Traffic conflict assessment using macroscopic traffic flow variables: A novel framework for real-time applications

Area of Science:

• Transportation Engineering
• Traffic Safety Analysis
• Data-Driven Modeling

Background:

• Vehicle-to-vehicle interactions occur in both lateral and longitudinal dimensions.
• Existing traffic conflict assessment often lacks a comprehensive macroscopic approach.
• Macroscopic traffic state variables offer potential for real-time safety evaluation.

Purpose of the Study:

• To develop a comprehensive traffic conflict assessment framework using macroscopic traffic state variables.
• To propose a two-dimensional framework for evaluating time spent in conflict (TSC).
• To model TSCs using both econometric and machine learning approaches.

Main Methods:

• Utilized vehicular trajectory data from a ten-lane divided urban expressway in India.
• Developed a two-dimensional framework based on the subject vehicle's influence zone to evaluate TSCs.
• Employed a two-step modeling process: Grouped Random Parameter Tobit (GRP-Tobit) model followed by machine learning models (including Random Forest).

Main Results:

• Intermediately congested traffic flow conditions were identified as critical for traffic safety.
• Macroscopic traffic variables (density, speed, standard deviation in speed, composition) positively influence TSC.
• The Random Forest model demonstrated the best performance in predicting TSC based on macroscopic traffic variables.

Conclusions:

• The developed framework effectively assesses traffic conflicts using macroscopic variables.
• Real-time traffic safety monitoring is facilitated by the predictive capabilities of the machine learning models.
• Understanding the relationship between macroscopic traffic states and conflict indicators is crucial for improving road safety.