Estimate Crowd Flow Including Side-trip Behavior When Exiting from Large-scale Event Venues

This study introduces a novel extended crowd simulation model for the task of estimating crowd flows in large-scale environments under unforced situations, where individuals freely select paths to their destination, such as exiting an event venue after participating in activities. The crowd flows in unforced situations, including side-trip behavior, such as stopping at roadside vending machines, stores, restrooms, and trash cans before arriving at their primary destination. The traditional models in a large-scale environment often focus on simulating forced situations such as evacuation. To accurately estimate unforced situations, we propose a minimal extended model to represent a stochastic side-trip. The proposed model contributes to evaluating crowd behavior in an entire building based on partial measurement data and reducing the high costs of deploying extensive measurement equipment and personnel. To evaluate the model, data on spectators exiting the Tokyo Dome were measured using LiDAR and 8K cameras. Evaluation results using measured data indicate that the proposed model achieves a 9.25% improvement in simulation accuracy over previously established models when evaluated on the same-day data. Furthermore, the model maintains superior performance, with a 3.15% improvement, even when calibrated parameters are applied to data from different day’s scenarios, demonstrating its robustness and generalizability.