Spatial-Game Evacuation Model for Simulating Evacuation Dynamics with Heterogeneous Coordination Games

Description

This repository contains the code for executing the Spatial-Game Evacuation Model (SGEM) used for studying a novel representation of human behavior in evacuation dynamics. Built upon traditional floor field dynamics, the SGEM is an agent-based model that adds an asymmetric coordination game theoretic layer to operationalize human decision-making, and includes heterogeneous risk tolerance and evolving stress states among all agents. This code repository contains all model and simulation code as well preliminary validation against empirical evacuation dataset references and results visualization in code notebooks.

Steps to reproduce

The model is implemented in Python, v3.12, with dependencies enumerated in requirements.txt. To run a single simulation, open and define the simulation with "config.py", then run "python run.py". Experimental results may be replicated by running the full experiment suite with "python experiments/run_tuning.py" and validation results may be replicated with "python experiments/run_tuning.py". Results may be visualized in Jupyter notebooks by generating the notebook file with "python analysis/build_notebook.py" and "python analysis/build_tuning_notebook.py" and then running the associated generated notebooks. Specific experiments, validation runs, or one-off visuals may also be produced by following documentation within the repository.

Institutions

• George Washington University

  District of Columbia, Washington

Categories

Licence