Rigid body agent-based crowd simulation framework utilizing simplified ray casting, social forces model, and heat map parameters fitting

Abstract

This paper presents a novel rigid body agent-based crowd simulation framework utilizing a simplified Social Forces Model based on ray casting. This approach is specially designed to be integrated with simulation engines (which enable ray cast-based collision detection) and to take advantage of their other functionalities. We also present a parameter fitting method that minimizes the root mean square error (RMSE) between heat maps generated by real-world and simulated crowd motion. The intuitive RMSE value can therefore be used to evaluate the consistency of the simulation with the reference (real-world) data and can potentially be used for model parameter fitting. The dataset and source code are available for download so that our experiments can be reproduced.