Prediction of Dangerous Driving Behavior Based on Vehicle Motion State and Passenger Feeling Using Cloud Model and...

Area of Science:

• Traffic Safety
• Artificial Intelligence
• Road Accident Prevention

Background:

• Dangerous driving behavior is a primary cause of road traffic accidents.
• Existing methods often focus on driver characteristics or sensor-based vehicle state estimation.
• Accurate and robust prediction of dangerous driving behavior remains a significant challenge in traffic safety management.

Purpose of the Study:

• To propose a novel hybrid model for predicting dangerous driving behavior.
• To integrate vehicle motion state estimation with passenger subjective feeling scores for intuitive perception.
• To enhance the accuracy and robustness of dangerous driving behavior prediction.

Main Methods:

• Development of a hybrid model combining a cloud model and Elman neural network (CM-ENN).
• Utilizing vehicle motion state estimation as input.
• Incorporating passenger subjective feeling scores for enhanced perception of dangerous driving.
• Real-world data acquisition in ShenZhen city, China.

Main Results:

• The proposed CM-ENN method demonstrated superior accuracy compared to classical neural network methods.
• The hybrid model showed enhanced robustness in predicting dangerous driving behaviors.
• Experimental results validated the effectiveness of the integrated approach.

Conclusions:

• The CM-ENN hybrid model offers a more accurate and robust solution for dangerous driving behavior prediction.
• Integrating vehicle motion and subjective feelings provides a more intuitive perception of driving risks.
• This approach contributes to advancing traffic safety management systems.