HyPedSim: A Multi-Level Crowd-Simulation Framework-Methodology, Calibration, and Validation.

Area of Science:

• Complex systems modeling
• Computational social science
• Traffic and crowd dynamics

Background:

• Modeling large-scale crowd phenomena is challenging due to varying pedestrian behaviors and densities.
• Existing microscopic and mesoscopic models have limitations in low-density and high-density scenarios, respectively.

Purpose of the Study:

• To propose and evaluate a novel agent-based model that dynamically switches between microscopic and mesoscopic operational models based on local crowd density.
• To enable multi-scale simulation capabilities for pedestrian flow.

Main Methods:

• Development of a hybrid agent-based model integrating microscopic and mesoscopic approaches.
• Calibration and validation using a genetic algorithm against real-world pedestrian outflow data from the Festival of Lights, Lyon.
• Local sensitivity analysis to identify key model parameters.

Main Results:

• The hybrid model effectively simulates pedestrian flow across diverse density conditions.
• Demonstrated optimization in computational performance compared to existing models.
• Successful calibration and validation using real-world data.

Conclusions:

• The proposed hybrid model offers a robust and computationally efficient solution for simulating large-scale crowd phenomena.
• This approach enhances the accuracy and applicability of agent-based modeling for pedestrian dynamics in complex environments.