Driver Behaviour Analysis to Improve Road Safety: A Comprehensive Study

Road safety is a shared responsibility that includes following traffic regulations, using defensive driving methods, and prioritizing vehicle maintenance. Consistent alertness behind the wheel, following speed limits, and avoiding distractions like phone use are critical factors in avoiding avoidable road accidents. The number of road accidents, however, has not fallen considerably. Hence, there is a need for a comprehensive study to tackle this problem. This study introduces an innovative approach integrating driver support and infotainment systems to categorize drivers based on their skills and states. Utilizing a machine learning methodology, we classify drivers’ maneuvering skills by employing longitudinal and lateral controls, coupled with vehicle speed. This classification relies on a holistic model constructed from sensor data capturing the driving environment, driver behaviors, and vehicle responses. The developed model enables automatic classification of an unknown driver's safety level, providing a personalized approach to enhance road safety. In this paper, our focus shifts to assessing driving skills using results from a driving simulator, particularly in curve driving scenarios. Our methodology employed various machine learning algorithms to train the model, including K-Nearest Neighbors (KNN), Gradient Boosting, AdaBoost, and Random Forest classifiers. Among these, Random Forest exhibited the highest accuracy, achieving an impressive 83% classification accuracy in distinguishing between safe and aggressive drivers. This research contributes to the development of intelligent systems that not only enhance road safety and encourage conscientious driving behaviors but also personalize infotainment offerings, providing a comprehensive approach to driver satisfaction and safety)