Driver Monitoring Systems (DMS) are becoming decision-enabling components within modern vehicle software stacks. For automotive OEMs, the challenge is improving system response time, operational reliability, and integration depth across ADAS ecosystems within the DMS architecture.
Our Mobility CoE (Centre of Excellence) experts developed a real-time, AI-powered DMS to address these requirements. The solution delivers low-latency behavioural assessment, scientifically grounded fatigue scoring, and direct coordination with ADAS modules, enabling vehicles to respond faster and more intelligently to driver state.
Business Goals
Our objective was to facilitate automotive OEM partners with a reliable, accurate, and production-ready DMS that:
- Quantifies drowsiness using proven scientific models
- Offers proactive, adaptive warnings before risk escalates
- Works smoothly with existing ADAS and safety ECUs
- Performs efficiently and consistently on embedded platforms such as Jetson and Orin
Overall, we aimed to deliver a behaviour-aware, predictive safety system that enhances driver well-being and supports next-generation vehicle safety architectures.
Solution
The system tracks critical indicators, including eye closure patterns, gaze direction, head movement, yawning, phone usage, and attention focus. These observations are processed together to understand the driver’s alertness and convert it into a scientifically validated fatigue score.
The system then delivers adaptive visual and audio alerts when it detects early signs of drowsiness or distraction. If the driver reaches higher fatigue levels, the DMS shares this information with the ADAS modules, enabling features such as lane keeping assist or emergency braking to respond promptly.
In essence, we delivered a predictive, behavior-aware DMS designed for production-ready integrations, built to recognize patterns early, act intelligently, and support the broader vehicle safety ecosystem.
Key Highlights
Holistic Behaviour Monitoring
- The system continuously analyses six behavioural dimensions: eye closure, gaze direction, head posture, yawning, phone usage, and overall attention. By combining these signals, the DMS develops a complete and reliable understanding of driver alertness, even when one cue is partially obscured.
Robust AI Perception for Real Driving Conditions
- Using accurate face detection and 3D facial landmark estimation techniques, the system maintains accuracy in challenging real-world scenarios, including variable lighting conditions, head movement, partial occlusions, and rapid gaze changes. All models are optimised to run on automotive-grade platforms.
Intelligent, Adaptive Alerts
- Instead of repetitive or intrusive warnings, the system adjusts the alert intensity based on the severity of fatigue. All alerts are logged, providing manufacturers and fleet operators with insights into behavioral trends, compliance needs, and long-term safety improvements.
Seamless ADAS Integration
- Through standard ROS interfaces, the DMS connects directly with lane-keeping assist, adaptive cruise control, and emergency braking features. This enables a coordinated safety response where the vehicle understands both its environment and the driver’s readiness to react.
Outcomes
- Improved driver attentiveness capture through continuous, data-driven fatigue evaluation.
- 40% reduction in distracted-driving events during controlled validation tests.
- Scientific fatigue scoring enabled drowsiness trend analysis and intelligent alerting strategies.
- Validated across 60+ hours of real-world driving and simulator-based testing.
- Production-ready deployment, designed for integration into in-vehicle ADAS architectures.
Technologies Used
- ROS Noetic – Enables message passing, modular node orchestration, and system scalability.
- Python, OpenCV, MediaPipe – Power real-time image processing and facial landmark detection.
- YOLOv8 – Detects distraction objects such as mobile phones.
- Custom HMI Interface – Displays real-time alerts and driver status on an interactive dashboard.
