| Abstract: |
| This presentation explores the application of Fourier Neural Operators (FNOs) to solve the Hughes pedestrian flow model. The Hughes model, a macroscopic crowd dynamics model, incorporates a velocity field dependent on the density distribution and a turning point representing the direction change due to congestion. The approach leverages the FNO framework to learn solutions of the model efficiently. Training data are generated using the wave-front tracking scheme, ensuring adherence to the entropy condition. The presentation outlines the mathematical formulation of the Hughes model, discusses the training methodology and dataset preparation, and demonstrates the FNO architecture`s capacity to predict accurate density profiles and turning points. This study highlights the potential of operator learning in tackling complex PDE-constrained problems in crowd dynamics. |
|