Natural Disaster Mobility Model and Scenarios

This repository contains a mobility model for natural disasters as well as two specific disaster scenarios: 2013 Typhoon Haiyan (for the city of Tacloban) and 2010 Haiti earthquake (for the surroundings of Port au Prince).

It also contains a number of non-upstream patches to the ONE, e.g., b15434a99da132f1376e80fd071a5f6f8316ed1a, 4f7fa115373a972883991a2aca285e2c82d784f0, 0707229877c471b7ef6c343b2cd2f7e9466fa8eb, b27d8313512d8fa98cee4f78794d730eb1aaf8eb.

Code Navigation

This repository is a fork of the Opportunistic Network Environment (ONE) simulator and, thus, we place our code and data in the following locations:

• data/ contains map data and POI waypoints for the city of Tacloban and surroundings of PortauPrince
• src/movement/ contains the implementations of the sub-mobility models
  • NaturalDisasterMovementModel.java: the main mobility model class schedules role-dependent activities
  • naturaldisaster/ contains the role-dependent activities (go to the airport, sleep, etc.)
• the base directory contains the scenario instantiations (tacloban_settings.txt and portauprince_settings.txt) in form of ONE settings files.

Usage

To use the model, simply base your evaluation scenario on one of the scenario instantiations. If you plan to create your scenario from scratch, you need to define:

• GroupX.role = {Healthy, Injured, Scientist, UN, Government, SnR, DRO} for each group. This setting is independent from GroupX.groupID, so you can have different groups with the same role.
• Group.nbrOfDays as the number of days that the scenario will be run for (in a future version of this model, this could be derived from the Scenario.endTime setting)

Publications

For more information on the underlying model, please refer to our paper. If you use this model or code in your own work, we are happy to receive a citation.

Milan Stute, Max Maass, Tom Schons, and Matthias Hollick, “Reverse Engineering Human Mobility in Large-scale Natural Disasters,” to appear in ACM International Conference on Modeling, Analysis and Simulation of Wireless and Mobile Systems (MSWiM), November 2017, Miami Beach, USA. [PDF]

Reproducibility

For the sake of reproducibility, we provide the complete experimental data set that was used for the plots in our paper (requires Python packages numpy and matplotlib) at 10.5281/zenodo.836815.

Alternatively, you may generate the data set yourself by running (warning: simulations might take several days to complete):

./exp.py

After completion, experiment data and plots will be available in out/<DATE>/{reports,plots}.