rlberry_scool.envs.GridWorld

class rlberry_scool.envs.GridWorld(nrows=5, ncols=5, start_coord=(0, 0), terminal_states=None, success_probability=0.9, reward_at=None, walls=((1, 1), (2, 2)), default_reward=0.0)

Bases: RenderInterface2D, FiniteMDP

Simple GridWorld environment.

Parameters:

nrowsint

number of rows

ncolsint

number of columns

start_coordtuple

tuple with coordinates of initial position

terminal_statestuple

((row_0, col_0), (row_1, col_1), …) = coordinates of terminal states

success_probabilitydouble

probability of moving in the chosen direction

reward_at: dict

dictionary, keys = tuple containing coordinates, values = reward at each coordinate

wallstuple

((row_0, col_0), (row_1, col_1), …) = coordinates of walls

default_rewarddouble

reward received at states not in ‘reward_at’

Attributes:

np_random

Returns the environment’s internal _np_random that if not set will initialise with a random seed.

render_mode

rng

Random number generator.

spec

unwrapped

Returns the base non-wrapped environment.

Methods

close()	After the user has finished using the environment, close contains the code necessary to "clean up" the environment.
from_layout([layout, success_probability])	Create GridWorld instance from a layout.
get_background()	Return a scene (list of shapes) representing the background
get_layout_array([state_data, fill_walls_with])	Returns an array 'layout' of shape (nrows, ncols) such that:
get_layout_img([state_data, colormap_name, ...])	Returns an image array representing the value of state_data on the gridworld layout.
get_params([deep])	Get parameters for this model.
get_scene(state)	Return scene (list of shapes) representing a given state
get_video([framerate])	Get video data.
get_wrapper_attr(name)	Gets the attribute name from the environment.
is_generative()	Returns true if sample() method is implemented
is_online()	Returns true if reset() and step() methods are implemented
is_terminal(state)	Returns true if a state is terminal.
log()	Print the structure of the MDP.
render([loop])	Function to render an environment that implements the interface.
reseed([seed_seq])	Get new random number generator for the model.
reset([seed, options])	Reset the environment to a default state.
reward_fn(state, action, next_state)	Reward function.
sample(state, action)	Sample a transition s' from P(s'|state, action).
save_video(filename[, framerate])	Save video file.
set_initial_state_distribution(distribution)	Parameters:
step(action)	Run one timestep of the environment's dynamics using the agent actions.

append_state_for_rendering
clear_render_buffer
disable_rendering
display_values
enable_rendering
get_renderer
get_transition_support
is_render_enabled
print_transition_at
render_ascii
save_gif
set_clipping_area
set_refresh_interval

close()

After the user has finished using the environment, close contains the code necessary to “clean up” the environment.

This is critical for closing rendering windows, database or HTTP connections. Calling close on an already closed environment has no effect and won’t raise an error.

classmethod from_layout(layout: str = '\nIOOOO # OOOOO  O OOOOR\nOOOOO # OOOOO  # OOOOO\nOOOOO O OOOOO  # OOOOO\nOOOOO # OOOOO  # OOOOO\nIOOOO # OOOOO  # OOOOr\n', success_probability=0.95)

Create GridWorld instance from a layout.

Layout symbols:

‘#’ : wall ‘r’ : reward of 1, terminal state ‘R’ : reward of 1, non-terminal state ‘T’ : terminal state ‘I’ : initial state (if several, start uniformly among I) ‘O’ : empty state any other char : empty state

Layout example:

IOOOO # OOOOO O OOOOR OOOOO # OOOOO # OOOOO OOOOO O OOOOO # OOOOO OOOOO # OOOOO # OOOOO IOOOO # OOOOO # OOOOr

get_background()

Return a scene (list of shapes) representing the background

get_layout_array(state_data=None, fill_walls_with=nan)

Returns an array ‘layout’ of shape (nrows, ncols) such that:

layout[row, col] = state_data[self.coord2idx[row, col]]

If (row, col) is a wall:

layout[row, col] = fill_walls_with

Parameters:

state_datanp.array, default = None

Array of shape (self.observation_space.n,)

fill_walls_withfloat, default: np.nan

Value to set in the layout in the coordinates corresponding to walls.

