Evacuation Simulation

A discrete agentic automation simulation of pedestrian evacuation dynamic implemented in Python using Pygame. It is based on the research paper published about floor field models introduced by Burstedde et al. (2001).

Agents navigate a room toward a single exit using static floor field computed by Dijkstra's algorithm which assigns every cell a cost base on its distance to the exit. Agents choose their next cell stochastically using transition probability formula from the paper:

p_ij = exp(KS * static_ij + KD * dynamic_ij) * (1 - occupied_ij) * wall_ij

Where KS * static_ij is the difference in static field between the neighbor and current cell, KD * static_ij is the dynamic field which is the pheromone left behind each agent, (1 - occupied_ij) blocks occupied cells, and wall_ij blocks walls. Conflict resolution is handled by randomly selecting one winner when multiple agents compete for the same cell.

The simulation is designed to demonstrate the faster-is-slower effect the counterintuitive phenomenon where increasing agent urgency (via KD) leads to jamming near the exit and slower overall evacuation times.

Built with: Python, Pygame, NumPy

Set up the virtual environment

Make sure Python is installed and in the PATH Run command to create venv: python -m venv .venv

Windows (activate venv)

.venv\Scripts\activate

Linux

source .venv/bin/activate

Use requirements.txt to get the dependencies

• After activating virtual env, run:

  pip install -r requirements.txt

To change the experiment outcomes

Go to config.py:

• make changes to the room dimension
• change the KD and KS which in turn changes the urgency of each agent
• ALHPA and DELTA changes the diffusion and decay. Meaning, it changes how fast the decay and diffusion happens.

To run the experiment:

• python app.py