A personalized driving risk assessment and rolling prediction method by integrating multiple indicators.

Area of Science:

• Traffic Safety and Intelligent Transportation Systems
• Biomedical Engineering and Human Factors

Background:

• Driving styles vary significantly, necessitating personalized risk assessment for proactive traffic safety.
• Existing systems lack real-time, personalized prediction of accident risk based on integrated physiological and vehicle data.
• Intelligent driver assistance systems require accurate risk prediction for effective intervention.

Purpose of the Study:

• To propose a personalized driving risk assessment and rolling prediction method.
• To integrate vehicle dynamics and Heart Rate Variability (HRV) for real-time accident risk prediction.
• To develop an advanced model for continuous driving risk prediction.

Main Methods:

• A natural driving experiment collected Critical Incident Events (CIEs) and indicator data.
• An improved Bayesian network identified key personalized indicators influencing CIEs.
• The Bayesian Optimization (BO)-based Bidirectional Gated Recurrent Unit (BiGRU) integrated with CNN and EKF (BCBGE model) was developed for prediction.

Main Results:

• Driving risk was categorized into four distinct levels.
• Personalized analysis revealed 3-4 key vehicle dynamics or HRV indicators per CIE type.
• The BCBGE model achieved 92.03% prediction accuracy, surpassing other models by 3.37%-15.88%.

Conclusions:

• The proposed personalized method effectively assesses and predicts driving risk in real-time.
• Integration of vehicle dynamics and HRV provides a robust basis for intelligent driver assistance.
• The BCBGE model demonstrates superior performance in continuous driving risk prediction.