PimvanderLoos / c18-virus-model

DMAS Project 20/21

c18-virus-model

DMAS Project 20/21

Code for the implementation of our project called 'Contact Tracing in Virus Models' for the course 'Design of Multi-agents Systems'. The code uses Mesa, the Apache2 licensed agent-based modeling framework in Python.

Summary

The model simulates the spread of Covid-19 in the setting of a university building using either no mitigation strategies or contact tracing, a mitigation measure that uses a mobile application to not only track and quarantine infected persons but also contacts of infected persons.
Every agent in the model follows their own schedule which defines when they have to sit in which room.

Every day in the model simulates 8 hours, split into 4 2-hour lecture slots. Because this model only looks at the spread of the virus in the setting of a university, the remaining 16 hours and the weekends are not modeled, though they do count towards time-based variables such as disease progression.

Once an agent is infected, they go through several stages: Infected -> Testable -> Infectious -> Symptomatic.
Once an agent becomes testable (including later stages), their tests can return a positive result. Once an agent becomes infectious (including later stages) they can infect other agents. After the symptomatic stage agents will either die or recover. Once recovered, they cannot get infected again.

Running the model

1. Install Mesa and its dependencies:

$ pip install -r https://raw.githubusercontent.com/PimvanderLoos/c18-virus-model/master/requirements.txt?token=AAXYNY2NM73QHJ7GULYUP5K7VKETW

2. Clone the online repository:

$ git clone https://github.com/PimvanderLoos/c18-virus-model.git

3. Run the program (from Master branch)

$ python run_model.py

The visualization should open in a new browser window.

The visualization

The image above shows what the model looks like. The big rectangle at the top is the break room, where agents go to when they have a break in their schedule. The squares below that are the lecture rooms, with the colored squares representing the seats.

The different colors of the agents represent their disease states, as further explained in the legend.

Run multiple experiments

To run multiple experiments, we provide a way to run the model without its visualization.

1. Define/Review the list of experiments to run in bulk_experiments.txt. This file contains one experiment definition per line using the commandline arguments of the run_model_noviz module.
   Here is an example of 2 experiment definitions, one without any mitigation strategies and one that uses contact tracing:

simulation_output_no_mitigation --stepCount 2000 --mitigation no_measures --num_agents 800 --baseInfection 3 --spreadChance 8 --spreadDistance 2 --testDelay 5 --random-seed 42 --write-plots --room_size 15 --room_count 10 --break_room_size 22
simulation_output_contact_tracing --stepCount 2000 --mitigation contact_tracing --num_agents 800 --baseInfection 3 --spreadChance 8 --spreadDistance 2 --testDelay 5 --random-seed 42 --write-plots --room_size 15 --room_count 10 --break_room_size 22

2. Run the experiments:

$ python run_bulk_experiments.py bulk_experiments.txt

The combined results of all the experiments can be found in output.csv, which will give you the following statistics about each experiment:

• Death count.
• Total number of agents that got infected during the simulation.
• Total number of tests performed (and whether they are negative or positive).
• The peak number of infected agents.
• The peak number of quarantined agents (and whether they are healthy or infected).
• The peak number of infected agents that were not quarantined.
• The peak difference in number of healhty quarantined agents compared to infected quarantined agents.

For every individual experiment a folder with its name (first entry of the experiment definition) will be created.
If --write-plots was specified, 4 plots will be generated to visualize the results. These are some example plots for a simulation without contact tracing:

An application of generating results to investigate the effect of the participation rate in contact tracing on the statistics can be found in the folder results.