Wolf-Sheep Predation Model#

Summary#

A simple ecological model, consisting of three agent types: wolves, sheep, and grass. The wolves and the sheep wander around the grid at random. Wolves and sheep both expend energy moving around, and replenish it by eating. Sheep eat grass, and wolves eat sheep if they end up on the same grid cell.

If wolves and sheep have enough energy, they reproduce, creating a new wolf or sheep (in this simplified model, only one parent is needed for reproduction). The grass on each cell regrows at a constant rate. If any wolves and sheep run out of energy, they die.

The model is tests and demonstrates several Mesa concepts and features:

MultiGrid
Multiple agent types (wolves, sheep, grass)
Overlay arbitrary text (wolf’s energy) on agent’s shapes while drawing on CanvasGrid
Agents inheriting a behavior (random movement) from an abstract parent
Writing a model composed of multiple files.
Dynamically adding and removing agents from the schedule

Installation#

To install the dependencies use pip and the requirements.txt in this directory. e.g.

    # First, we clone the Mesa repo
    $ git clone https://github.com/projectmesa/mesa.git
    $ cd mesa
    # Then we cd to the example directory
    $ cd examples/wolf_sheep
    $ pip install -r requirements.txt

How to Run#

To run the model interactively, run mesa runserver in this directory. e.g.

    $ mesa runserver

Then open your browser to http://127.0.0.1:8521/ and press Reset, then Run.

Files#

wolf_sheep/random_walk.py: This defines the RandomWalker agent, which implements the behavior of moving randomly across a grid, one cell at a time. Both the Wolf and Sheep agents will inherit from it.
wolf_sheep/test_random_walk.py: Defines a simple model and a text-only visualization intended to make sure the RandomWalk class was working as expected. This doesn’t actually model anything, but serves as an ad-hoc unit test. To run it, cd into the wolf_sheep directory and run python test_random_walk.py. You’ll see a series of ASCII grids, one per model step, with each cell showing a count of the number of agents in it.
wolf_sheep/agents.py: Defines the Wolf, Sheep, and GrassPatch agent classes.
wolf_sheep/scheduler.py: Defines a custom variant on the RandomActivationByType scheduler, where we can define filters for the get_type_count function.
wolf_sheep/model.py: Defines the Wolf-Sheep Predation model itself
wolf_sheep/server.py: Sets up the interactive visualization server
run.py: Launches a model visualization server.

Agents#

from mesa.experimental.cell_space import CellAgent, FixedAgent


class Animal(CellAgent):
    """The base animal class."""

    def __init__(
        self, model, energy=8, p_reproduce=0.04, energy_from_food=4, cell=None
    ):
        """Initialize an animal.

        Args:
            model: Model instance
            energy: Starting amount of energy
            p_reproduce: Probability of reproduction (asexual)
            energy_from_food: Energy obtained from 1 unit of food
            cell: Cell in which the animal starts
        """
        super().__init__(model)
        self.energy = energy
        self.p_reproduce = p_reproduce
        self.energy_from_food = energy_from_food
        self.cell = cell

    def spawn_offspring(self):
        """Create offspring by splitting energy and creating new instance."""
        self.energy /= 2
        self.__class__(
            self.model,
            self.energy,
            self.p_reproduce,
            self.energy_from_food,
            self.cell,
        )

    def feed(self):
        """Abstract method to be implemented by subclasses."""

    def step(self):
        """Execute one step of the animal's behavior."""
        # Move to random neighboring cell
        self.move()

        self.energy -= 1

        # Try to feed
        self.feed()

        # Handle death and reproduction
        if self.energy < 0:
            self.remove()
        elif self.random.random() < self.p_reproduce:
            self.spawn_offspring()


class Sheep(Animal):
    """A sheep that walks around, reproduces (asexually) and gets eaten."""

    def feed(self):
        """If possible, eat grass at current location."""
        grass_patch = next(
            obj for obj in self.cell.agents if isinstance(obj, GrassPatch)
        )
        if grass_patch.fully_grown:
            self.energy += self.energy_from_food
            grass_patch.fully_grown = False

    def move(self):
        """Move towards a cell where there isn't a wolf, and preferably with grown grass."""
        cells_without_wolves = self.cell.neighborhood.select(
            lambda cell: not any(isinstance(obj, Wolf) for obj in cell.agents)
        )
        # If all surrounding cells have wolves, stay put
        if len(cells_without_wolves) == 0:
            return

