How do cyclists interact with motorized vehicles at unsignalized intersections? Modeling cyclists' yielding behavior...

Area of Science:

• Traffic Safety
• Human-Computer Interaction
• Robotics

Background:

• Cyclist fatalities at unsignalized intersections remain a concern.
• Existing models lack cyclist behavioral cues, limiting prediction accuracy.
• Automated vehicle safety necessitates accurate prediction of cyclist intentions.

Purpose of the Study:

• To develop a quantitative model predicting cyclist crossing intentions at unsignalized intersections.
• To investigate the impact of behavioral cues on prediction accuracy.
• To enhance threat assessment algorithms for active safety systems.

Main Methods:

• Extracted interaction events from naturalistic trajectory data.
• Enriched data with cyclist behavioral cues from sensors (pedaling, head movement).
• Developed a quantitative model incorporating kinematics and behavioral cues.

Main Results:

• Both kinematics and behavioral cues were statistically significant predictors of cyclist yielding behavior.
• Non-verbal communication cues demonstrably improve prediction models.
• The proposed model shows enhanced predictive capabilities.

Conclusions:

• Incorporating cyclist behavioral cues into threat assessment algorithms is vital for automated vehicle safety.
• This research provides a foundation for more sophisticated cyclist behavior prediction.
• Improved prediction models can lead to reduced cyclist-vehicle conflicts and fatalities.