Agent-based traffic simulation of emergency evacuation scenarios using SUMO and Python. This project analyzes evacuation performance, congestion patterns, and routing strategies to inform traffic planning.
Traffic evacuation modeling is essential for emergency planning (e.g., natural disasters, industrial accidents, urban evacuation drills). This repository provides a reproducible workflow that uses SUMO (Simulation of Urban MObility) to simulate evacuation scenarios and collect meaningful metrics on how traffic behaves under high-demand stress.
Efficient evacuation performance can save lives in real emergencies.
Key questions this project explores:
- How does congestion evolve as evacuation trips increase?
- What effects do different routing strategies have on total clearance time?
- Which network bottlenecks contribute most to overall delay?
By simulating evacuation demand over a realistic road network, we can explore these questions quantitatively.
| Component | Purpose |
|---|---|
| SUMO | Traffic simulation engine |
| TraCI (Python API) | Real-time interaction with SUMO |
| Python | Data processing & automation |
matplotlib / pandas |
Analysis & visualization |
The repository was originally developed as a group project for the Modeling and Simulation course at the Vienna University of Technology.
- Refactored the original codebase to eliminate significant code duplication and ensure reproducibility through the use of random seeds and configuration files.
- Modularized simulation scripts for flexible scenario configuration
- Automated simulation execution and metric collection
- Analysis of simulation results.
Neulengbach, Austria was chosen as the setting for the simulations. A danger zone that's 800 m in radius was defined at the center of the town. In some scenarios, a road was blocked, which is marked in the image below.
A histogram of the total evacuation time for various configurations is shown below. Total evacuation time is measured as the time at which the last car leaves the danger zone. Scenario 1 refers to all roads being open, while scenario 2 is the road as marked above being blocked. In the latter case, the cars are informed beforehand and plan a route avoiding the blockage during initialization. Each configuration was run 100 times.
As expected, total evacuation time increases with the number of cars. What's a bit more unexpected is that blocking a major road at the center of the danger zone only had a minimal effect on the total evacuation time (blue vs yellow).
- Python 3.12
- SUMO installed and added to system path.
- Important:
sumoorsumo-guimust point to the correct binaries.
- Important:
- Install required packages.
pip install -r requirements.txt - Adjust
config.yamlas needed and rundriver.py.- Note that it may take a minute for SUMO to activate after running
driver.pyfor the first time.
- Note that it may take a minute for SUMO to activate after running