Module moog.action_spaces.set_position

SetPosition action space for controlling agent avatars.

This action space sets the position of an agent. This is commonly used with a touch-screen or mouse-click interface.

"""SetPosition action space for controlling agent avatars.

This action space sets the position of an agent. This is commonly used with a
touch-screen or mouse-click interface.
"""

from . import abstract_action_space
from dm_env import specs
import numpy as np


class SetPosition(abstract_action_space.AbstractActionSpace):
    """SetPosition action space."""

    def __init__(self, action_layers='agent', inertia=0.):
        """Constructor.
        
        Args:
            agent_layer: String or iterable of strings. Elements (or itself if
                string) must be keys in the environment state. All sprites in
                these layers will be acted upon by this action space.
            inertia: Float in [0, 1]. Inertia of the action layer sprite
                positions. Zero means position is set instantly to action value,
                and 1 means the action has no effect on sprite positions.
        """
        if not isinstance(action_layers, (list, tuple)):
            action_layers = (action_layers,)
        self._action_layers = action_layers
        self._inertia = inertia

        self._action_spec = specs.BoundedArray(
            shape=(2,), dtype=np.float32, minimum=0, maximum=1)

    def step(self, state, action):
        """Apply action to environment state.

        Args:
            state: Ordereddict. Environment state.
            action: Numpy float array of size (2), in [0, 1]. Position to set
                for the agent(s) in self._action_layers.
        """
        for action_layer in self._action_layers:
            for sprite in state[action_layer]:
                sprite.position = (
                    self._inertia * sprite.position +
                    (1 - self._inertia) * action
                )

    def random_action(self):
        """Return randomly sampled action."""
        return np.random.uniform(0., 1., size=(2,))
            
    def action_spec(self):
        return self._action_spec

Classes

class SetPosition (action_layers='agent', inertia=0.0)

SetPosition action space.

Constructor.

Args

agent_layer

String or iterable of strings. Elements (or itself if string) must be keys in the environment state. All sprites in these layers will be acted upon by this action space.

inertia

Float in [0, 1]. Inertia of the action layer sprite positions. Zero means position is set instantly to action value, and 1 means the action has no effect on sprite positions.

Expand source code

class SetPosition(abstract_action_space.AbstractActionSpace):
    """SetPosition action space."""

    def __init__(self, action_layers='agent', inertia=0.):
        """Constructor.
        
        Args:
            agent_layer: String or iterable of strings. Elements (or itself if
                string) must be keys in the environment state. All sprites in
                these layers will be acted upon by this action space.
            inertia: Float in [0, 1]. Inertia of the action layer sprite
                positions. Zero means position is set instantly to action value,
                and 1 means the action has no effect on sprite positions.
        """
        if not isinstance(action_layers, (list, tuple)):
            action_layers = (action_layers,)
        self._action_layers = action_layers
        self._inertia = inertia

        self._action_spec = specs.BoundedArray(
            shape=(2,), dtype=np.float32, minimum=0, maximum=1)

    def step(self, state, action):
        """Apply action to environment state.

        Args:
            state: Ordereddict. Environment state.
            action: Numpy float array of size (2), in [0, 1]. Position to set
                for the agent(s) in self._action_layers.
        """
        for action_layer in self._action_layers:
            for sprite in state[action_layer]:
                sprite.position = (
                    self._inertia * sprite.position +
                    (1 - self._inertia) * action
                )

    def random_action(self):
        """Return randomly sampled action."""
        return np.random.uniform(0., 1., size=(2,))
            
    def action_spec(self):
        return self._action_spec

Ancestors

• AbstractActionSpace
• abc.ABC

Methods

def random_action(self)

Return randomly sampled action.

Expand source code

def random_action(self):
    """Return randomly sampled action."""
    return np.random.uniform(0., 1., size=(2,))

def step(self, state, action)

Apply action to environment state.

Args

state

Ordereddict. Environment state.

action

Numpy float array of size (2), in [0, 1]. Position to set for the agent(s) in self._action_layers.

Expand source code

def step(self, state, action):
    """Apply action to environment state.

    Args:
        state: Ordereddict. Environment state.
        action: Numpy float array of size (2), in [0, 1]. Position to set
            for the agent(s) in self._action_layers.
    """
    for action_layer in self._action_layers:
        for sprite in state[action_layer]:
            sprite.position = (
                self._inertia * sprite.position +
                (1 - self._inertia) * action
            )

Inherited members

• AbstractActionSpace:
  • action_spec
  • reset