monitoring solution designed for widespread, real-world adoption.

DriveIntel uses two simple cues:

1. Inactivity timers as a proxy for drowsiness

2. An MQ3 gas sensor to detect alcohol or smoke in the cabin

Processing happens locally on an ESP32, allowing alerts (vibration → buzzer/LED → cloud notification) even without network connectivity. A secure dashboard provides real-time visualization for drivers, guardians, or fleet operators. The document presents the design, firmware logic (moving average, hysteresis, exponential backoff), performance results, and placement of the system within existing research.

Literature Review Summary

1. Vision-Based Monitoring

Camera-based drowsiness detection using CNNs and temporal models can track eye closure, yawns, and head movement. These systems work well under controlled lighting but fail at night, with occlusions, or with uncooperative users. They also raise privacy concerns. DriveIntel avoids cameras entirely to reduce cost and protect privacy.

2. Gas/Alcohol Detection

Prior work using MQ-series sensors demonstrates reliable ethanol detection but faces issues like humidity interference, drift, and recovery delays. DriveIntel addresses these challenges using baseline calibration and hysteresis-based logic for stable readings.

3. IoT-Based Vehicle Monitoring

IoT systems enable remote alerts and cloud analytics, but many depend heavily on continuous connectivity. DriveIntel prioritizes edge-first processing, ensuring core safety functionality even during Wi-Fi loss, with automatic cloud syncing on reconnection.

4. Cost-Effective Embedded Systems

Research on embedded fatigue-detection systems shows that microcontrollers can support low-latency safety solutions, though many lack cloud integration. DriveIntel enhances these approaches by combining real-time edge detection with cloud persistence and visualization.

5. Behavioral and Time-Series Analytics

Time-series analysis helps detect long-term driver behavior patterns but usually requires large training datasets. DriveIntel instead uses simple rule-based temporal logic (thresholds + dwell times), enabling fast deployment without training data.

Overall Conclusion: Existing systems are accurate but usually too expensive, invasive, or connectivity-dependent for practical deployment in developing regions. DriveIntel fills this gap with a lightweight, low-cost, privacy-friendly, and extensible solution.

Problem Statement Summary

The goal is to create a single, camera-free driver-monitoring system that detects drowsiness and alcohol/smoke using inexpensive, non-intrusive sensors. It must:

• Alert the driver and emergency contacts

• Operate reliably despite humidity, airflow, or temperature changes

• Remain functional offline

• Sync events to the cloud when connectivity resumes

• Provide real-time monitoring with no dependency on third-party services

Methodology Summary

The DriveIntel pipeline includes:

• Sensing & Sampling: MQ3 readings at 10–20 Hz, smoothed using a moving average.

• Baseline Calibration: Establishes a clean-air reference to manage drift.

• Finite-State Machine (FSM): Three states—Normal, Caution, Alert—with thresholds, dwell times, and hysteresis.

• Actuation: Vibration (Caution), buzzer + LED (Alert).

• Cloud Sync: Buffered JSON events sent to Firebase with exponential backoff.

• Security: No audio/video collection, authenticated access, strict Firestore rules.

• Development Workflow: Component selection, assembly, firmware design, dashboard creation, testing under Wi-Fi loss, and iterative refinement.

Proposed Architecture Summary

DriveIntel consists of five layers:

1. Sensing Layer: MQ3 detects alcohol/smoke; moving-average window reduces noise; baseline handles environmental changes.

2. Edge Processing Layer: ESP32 runs the FSM, ensuring deterministic, low-latency decisions even without internet.

3. Actuation Layer: Vibration motor, buzzer, and LED provide escalating alerts via isolated power rails.

4. Cloud Layer: JSON events uploaded to Firebase; locally buffered when offline; fault-tolerant and suitable for forensic logs.

5. Dashboard Layer: Built using Next.js/React; shows real-time alerts, trends, driver profiles, and emergency alerts.

A hybrid offline-first design ensures safety in rural or highway areas with unstable networks. The system is modular and future-proof, allowing new sensors (IR eye-blink detector, gyroscope, etc.) via GPIO/I²C.

Hardware Summary

Hardware is designed for reliability, modularity, and low cost.
Key components:

• ESP32 Dev Kit (processing + Wi-Fi)

• MQ3 sensor (alcohol/smoke detection)

• Vibration motor (haptic warning)

• LED + buzzer (visual/auditory alerts)

• Relay + regulated power supply

The design ensures easy replication by students, researchers, and fleet operators.

