Physics-Informed Neural Network Modeling of Nonlocal Crowd Dynamics for Evacuation Scenarios
Abstract
We investigate a nonlocal continuum model of crowd dynamics using a physics-informed neural network approach. The crowd is described by a system of nonlinear conservation laws in which the flux incorporates advection, diffusion, and nonlocal interaction terms accounting for density-dependent motion and limited perception of surrounding agents. Nonlocal effects are modeled through spatial convolutions with smooth kernels, enabling agents to respond to averaged density gradients rather than purely local information. The governing system of partial differential equations is solved using a physics-informed neural network known as PINN, which approximates the solution over the entire space–time domain while enforcing the physical constraints through automatic differentiation. The nonlocal interaction terms are implemented in a stable discrete convolution form, ensuring numerical robustness during training. The approach is demonstrated on the interaction of two pedestrian groups moving in opposite directions in a one-dimensional corridor. The results exhibit the formation and propagation of density fronts, the gradual merging of flows, and the emergence of stable mixed zones. A characteristic feature of the solution is the partial interpenetration of the groups without rigid collisions, reflecting realistic collective motion. To validate the method, the PINN solution is compared with a reference finite-difference scheme based on a Rusanov flux. Qualitative agreement is observed in front structure and mixing dynamics, while quantitative deviations in key characteristics remain
within a few percent. A systematic parameter study shows that the PINN-based solution remains stable under variations of advection velocity, diffusion coefficient, and nonlocal interaction radius, in contrast to the finite-difference scheme, which exhibits strong stability limitations. These results demonstrate that PINN provides a robust and physically consistent tool for modeling nonlinear nonlocal crowd dynamics.
Downloads
References
L.F. Henderson, "The Statistics of Crowd Fluids," Nature, 229(5284), 381–383 (1971). https://doi.org/10.1038/229381a0
L.F. Henderson, "On the fluid mechanics of human crowd motion," Transportation Research, 8, 509–515 (1974), https://doi.org/10.1016/0041-1647(74)90027-6
N. Bellomo, and C. Dogbe, "On the Modeling of Traffic and Crowds: A Survey of Models, Speculations, and Perspectives," SIAM Review, 53(3), 409 (2011). https://doi.org/10.1137/090746677
R.L. Hughes, "A continuum theory for the flow of pedestrians," Transportation Research Part B: Methodological, 36(6), 507-535 (2002). https://doi.org/10.1016/s0191-2615(01)00015-7
R.M. Colombo, and M. Lécureux-Mercier, "Nonlocal Crowd Dynamics Models for Several Populations," Acta Mathematica Scientia, 32(1), 177-196 (2012). https://doi.org/10.1016/s0252-9602(12)60011-3
R.J. LeVeque, Finite Volume Methods for Hyperbolic Problems, (Cambridge University Press, Cambridge, 2002), https://doi.org/10.1017/CBO9780511791253.
J. P. Boyd, Chebyshev and Fourier Spectral Methods, (Dover Publications, Mineola, NY, 2001).
M. Raissi, P. Perdikaris, and G. E. Karniadakis, "Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations," Journal of Computational Physics, 378, 686–707 (2019). https://doi.org/10.1016/j.jcp.2018.10.045.
A. D. Jagtap, E. Kharazmi and G. E. Karniadakis, "Conservative physics-informed neural networks on discrete domains for conservation laws: Applications to forward and inverse problems," Computer Methods in Applied Mechanics and Engineering, 365, 113028 (2020). https://doi.org/10.1016/j.cma.2020.113028.
S. Wang, Y. Teng, and P. Perdikaris, "Understanding and mitigating gradient pathologies in physics-informed neural networks," Journal of Computational Physics, 449, 110768 (2021). https://doi.org/10.1137/20m1318043.
R.J. LeVeque, and J. Randall, Numerical Methods for Conservation Laws, (Birkhäuser Verlag, Basel, 1992).
Copyright (c) 2026 A. Naumovets, P. Kuznietsov, V. Cherkashyn, A. Gakh
This work is licensed under a Creative Commons Attribution 4.0 International License.
Authors who publish with this journal agree to the following terms:
- Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution License that allows others to share the work with an acknowledgment of the work's authorship and initial publication in this journal.
- Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgment of its initial publication in this journal.
- Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work (See The Effect of Open Access).
