The Parameter Calibration of Social Force Model for Pedestrian Flow Simulation Based on YOLOv5

Area of Science:

• Urban planning and transportation science
• Computational social science
• Traffic engineering

Background:

• Subway systems are vital for urban transit, necessitating accurate pedestrian flow simulation for safety and management.
• Current simulation models require calibration with real-world data to enhance realism.
• Existing methods often yield significant discrepancies between simulated and actual pedestrian behavior.

Purpose of the Study:

• To develop and validate a calibration approach for pedestrian flow simulation models in subway stations.
• To improve the accuracy of the social force model using real-world pedestrian flow data.
• To reduce the gap between simulated and actual pedestrian dynamics in transit environments.

Main Methods:

• Utilized YOLOv5 for pedestrian detection and tracking to collect real-world flow data.
• Applied a calibration approach to optimize parameters within the social force model in Anylogic.
• Compared simulation outputs with collected real-world data to assess model performance.

Main Results:

• Calibrated model parameters successfully reproduced key characteristics of pedestrian flow in subway stations.
• The calibration method significantly decreased overall simulation errors.
• The approach mitigated the common issue of substantial differences between simulation results and reality.

Conclusions:

• The YOLOv5-based calibration approach enhances the fidelity of pedestrian flow simulations in subway environments.
• Accurate simulation is crucial for effective station management and risk assessment.
• This method provides a reliable tool for improving the predictive capabilities of pedestrian flow models.