Conclusion

DriveIntel, an Intelligent Driver Behavior Detection System, was introduced in this study. Its purpose is to detect and notify drivers of dangerous driving behaviors, particularly those involving alcohol use, smoking, and sleepiness. The main objective of the system was to develop an inexpensive, dependable, and private driver assistance tool that could be used even in older or less expensive cars without sophisticated safety features. The created prototype successfully combines cloud services, software, and hardware into a single architecture. The ESP32 Dev Kit V1 microcontroller effectively managed local event processing and sensor sampling, and the MQ3 gas sensor reliably detected the levels of smoke and alcohol. Quick driver response was ensured by the multi-modal alerting made possible by the vibration motor, LED indicator, and buzzer. Secure, real-time data logging and visualization without the hassle of complicated server maintenance was made possible by the use of Firebase Cloud Firestore and a Next.js dashboard. The system’s technical viability and resilience were validated through controlled laboratory testing. Alerts were triggered in less than two seconds for local actuation and less than five seconds for cloud synchronization, demonstrating the prototype’s low-latency event response. The system’s stability over extended operation was confirmed by its consistent uptime of over 97 Common privacy and financial constraints observed in many commercial driver monitoring systems are addressed by the system’s non-intrusive and camera-free design. Additionally, DriveIntel is appropriate for future integration into both private automobiles and commercial fleets due to its scalable architecture, modular hardware design, and straightforward firmware. The overall results of the experiments confirm that the suggested design satisfies its primary goals of affordability, responsiveness, and dependability, providing a strong basis for deployment in the future

References

[1] Y. Jebraeily, Y. Sharafi, and M. Teshnehlab, “Driver Drowsiness Detection Based on CNN Architecture Optimization Using Genetic Algorithm,” IEEE Access, 2024. [2] H. R. Ansari, Z. Kordrostami, and A. Mirzaei, “In-vehicle Wireless Driver Breath Alcohol Detection System Using a Microheater Integrated [3] Gas Sensor Based on Sn-doped CuO Nanostructures,” Scientific Reports, 2023. [4] R. Gamal, M. Al-Qutt, and S. Ghoniemy, “Driver Behavior Detection in Time Series: A Decade Review,” Int. J. Intelligent Computing and Information Sciences, 2023. [5] B. Adhikari, “Using Visual and Vehicular Sensors for Driver Behavior Analysis: A Survey,” George Mason University, 2023. [6] A. Banerjee, R. Saha, and T. Dutta, “Cloud-enabled Smart Vehicle Safety Systems: A Review,” IEEE Internet of Things Journal, 2023. [7] S. Gupta, A. Banerjee, and V. Sharma, “Internet of Vehicles: Architectures, Challenges, and Future Directions,” IEEE Access, 2022. [8] K. Lin, X. Zhang, and J. Wang, “Edge-Cloud Collaboration for RealTime Safety Analytics in Intelligent Transport,” IEEE Communications Surveys & Tutorials, 2021. [9] P. Singh, R. Kumar, and S. Gupta, “Low-cost Embedded Driver Fatigue Monitoring System,” IEEE Sensors Journal, 2020. [10] G. Rao and H. Singh, “Wireless Sensor Network Integration for Vehicle Safety Applications,” IEEE Vehicular Technology Magazine, 2020. [11] M. Patel and N. Joshi, “Practical Considerations for Low-Cost Gas Sensing in Automotive Cabins,” Sensors, 2022. [12] L. Zhao and Y. Chen, “Driver State Estimation using Wearable Sensors: A Survey,” IEEE Trans. on Intelligent Vehicles, 2019. [13] R. M. Thomas and D. S. B., “An Intelligent Application for Detecting and Alerting on Dangerous Driving Behaviors,” in Proc. ICCIDT, 2023. [14] J. Chandrasekharan et al., “An Intelligent Driver Monitoring System,” in Proc. IEEE C2I, 2021. [15] W. Li, “A Real-Time Driver Behavior Monitoring and Feedback Framework,” Master’s Thesis, Politecnico di Torino, 2024. [16] J. Park and H. Kim, “A Comparative Study of Alcohol Detection Techniques for In-vehicle Systems,” IEEE Access, 2020. [17] M. Rahman et al., “IoT-based Road Safety and Accident Prevention Frameworks: A Contemporary Review,” IEEE Access, 2024.

Copyright

Copyright © 2025 Ashish Trivedi, Vaishnavi Diyewar, Vinay Taksale, Kashish Deshmukh, Yash Wanjari, Surabhi Chandolkar, Sakshi Bhagat. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.