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

How to Run#

Install Mesa with recommended dependencies:

pip install “mesa[rec]”

Then run the example:

solara run app.py

Open the displayed local URL in your browser.

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.

Further Reading#

This model is closely based on the NetLogo Wolf-Sheep Predation Model:

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.

See also the Lotka–Volterra equations for an example of a classic differential-equation model with similar dynamics.

Agents#

from mesa.discrete_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)), None
        )
        if grass_patch is not None and grass_patch.fully_grown:
            self.energy += self.energy_from_food
            grass_patch.get_eaten()

    def move(self):
        """Move towards a cell where there isn't a wolf, and preferably with grown grass."""
        cells_without_wolves = []
        cells_with_grass = []

        for cell in self.cell.neighborhood:
            has_wolf = False
            has_grass = False

            for obj in cell.agents:
                # If there's a wolf, we can early exit
                if isinstance(obj, Wolf):
                    has_wolf = True
                    break
                elif isinstance(obj, GrassPatch) and obj.fully_grown:
                    has_grass = True

            # Prefer cells without wolves
            if not has_wolf:
                cells_without_wolves.append(cell)

                # Among safe cells, pick those with grown grass
                if has_grass:
                    cells_with_grass.append(cell)

        # If all surrounding cells have wolves, stay put
        if len(cells_without_wolves) == 0:
            return

        # 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 = self.random.choice(target_cells)


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."""

    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.schedule_event(self.regrow, after=countdown)

    def regrow(self):
        """Regrow the grass."""
        self.fully_grown = True

    def get_eaten(self):
        """Mark grass as eaten and schedule regrowth."""
        self.fully_grown = False
        self.model.schedule_event(self.regrow, after=self.grass_regrowth_time)

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.discrete_space import OrthogonalVonNeumannGrid
from mesa.examples.advanced.wolf_sheep.agents import GrassPatch, Sheep, Wolf
from mesa.experimental.scenarios import Scenario


class WolfSheepScenario(Scenario):
    """Scenario parameters for the Wolf-Sheep model.

    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
        rng: Random rng
    """

    width: int = 20
    height: int = 20
    initial_sheep: int = 100
    initial_wolves: int = 50
    sheep_reproduce: float = 0.04
    wolf_reproduce: float = 0.05
    wolf_gain_from_food: float = 20.0
    grass: bool = True
    grass_regrowth_time: int = 30
    sheep_gain_from_food: float = 4.0


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, scenario: WolfSheepScenario = WolfSheepScenario):
        """Create a new Wolf-Sheep model with the given parameters.

        Args:
            scenario: WolfSheepScenario containing model parameters.
        """
        super().__init__(scenario=scenario)

        # Initialize model parameters
        self.height = scenario.height
        self.width = scenario.width
        self.grass = scenario.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 self.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,
            scenario.initial_sheep,
            energy=self.rng.random((scenario.initial_sheep,))
            * 2
            * scenario.sheep_gain_from_food,
            p_reproduce=scenario.sheep_reproduce,
            energy_from_food=scenario.sheep_gain_from_food,
            cell=self.random.choices(
                self.grid.all_cells.cells, k=scenario.initial_sheep
            ),
        )
        # Create Wolves:
        Wolf.create_agents(
            self,
            scenario.initial_wolves,
            energy=self.rng.random((scenario.initial_wolves,))
            * 2
            * scenario.wolf_gain_from_food,
            p_reproduce=scenario.wolf_reproduce,
            energy_from_food=scenario.wolf_gain_from_food,
            cell=self.random.choices(
                self.grid.all_cells.cells, k=scenario.initial_wolves
            ),
        )

        # Create grass patches if enabled
        if self.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, scenario.grass_regrowth_time)
                )
                GrassPatch(self, countdown, scenario.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, WolfSheepScenario
from mesa.visualization import (
    CommandConsole,
    Slider,
    SolaraViz,
    SpaceRenderer,
    make_plot_component,
)
from mesa.visualization.components import AgentPortrayalStyle


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

    portrayal = AgentPortrayalStyle(
        size=50,
        marker="o",
        zorder=2,
    )

    if isinstance(agent, Wolf):
        portrayal.update(("color", "red"))
    elif isinstance(agent, Sheep):
        portrayal.update(("color", "cyan"))
    elif isinstance(agent, GrassPatch):
        if agent.fully_grown:
            portrayal.update(("color", "tab:green"))
        else:
            portrayal.update(("color", "tab:brown"))
        portrayal.update(("marker", "s"), ("size", 125), ("zorder", 1))

    return portrayal


model_params = {
    "rng": {
        "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))


lineplot_component = make_plot_component(
    {"Wolves": "tab:orange", "Sheep": "tab:cyan", "Grass": "tab:green"},
    post_process=post_process_lines,
)

model = WolfSheep(scenario=WolfSheepScenario(grass=True))

renderer = SpaceRenderer(
    model,
    backend="matplotlib",
).setup_agents(wolf_sheep_portrayal)
renderer.post_process = post_process_space
renderer.draw_agents()

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