envs.EvoGymBase

Extend this class for creating gym-like environments (inherits from gym.Env).

Quicklinks: EvoWorld, EvoSim, EvoViewer, WorldObject, envs.EvoGymBase, helper functions and metadata

Page Outline

class EvoGymBase

`class` `EvoGymBase`

Base class for all Evolution Gym environments.

Args:

world (EvoWorld): object specifying the voxel layout of the environment.

`function` `init`

__init__(world: EvoWorld) → None

`property` default_viewer

Returns the environment’s default viewer.

Returns:

EvoSim: viewer object to return.

`property` sim

Returns the environment’s simulation.

Returns:

EvoSim: simulation object to return.

`property` unwrapped

Completely unwrap this env.

Returns:

gym.Env: The base non-wrapped gym.Env instance

`function` `close`

close() → None

Close the simulation.

`function` `get_actuator_indices`

get_actuator_indices(robot_name: str) → ndarray

Returns the voxel indices a target robot’s actuators in the environment’s simulation.

Args:

robot_name (str): name of robot.

Returns:

np.ndarray: (n,) array of actuator indices, where n is the number of actuators.

`function` `get_ceil_obs`

get_ceil_obs(
    object_name: str,
    terrain_list: list,
    sight_dist: int,
    sight_range: float = 5
) → ndarray

Observation helper-function. Computes an observation describing the shape of the terrain above the target object. Specifically, for each voxel to the left and right of the target object’s center of mass (along with the voxel containing the center of mass), the following observation is computed: min(y-distance in voxels to the nearest terrain object above the target object’s center of mass, sight_range). Results are returned in a 1D numpy array.

Args:

object_name (str): name of target object.
terrain_list (list[str]): names of objects to be considered terrain in the computation.
sight_dist (int): number of voxels to the left and right of the target object’s center of mass for which an observation should be returned.
sight_range (float): the max number of voxels above the object that can be seen. (default = 5)

Returns:

np.ndarray: (2 * sight_range + 1, ) array of distance observations.

`function` `get_dim_action_space`

get_dim_action_space(robot_name: str) → int

Returns the number of actuators for a target robot in the environment’s simulation.

Args:

robot_name (str): name of robot.

Returns:

int: number of actuators.

`function` `get_floor_obs`

get_floor_obs(
    object_name: str,
    terrain_list: list,
    sight_dist: int,
    sight_range: float = 5
) → ndarray

Observation helper-function. Computes an observation describing the shape of the terrain below the target object. Specifically, for each voxel to the left and right of the target object’s center of mass (along with the voxel containing the center of mass), the following observation is computed: min(y-distance in voxels to the nearest terrain object below the target object’s center of mass, sight_range). Results are returned in a 1D numpy array.

Args:

object_name (str): name of target object.
terrain_list (list[str]): names of objects to be considered terrain in the computation.
sight_dist (int): number of voxels to the left and right of the target object’s center of mass for which an observation should be returned.
sight_range (float): the max number of voxels below the object that can be seen. (default = 5)

Returns:

np.ndarray: (2 * sight_range + 1, ) array of distance observations.

`function` `get_ort_obs`

get_ort_obs(object_name: str)

Observation helper-function. Returns the orientation of a target object.

Args:

object_name (str): name of object

Returns:

np.ndarray: (1,) array of the object’s orientation.

`function` `get_pos_com_obs`

get_pos_com_obs(object_name: str) → ndarray

Observation helper-function. Computes the position of the center of mass of a target object by averaging the positions of the object’s point masses.

Args:

object_name (str): name of object

Returns:

np.ndarray: (2,) array of the position of the center of mass.

`function` `get_relative_pos_obs`

get_relative_pos_obs(object_name: str)

Observation helper-function. Computes the positions of a target object’s point masses relative to their center of mass.

Args:

object_name (str): name of object

Returns:

np.ndarray: (2n,) array of positions, where n is the number of point masses.

`function` `get_time`

get_time() → int

Returns the current time as defined in the environment’s simulator. Time starts at 0 and is incremented each time the environment steps. Time resets to 0 when the environment is reset.

Returns:

int: the current time.

`function` `get_vel_com_obs`

get_vel_com_obs(object_name: str) → ndarray

Observation helper-function. Computes the velocity of the center of mass of a target object by averaging the velocities of the object’s point masses.

