Abstract:Traditional rule-based physical models are limited by their reliance on singular physical formulas and parameters, making it difficult to effectively tackle the intricate tasks associated with crowd simulation. Recent research has introduced deep learning methods to tackle these issues, but most current approaches focus primarily on generating pedestrian trajectories, often lacking interpretability and failing to provide real-time dynamic <a class="link-external link-http" href="http://simulations.To" rel="external noopener nofollow">this http URL</a> address the aforementioned issues, we propose a novel data-driven crowd simulation framework that integrates Physics-informed Machine Learning (PIML) with navigation potential fields. Our approach leverages the strengths of both physical models and PIML. Specifically, we design an innovative Physics-informed Spatio-temporal Graph Convolutional Network (PI-STGCN) as a data-driven module to predict pedestrian movement trends based on crowd spatio-temporal data. Additionally, we construct a physical model of navigation potential fields based on flow field theory to guide pedestrian movements, thereby reinforcing physical constraints during the simulation. In our framework, navigation potential fields are dynamically computed and updated based on the movement trends predicted by the PI-STGCN, while the updated crowd dynamics, guided by these fields, subsequently feed back into the PI-STGCN. Comparative experiments on two publicly available large-scale real-world datasets across five scenes demonstrate that our proposed framework outperforms existing rule-based methods in accuracy and fidelity. The similarity between simulated and actual pedestrian trajectories increases by 10.8%, while the average error is reduced by 4%. Moreover, our framework exhibits greater adaptability and better interpretability compared to methods that rely solely on deep learning for trajectory generation.

Exemplar-AMMs: Recognizing Crowd Movements From Pedestrian Trajectories

CrowdMove: Autonomous Mapless Navigation in Crowded Scenarios

Anomaly Detection and Localisation in the Crowd Scenes Using a Block-Based Social Force Model

Real-time Crowd Tracking using Parameter Optimized Mixture of Motion Models

Diverse Motions and Responses in Crowd Simulation

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

Crowd Behavior Analysis Via Curl and Divergence of Motion Trajectories

An Intelligent Extraversion Analysis Scheme from Crowd Trajectories for Surveillance

Crowd Saliency Detection via Global Similarity Structure

Model-based Crowd Behaviours in Human-solution Space

Appearance and motion-based unusual crowd events detection using multiple moving objects

Representing Dense Crowd Patterns Using Bag of Trajectory Graphs

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

Exploring Coherent Motion Patterns Via Structured Trajectory Learning for Crowd Mood Modeling

LCrowdV: Generating Labeled Videos for Simulation-based Crowd Behavior Learning

Anomaly Detection and Localization in Crowded Scenes Using Short-Term Trajectories

Detecting and Quantifying Crowd-Level Abnormal Behaviors in Crowd Events

Feedback-based metric learning for activity recognition

Toward Accurate Person-level Action Recognition in Videos of Crowed Scenes

A novel dynamic model for multiple pedestrians tracking in extremely crowded scenarios

Socially Aware Crowd Navigation with Multimodal Pedestrian Trajectory Prediction for Autonomous Vehicles