Context-Aware Quantitative Risk Assessment Machine Learning Model for Drivers Distraction

TL;DR

The paper addresses distracted driving by proposing MDDRA, a context-aware, multiclass risk assessment framework that uses driver, vehicle, and environmental data to classify risk as safe, careless, or dangerous. It combines a risk assessment matrix with a probabilistic, frame-by-frame model (including a dynamic Bayesian network) and evaluates multiple ML approaches on TeleFOT data, finding Ensemble Bagged Trees to be the most accurate at 96.2%. Key contributions include a frame-level severity score S^*, a probabilistic data model for temporal inference, and a vehicle takeover mechanism triggered by high-severity frames. The study demonstrates that incorporating contextual cues from road type, weather, illumination, eye gaze, and hand state improves risk discrimination and enables rapid, robust takeover decisions in ADAS contexts. The results suggest practical potential for reducing distraction-related crashes, while pointing to future work in deep learning architectures and larger, more diverse datasets to further boost performance and generalization.

Abstract

Risk mitigation techniques are critical to avoiding accidents associated with driving behaviour. We provide a novel Multi-Class Driver Distraction Risk Assessment (MDDRA) model that considers the vehicle, driver, and environmental data during a journey. MDDRA categorises the driver on a risk matrix as safe, careless, or dangerous. It offers flexibility in adjusting the parameters and weights to consider each event on a specific severity level. We collect real-world data using the Field Operation Test (TeleFOT), covering drivers using the same routes in the East Midlands, United Kingdom (UK). The results show that reducing road accidents caused by driver distraction is possible. We also study the correlation between distraction (driver, vehicle, and environment) and the classification severity based on a continuous distraction severity score. Furthermore, we apply machine learning techniques to classify and predict driver distraction according to severity levels to aid the transition of control from the driver to the vehicle (vehicle takeover) when a situation is deemed risky. The Ensemble Bagged Trees algorithm performed best, with an accuracy of 96.2%.

Figures (13)

• Figure 1: MDDRA Model Architecture
• Figure 2: Context-aware probabilistic model for severity classification
• Figure 3: TeleFOT Dataset
• Figure 4: Distraction Interdependence State Diagram
• Figure 5: In-Vehicle State of Hand, Eye Gaze, and Face Orientation