Fast Flocking of Protesters on Street Networks

2024-06-03

AI Summary

This study addresses the challenge of modeling rapid, stable self-organized aggregation of high-density pedestrian crowds on urban road networks. Conventional flocking models (e.g., Boids) fail to accommodate the non-Euclidean, directed, and irregular topological structure of real-world road networks. To overcome this limitation, we propose the first biologically inspired flocking framework adapted to actual road infrastructure—a distributed multi-agent decision-making architecture grounded in local navigation-aware graph convolution. The framework jointly integrates road network embedding, graph neural networks (GNNs), and distributed reinforcement learning. Evaluated on 12 real-world urban road network datasets, our method achieves ≥92% path consistency and triples aggregation speed compared to baseline approaches—including extended Boids models and A*-based group path planning—while ensuring scalability and topological adaptability. This work establishes a novel, topology-aware paradigm for dynamic crowd coordination and emergency evacuation simulation.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Multiagent Systems: Agent-Based Simulation and Emergent Behavior
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Economics, Online Markets and Human Computation: Architectures and workflows that use LLMs for crowd work
• Search and Retrieval-Augmented AI: Agentic search

Problem

• Crowd Dynamics
• Gathering Efficiency
• Urban Environment

Innovation

• Simulated Game
• Crowd Aggregation
• Directional Consistency