        # Among safe cells, prefer those with grown grass
        cells_with_grass = cells_without_wolves.select(
            lambda cell: any(
                isinstance(obj, GrassPatch) and obj.fully_grown for obj in cell.agents
            )
        )
        # Move to a cell with grass if available, otherwise move to any safe cell
        target_cells = (
            cells_with_grass if len(cells_with_grass) > 0 else cells_without_wolves
        )
        self.cell = target_cells.select_random_cell()


class Wolf(Animal):
    """A wolf that walks around, reproduces (asexually) and eats sheep."""

    def feed(self):
        """If possible, eat a sheep at current location."""
        sheep = [obj for obj in self.cell.agents if isinstance(obj, Sheep)]
        if sheep:  # If there are any sheep present
            sheep_to_eat = self.random.choice(sheep)
            self.energy += self.energy_from_food
            sheep_to_eat.remove()

    def move(self):
        """Move to a neighboring cell, preferably one with sheep."""
        cells_with_sheep = self.cell.neighborhood.select(
            lambda cell: any(isinstance(obj, Sheep) for obj in cell.agents)
        )
        target_cells = (
            cells_with_sheep if len(cells_with_sheep) > 0 else self.cell.neighborhood
        )
        self.cell = target_cells.select_random_cell()


class GrassPatch(FixedAgent):
    """A patch of grass that grows at a fixed rate and can be eaten by sheep."""

    @property
    def fully_grown(self):
        """Whether the grass patch is fully grown."""
        return self._fully_grown

    @fully_grown.setter
    def fully_grown(self, value: bool) -> None:
        """Set grass growth state and schedule regrowth if eaten."""
        self._fully_grown = value

        if not value:  # If grass was just eaten
            self.model.simulator.schedule_event_relative(
                setattr,
                self.grass_regrowth_time,
                function_args=[self, "fully_grown", True],
            )

    def __init__(self, model, countdown, grass_regrowth_time, cell):
        """Create a new patch of grass.

        Args:
            model: Model instance
            countdown: Time until grass is fully grown again
            grass_regrowth_time: Time needed to regrow after being eaten
            cell: Cell to which this grass patch belongs
        """
        super().__init__(model)
        self._fully_grown = countdown == 0
        self.grass_regrowth_time = grass_regrowth_time
        self.cell = cell

        # Schedule initial growth if not fully grown
        if not self.fully_grown:
            self.model.simulator.schedule_event_relative(
                setattr, countdown, function_args=[self, "fully_grown", True]
            )

Model#

"""
Wolf-Sheep Predation Model
================================

Replication of the model found in NetLogo:
    Wilensky, U. (1997). NetLogo Wolf Sheep Predation model.
    http://ccl.northwestern.edu/netlogo/models/WolfSheepPredation.
    Center for Connected Learning and Computer-Based Modeling,
    Northwestern University, Evanston, IL.
"""

import math

from mesa import Model
from mesa.datacollection import DataCollector
from mesa.examples.advanced.wolf_sheep.agents import GrassPatch, Sheep, Wolf
from mesa.experimental.cell_space import OrthogonalVonNeumannGrid
from mesa.experimental.devs import ABMSimulator


class WolfSheep(Model):
    """Wolf-Sheep Predation Model.

    A model for simulating wolf and sheep (predator-prey) ecosystem modelling.
    """

    description = (
        "A model for simulating wolf and sheep (predator-prey) ecosystem modelling."
    )

    def __init__(
        self,
        width=20,
        height=20,
        initial_sheep=100,
        initial_wolves=50,
        sheep_reproduce=0.04,
        wolf_reproduce=0.05,
        wolf_gain_from_food=20,
        grass=True,
        grass_regrowth_time=30,
        sheep_gain_from_food=4,
        seed=None,
        simulator: ABMSimulator = None,
    ):
        """Create a new Wolf-Sheep model with the given parameters.

        Args:
            height: Height of the grid
            width: Width of the grid
            initial_sheep: Number of sheep to start with
            initial_wolves: Number of wolves to start with
            sheep_reproduce: Probability of each sheep reproducing each step
            wolf_reproduce: Probability of each wolf reproducing each step
            wolf_gain_from_food: Energy a wolf gains from eating a sheep
            grass: Whether to have the sheep eat grass for energy
            grass_regrowth_time: How long it takes for a grass patch to regrow
                                once it is eaten
            sheep_gain_from_food: Energy sheep gain from grass, if enabled
            seed: Random seed
            simulator: ABMSimulator instance for event scheduling
        """
        super().__init__(seed=seed)
        self.simulator = simulator
        self.simulator.setup(self)

        # Initialize model parameters
        self.height = height
        self.width = width
        self.grass = grass

        # Create grid using experimental cell space
        self.grid = OrthogonalVonNeumannGrid(
            [self.height, self.width],
            torus=True,
            capacity=math.inf,
            random=self.random,
        )