Returns:

Gridworld layout array of shape (nrows, ncols).

get_layout_img(state_data=None, colormap_name='cool', wall_color=(0.0, 0.0, 0.0))

Returns an image array representing the value of state_data on the gridworld layout.

Parameters:

state_datanp.array, default = None

Array of shape (self.observation_space.n,)

colormap_namestr, default = ‘cool’

Colormap name. See https://matplotlib.org/tutorials/colors/colormaps.html

wall_colortuple

RGB color for walls.

Returns

——-

Gridworld image array of shape (nrows, ncols, 3).

get_params(deep=True)

Get parameters for this model.

Parameters:

deepbool, default=True

If True, will return the parameters for this model and contained subobjects.

Returns:

paramsdict

Parameter names mapped to their values.

get_scene(state)

Return scene (list of shapes) representing a given state

get_video(framerate=25, **kwargs)

Get video data.

get_wrapper_attr(name: str) → Any

Gets the attribute name from the environment.

is_generative()

Returns true if sample() method is implemented

is_online()

Returns true if reset() and step() methods are implemented

is_terminal(state)

Returns true if a state is terminal.

log()

Print the structure of the MDP.

property np_random: Generator

Returns the environment’s internal _np_random that if not set will initialise with a random seed.

Returns:

Instances of np.random.Generator

render(loop=True, **kwargs)

Function to render an environment that implements the interface.

reseed(seed_seq=None)

Get new random number generator for the model.

Parameters:

seed_seqnp.random.SeedSequence, rlberry.seeding.Seeder or int, defaultNone

Seed sequence from which to spawn the random number generator. If None, generate random seed. If int, use as entropy for SeedSequence. If seeder, use seeder.seed_seq

reset(seed=None, options=None)

Reset the environment to a default state.

reward_fn(state, action, next_state)

Reward function. Returns mean reward at (state, action) by default.

Parameters:

stateint

current state

actionint

current action

next_state

next state

Returns:

reward : float

property rng

Random number generator.

sample(state, action)

Sample a transition s’ from P(s’|state, action).

save_video(filename, framerate=25, **kwargs)

Save video file.

set_initial_state_distribution(distribution)

Parameters:

distributionnumpy.ndarray or int

array of size (S,) containing the initial state distribution or an integer representing the initial/default state

step(action)

Run one timestep of the environment’s dynamics using the agent actions.

When the end of an episode is reached (terminated or truncated), it is necessary to call reset() to reset this environment’s state for the next episode.

Changed in version 0.26: The Step API was changed removing done in favor of terminated and truncated to make it clearer to users when the environment had terminated or truncated which is critical for reinforcement learning bootstrapping algorithms.

Args:

action (ActType): an action provided by the agent to update the environment state.

Returns:

observation (ObsType): An element of the environment’s observation_space as the next observation due to the agent actions.

An example is a numpy array containing the positions and velocities of the pole in CartPole.

reward (SupportsFloat): The reward as a result of taking the action. terminated (bool): Whether the agent reaches the terminal state (as defined under the MDP of the task)

which can be positive or negative. An example is reaching the goal state or moving into the lava from the Sutton and Barton, Gridworld. If true, the user needs to call reset().

truncated (bool): Whether the truncation condition outside the scope of the MDP is satisfied.

Typically, this is a timelimit, but could also be used to indicate an agent physically going out of bounds. Can be used to end the episode prematurely before a terminal state is reached. If true, the user needs to call reset().

info (dict): Contains auxiliary diagnostic information (helpful for debugging, learning, and logging).

This might, for instance, contain: metrics that describe the agent’s performance state, variables that are hidden from observations, or individual reward terms that are combined to produce the total reward. In OpenAI Gym <v26, it contains “TimeLimit.truncated” to distinguish truncation and termination, however this is deprecated in favour of returning terminated and truncated variables.

done (bool): (Deprecated) A boolean value for if the episode has ended, in which case further step() calls will

return undefined results. This was removed in OpenAI Gym v26 in favor of terminated and truncated attributes. A done signal may be emitted for different reasons: Maybe the task underlying the environment was solved successfully, a certain timelimit was exceeded, or the physics simulation has entered an invalid state.

property unwrapped

Returns the base non-wrapped environment.

Returns:

Env: The base non-wrapped gymnasium.Env instance