EvacSim

EvacSim is a deterministic simulation that uses graph theory and linear programming to model the effects of hurricanes on communities in order to predict optimal evacuation routes. This project was originally developed for HackRPI 2019, a 24-hour hackathon in which it won 2nd place.

Background

This project was originally developed for HackRPI 2019, a 24-hour hackathon in which it won 2nd place. The idea was born after witnessing the devastating effects of Hurricane Dorian as well as hearing from friends and family in tropical regions who were forced to evacuate their homes due to storms numerous times over the years.

How it works

Model

In the model, populations are represented by nodes and the infrastructure between them are represented by edges. These nodes and edges can then be used to construct a graph. The actual modeling happens in CSV files, such as the following nodes.csv file modeling populations around Troy, NY:

Enabled	Name	Latitude	Longitude	Population	Capacity
0	Troy	42.718	-73.687	50000	60000
0	Albany	42.649	-73.753	98000	120000
0	Guilderland	42.702	-73.909	36000	50000
0	EastGreenbush	42.588	-73.703	16000	30000
0	Brunswick	42.732	-73.562	16000	35000
0	Schenectady	42.806	-73.943	65000	70000

Similar models exist for edges, hurricanes, and the main setup:

• edges.csv: Models infrastructure (ie. roads) between populations by source node, destination node, transit time, and maximum capacity.
• hurricanes.csv: Models the hurricane by initial position, wind speed variation, velocity, and trajectory.
• main.csv: Models the start and end time for the simulation.

Examples can be found in the models folder.

Simulation

The simulation has three main parts: the simulator itself, which is written in Java, the AMPL model, and the boilerplate script, which is written in JavaScript with Node.js. The script first fetches the model files from where they are stored in the IBM cloud. The files are then passed to the simulator, which translates the model files to usable data. From the hurricane data, the path and area of effect of the storm is determined, which is then used to determine which populations should evacuate. The data for edges, nodes, and affected nodes are then passed to an AMPL model which runs a minimum flow algorithm on the graph to determine where each affected population can evacuate to without violating maximum capacities on communities and infrastructure. This data is passed back to the Java simulation, which generates a KML file showing the area of effect of the hurricane, the severity in each region, the evacuation routes for each affected community, and data such as the final populations in each community. This KML file can be displayed in a geographic browser such as Google Earth.

What's next

Some possibilities for extensions of the project include:

• Supporting multiple types of natural disasters, such as earthquakes and wildfires
• Improving the accuracy of hurricane representations by allowing for varying velocities and trajectories
• Using IBM Cloud services more extensively
• Creating a custom frontend to display evacuation routes rather than generating KML files