Args:

object_name (str): name of object

Returns:

np.ndarray: (2,) array of the velocity of the center of mass.

`function` `object_orientation_at_time`

object_orientation_at_time(time: int, object_name: str) → float

Returns an estimate of the orientation of an object at time time. Use EvoGymBase.get_time() to get current measurements.

Args:

time (int): time at which to return measurement.
object_name (str): name of object

Returns:

float: orientation with respect to x-axis in radians (increasing counter-clockwise) from the range [0, 2π].

`function` `object_pos_at_time`

object_pos_at_time(time: int, object_name: str) → ndarray

Returns positions of all point-masses in a target object at time time. Use EvoGymBase.get_time() to get current measurements.

Args:

time (int): time at which to return measurements.
object_name (str): name of object

Returns:

np.ndarray: (2, n) array of measurements, where n is the number of point-masses in the target object.

`function` `object_vel_at_time`

object_vel_at_time(time: int, object_name: str) → ndarray

Returns velocities of all point-masses in a target object at time time. Use EvoGymBase.get_time() to get current measurements.

Args:

time (int): time at which to return measurements.
object_name (str): name of object

Returns:

np.ndarray: (2, n) array of measurements, where n is the number of point-masses in the target object.

`function` `pos_at_time`

pos_at_time(time: int) → ndarray

Returns positions of all point-masses in the environment’s simulation at time time. Use EvoGymBase.get_time() to get current measurements.

Args:

time (int): time at which to return measurements.

Returns:

np.ndarray: (2, n) array of measurements, where n is the number of points in the environment’s simulation.

`function` `render`

render(
    mode: str = 'screen',
    verbose: bool = False,
    hide_background: bool = False,
    hide_grid: bool = False,
    hide_edges: bool = False,
    hide_voxels: bool = False
) → Optional[ndarray]

Render the simulation.

Args:

mode (str): values of ‘screen’ and ‘human’ will render to a debug window. If set to ‘img’ will return an image array.
verbose (bool): whether or not to print the rendering speed (rps) every second.
hide_background (bool): whether or not to render the cream-colored background. If shut off background will be white.
hide_grid (bool): whether or not to render the grid.
hide_edges (bool): whether or not to render edges around all objects.
hide_voxels (bool): whether or not to render voxels.

Returns:

Optional[np.ndarray]: if mode is set to img, will return an image array.

`function` `reset`

reset() → None

Reset the simulation to the initial state.

`function` `step`

step(action: Dict[str, ndarray]) → bool

Step the environment by running physcis computations.

Args:

action (Dict[str, np.ndarray]): dictionary mapping robot names to actions. Actions are (n,) arrays, where n is the number of actuators in the target robot.

Returns:

bool: whether or not the simulation has reached an unstable state and cannot be recovered (True = unstable).

`function` `vel_at_time`

vel_at_time(time: int) → ndarray

Returns velocities of all point-masses in the environment’s simulation at time time. Use EvoGymBase.get_time() to get current measurements.

Args:

time (int): time at which to return measurements.

Returns:

np.ndarray: (2, n) array of measurements, where n is the number of points in the environment’s simulation.

This file was automatically generated via lazydocs.

envs.EvoGymBase

Page Outline

class EvoGymBase

function __init__

property default_viewer

property sim

property unwrapped

function close

function get_actuator_indices

function get_ceil_obs

function get_dim_action_space

function get_floor_obs

function get_ort_obs

function get_pos_com_obs

function get_relative_pos_obs

function get_time

function get_vel_com_obs

function object_orientation_at_time

function object_pos_at_time

function object_vel_at_time

function pos_at_time

function render

function reset

function step

function vel_at_time

`class` `EvoGymBase`

`function` `init`

`property` default_viewer

`property` sim

`property` unwrapped

`function` `close`

`function` `get_actuator_indices`

`function` `get_ceil_obs`

`function` `get_dim_action_space`

`function` `get_floor_obs`

`function` `get_ort_obs`

`function` `get_pos_com_obs`

`function` `get_relative_pos_obs`

`function` `get_time`

`function` `get_vel_com_obs`

`function` `object_orientation_at_time`

`function` `object_pos_at_time`

`function` `object_vel_at_time`

`function` `pos_at_time`

`function` `render`

`function` `reset`

`function` `step`

`function` `vel_at_time`