Abstract:Crowd simulation acts as the basic component in traffic management, urban planning, and emergency management. Most existing approaches use physics-based models due to their robustness and strong generalizability, yet they fall short in fidelity since human behaviors are too complex and heterogeneous for a universal physical model to describe. Recent research tries to solve this problem by deep learning methods. However, they are still unable to generalize well beyond training distributions. In this work, we propose to jointly leverage the strength of the physical and neural network models for crowd simulation by a Physics-Infused Machine Learning (PIML) framework. The key idea is to let the two models learn from each other by iteratively going through a physics-informed machine learning process and a machine-learning-aided physics discovery process. We present our realization of the framework with a novel neural network model, Physics-informed Crowd Simulator (PCS), and tailored interaction mechanisms enabling the two models to facilitate each other. Specifically, our designs enable the neural network model to identify generalizable signals from real-world data better and yield physically consistent simulations with the physical model's form and simulation results as a prior. Further, by performing symbolic regression on the well-trained neural network, we obtain improved physical models that better describe crowd dynamics. Extensive experiments on two publicly available large-scale real-world datasets show that, with the framework, we successfully obtain a neural network model with strong generalizability and a new physical model with valid physical meanings at the same time. Both models outperform existing state-of-the-art simulation methods in accuracy, fidelity, and generalizability, which demonstrates the effectiveness of the PIML framework for improving simulation performance and its capability for facilitating scientific discovery and deepening our understandings of crowd dynamics. We release the codes at https://github.com/tsinghua-fib-lab/PIML.

Physics-infused Machine Learning for Crowd Simulation

A Data-driven Crowd Simulation Framework Integrating Physics-informed Machine Learning with Navigation Potential Fields

Model-based Crowd Behaviours in Human-solution Space

Enhanced Crowd Dynamics Simulation with Deep Learning and Improved Social Force Model

Social Physics Informed Diffusion Model for Crowd Simulation

Physics-Informed Machine Learning for Modeling and Control of Dynamical Systems

When Physics Meets Machine Learning: A Survey of Physics-Informed Machine Learning

Physics-guided machine learning from simulated data with different physical parameters

Understanding complex crowd dynamics with generative neural simulators

Physics-Informed Machine Learning: A Survey on Problems, Methods and Applications

Understanding and Modeling Collision Avoidance Behavior for Realistic Crowd Simulation

Diverse Motions and Responses in Crowd Simulation

The Application of Physics-Informed Machine Learning in Multiphysics Modeling in Chemical Engineering

Physics-Based Dynamic Models Hybridisation Using Physics-Informed Neural Networks

A Radar-Nearest-Neighbor Based Data-Driven Approach for Crowd Simulation

Multi-level Crowd Simulation Using Social LSTM

Applying physics-based loss functions to neural networks for improved generalizability in mechanics problems

Crowd evacuation simulation based on hierarchical agent model and physics‐based character control

Physics-informed machine learning

Physics-informed machine learning for moving load problems

Data-driven physics-based modeling of pedestrian dynamics