Usage

There are multiple ways to use Sadie.

Using pre-defined objects

Sadie comes fully loaded with a range of objects that can provide for a wide range of use cases. These can be directly instantiated and used.

• Agent objects are basic elements of agent-based models.
• Walker objects are agents that implement various random walk patterns, including having their own state and conditionality.
• Forager objects are special walkers that alter their behaviour as a function of their environment.

Creating your own objects

Changing azimuth and distance distributions

To create your own objects, subclass the closest base class and add your own desired functionality. Consider, for instance, the way the Uniform-Lévy walker is constructed:

class UniformLevyRandomWalker(BaseWalker):
    def retarget(self):
        self.set_polar_target(np.random.uniform(0, 2*π), stats.levy().rvs())

Simply overriding the retarget() method modifies the targeting behaviour of a walker – you can, for instance, pick a different distribution or set parameters to the distribution.

Changing wait/move behaviour

To change wait/move behaviour, e.g. to halt the object, consider overriding its update() method:

class WaitingUniformWalker(BaseWalker):
    def __init__(self, x_init: float, y_init: float, velocity: float = 1.0, wait_transition_probability: float = 0.4):
        super(WaitingUniformWalker, self).__init__(x_init=x_init, y_init=y_init, velocity=velocity)
        self.wait_transition_probability = wait_transition_probability

    def update(self):
        if self.target == (None, None):
            self.retarget()
        elif self.is_on_target:
            if np.random.random() > self.wait_transition_probability:
                self.retarget()
            else:
                self.wait()
        else:
            self.move()

Here, we override the __init__() method first, adding a new parameter to it to represent the wait transition probability. We assign this to a field, then, overriding the update() method, we provide a checkpoint: once an agent is on target and due for re-targeting, we insert a random value condition whereby a random float must exceed the pre-defined wait_transition_probability.

Creating new trackable values

To track other values, e.g. energy, you can create new class attributes by subclassing your Agent or Walker, then creating a new entry after invoking the super() call:

class EnergyTrackingWalker(BaseWalker):
    def __init__(self, x_init: float, y_init: float, velocity: float = 1.0):
        super(WaitingUniformWalker, self).__init__(x_init=x_init, y_init=y_init, velocity=velocity)
        self.energy: float = 0.0