Vehicle Directional Driving Behavior Segmentation Based on Cornering Strength

Abstract

With the rapid development in recent years, the autonomous driving system comes to the stage of commercial deployment now. However, it is not well accepted by society. This is due to the fact that the driving behavior features of autonomous driving vehicles are currently different from those of human drivers. It is important to understand how people drive for increasing acceptance. And the primary task of that is driving behavior identification. In this study, we proposed a segmentation method for vehicle driving process into basic directional driving behaviors. In order to clarify the relationship between vehicle motion state and driver operation, we rebuilt the vehicle motion state in the natural coordinate system. We innovatively proposed the concept of cornering strength as a metric of directional driving behavior intensity. To verify the validity of our approach, 22 participants were asked to maneuver a car through some common driving scenarios including turn, lane change, overtaking and U-turn. As a result, our method can extract directional driving behaviors with an average accuracy of 93.33%. In addition, the cornering strength of participants with different driving styles exhibit a significant difference though in the same driving scenario. Driving age has a significant effect on CS stability of participants. The variance of CS reaches a maximum of 0.22. The directional driving behavior features of the same participant can be different influenced by psychological or physiological aspects. Road adhesion condition and traffic congestion condition also have an effect on the CS values of participants.

Keywords

• Autonomous driving system
• cornering strength
• driving behavior segmentation
• maneuvering target tracking
• motion state rebuilding