        # Set up data collection
        model_reporters = {
            "Wolves": lambda m: len(m.agents_by_type[Wolf]),
            "Sheep": lambda m: len(m.agents_by_type[Sheep]),
        }
        if grass:
            model_reporters["Grass"] = lambda m: len(
                m.agents_by_type[GrassPatch].select(lambda a: a.fully_grown)
            )

        self.datacollector = DataCollector(model_reporters)

        # Create sheep:
        Sheep.create_agents(
            self,
            initial_sheep,
            energy=self.rng.random((initial_sheep,)) * 2 * sheep_gain_from_food,
            p_reproduce=sheep_reproduce,
            energy_from_food=sheep_gain_from_food,
            cell=self.random.choices(self.grid.all_cells.cells, k=initial_sheep),
        )
        # Create Wolves:
        Wolf.create_agents(
            self,
            initial_wolves,
            energy=self.rng.random((initial_wolves,)) * 2 * wolf_gain_from_food,
            p_reproduce=wolf_reproduce,
            energy_from_food=wolf_gain_from_food,
            cell=self.random.choices(self.grid.all_cells.cells, k=initial_wolves),
        )

        # Create grass patches if enabled
        if grass:
            possibly_fully_grown = [True, False]
            for cell in self.grid:
                fully_grown = self.random.choice(possibly_fully_grown)
                countdown = (
                    0 if fully_grown else self.random.randrange(0, grass_regrowth_time)
                )
                GrassPatch(self, countdown, grass_regrowth_time, cell)

        # Collect initial data
        self.running = True
        self.datacollector.collect(self)

    def step(self):
        """Execute one step of the model."""
        # First activate all sheep, then all wolves, both in random order
        self.agents_by_type[Sheep].shuffle_do("step")
        self.agents_by_type[Wolf].shuffle_do("step")

        # Collect data
        self.datacollector.collect(self)

App#

from mesa.examples.advanced.wolf_sheep.agents import GrassPatch, Sheep, Wolf
from mesa.examples.advanced.wolf_sheep.model import WolfSheep
from mesa.experimental.devs import ABMSimulator
from mesa.visualization import (
    Slider,
    SolaraViz,
    make_plot_component,
    make_space_component,
)


def wolf_sheep_portrayal(agent):
    if agent is None:
        return

    portrayal = {
        "size": 25,
    }

    if isinstance(agent, Wolf):
        portrayal["color"] = "tab:red"
        portrayal["marker"] = "o"
        portrayal["zorder"] = 2
    elif isinstance(agent, Sheep):
        portrayal["color"] = "tab:cyan"
        portrayal["marker"] = "o"
        portrayal["zorder"] = 2
    elif isinstance(agent, GrassPatch):
        if agent.fully_grown:
            portrayal["color"] = "tab:green"
        else:
            portrayal["color"] = "tab:brown"
        portrayal["marker"] = "s"
        portrayal["size"] = 75

    return portrayal


model_params = {
    "seed": {
        "type": "InputText",
        "value": 42,
        "label": "Random Seed",
    },
    "grass": {
        "type": "Select",
        "value": True,
        "values": [True, False],
        "label": "grass regrowth enabled?",
    },
    "grass_regrowth_time": Slider("Grass Regrowth Time", 20, 1, 50),
    "initial_sheep": Slider("Initial Sheep Population", 100, 10, 300),
    "sheep_reproduce": Slider("Sheep Reproduction Rate", 0.04, 0.01, 1.0, 0.01),
    "initial_wolves": Slider("Initial Wolf Population", 10, 5, 100),
    "wolf_reproduce": Slider(
        "Wolf Reproduction Rate",
        0.05,
        0.01,
        1.0,
        0.01,
    ),
    "wolf_gain_from_food": Slider("Wolf Gain From Food Rate", 20, 1, 50),
    "sheep_gain_from_food": Slider("Sheep Gain From Food", 4, 1, 10),
}


def post_process_space(ax):
    ax.set_aspect("equal")
    ax.set_xticks([])
    ax.set_yticks([])


def post_process_lines(ax):
    ax.legend(loc="center left", bbox_to_anchor=(1, 0.9))


space_component = make_space_component(
    wolf_sheep_portrayal, draw_grid=False, post_process=post_process_space
)
lineplot_component = make_plot_component(
    {"Wolves": "tab:orange", "Sheep": "tab:cyan", "Grass": "tab:green"},
    post_process=post_process_lines,
)

simulator = ABMSimulator()
model = WolfSheep(simulator=simulator, grass=True)

page = SolaraViz(
    model,
    components=[space_component, lineplot_component],
    model_params=model_params,
    name="Wolf Sheep",
    simulator=simulator,
)
page  # noqa