SEMT20001: Principles of Computational Modelling
COURSEWORK REASSESSMENT: QUESTION 2
Question 2: Agent-based Modelling — The Zombie Apocalypse: Zombie Edition
There are many examples of infection spreading in both nature and fiction, although none are quite so agressive as zombies.
In this coursework question, we will be using a simple ABM representing an SI infection spread model where the infected agents (zombies) actively try to infect the susceptible agents (humans).
We make the following assumptions:
(A1) A human bitten by a zombie becomes a zombie.
(A2) If a human is within range of a zombie then the zombie will attack and infect them.
(A3) Zombies have perfect sensing, but humans have erroneous sensing.
(A4) Humans and zombies move with the same speed.
(A5) A human will stay still if they sense a zombie nearby, otherwise they will move randomly.
(A6) Zombies have a greater sensing radius than humans.
(A7) The more a zombie moves, the more energy it expends, and the more likely it will die.
The overall aim is to answer the following questions:
• How many people will be turned into zombies and how fast will this occur?
• Which pursuit strategies lead to the extinction of humans?
You have been given the following three Python files:
• ZombieModel.py : This contains the ZombieWorld model class and some helper functions for calcu- lating the number of humans, zombies and dead agents. The ZombieWorld class has three methods: init , step and make agents. Your answers should only require you to edit the init method.
• ZombieAgent .py : This contains the Agent class. The Agent class has four types of methods: general agent class methods, human specific methods, zombie specific methods and movement helper methods. Your answers should only require you to edit the zombie specific methods, or write new methods.
• ZombieApp.py : This contains the code used to create the simulation visualisation. You can run it using solara run ZombieApp.py. You answers should not require you to edit this file.
All written answers and any code used to produce results and figures should be submitted in a Jupyter Notebook.
In all your answers, you should ensure that your results are reproducible, by giving all relevant methods, model parameters, random number seeds, etc. Also, don’t forget to think carefully about how long to run the simulations and the number of replications needed.
(a) Run the given model, with no modifications, and produce a plot showing how the number of human and zombie agents changes over time. (2 marks)
(b) Describe how the environment is represented in the model, including how agent movement, sensing, and interaction have been implemented in the starter model. (4 marks)
(c) In the current version of the model, the zombies will move to a random neighbouring cell if they sense a human nearby, otherwise they stand still. Your task is to design two different spatial pursuit strategies that the zombies could implement. Any strategies you implement must still satisfy the above assumptions. Implement these three strategies in your model, using the agent attribute zombie persuit strategy to switch between behaviours. Describe and explain how each of your two strategies should work, why you have chosen them, and how you have verified that they work as intended. (6 marks)
(d) Now we want a way of capturing the zombie’s energy expenditure in order to compare between strategies.
(i) Add a new attribute to the Agent class called zombie pursuit move flag that indicates whether a zombie has moved that timestep. Modify the move method of the Agent class so that this attribute is equal to 1 if the zombie has moved that timestep as part of its pursuit strategy (i.e. ignoring movement from try infect) and 0 if it has remained stationary. Then add a new model reporter to the Datacollector of the ZombieWorld class that calculates the total number of pursuit movements taken by zombies each timestep.
(ii) Produce a time series plot showing how the total number of zombie movement steps varies over time for the different pursuit strategies. Comment on this plot. (3 marks)
(e) Compare your pursuit strategies and the baseline random movement. Which strategy would you recommend zombies use, and why?
(i) Produce a time series plot showing how the number of infected or dead agents vs total agents changes over time, and how this varies for the different strategies. Comment on this plot.
(ii) Evaluate and compare the pursuit strategies using two metrics; the time to human extinction and the total number of pursuit movement steps taken by zombies. Comment on your findings, including any qualitative observations you might make using the visualisation, and how they influence your recommendation. (Hint: If human extinction does not occur in a particular simulation run, what is a reasonable stopping criterion for comparison?)
If you did not implement the zombie persuit move flag attribute in part (c), then use the median human survival time instead. (7 marks)
(f) Explore the sensitivity of your findings as the zombie sensing radius varies from 1 to 10 using the metrics you calculated in part (e)(ii). Comment on your findings, including any qualitative observations you might make using visualisation. Would you still recommend the same strategy? If not, why not?
If you cannot use the metrics from part (e)(ii) then use the number of zombies, humans and dead agents at a particular timestep instead. (3 marks)