aprel.basics package¶

aprel.basics.environment module¶

Environment-related modules.

class aprel.basics.environment.Environment(env: gym.core.Env, feature_func: Callable)¶

Bases: object

This is a wrapper around an OpenAI Gym environment, so that we can store the features function along with the environment itself.

Parameters

env (gym.Env) – An OpenAi Gym environment.
features (Callable) – Given a Trajectory, this function must return a numpy.array of features.

env¶

The wrapped environment.

Type: gym.Env

features¶

Features function.

Type: Callable

action_space¶: Inherits from env.

observation_space¶: Inherits from env.

reset¶

Inherits from env.

Type: Callable

step¶

Inherits from env.

Type: Callable

render¶

Inherits from env, if it exists; None otherwise.

Type: Callable

render_exists¶

True if render exists.

Type: bool

close¶

Inherits from env, if it exists; None otherwise.

Type: Callable

close_exists¶

True if close exists.

Type: bool

aprel.basics.trajectory module¶

Modules that are related to environment trajectories.

class aprel.basics.trajectory.Trajectory(env: aprel.basics.environment.Environment, trajectory: List[Tuple[numpy.array, numpy.array]], clip_path: Optional[str] = None)¶

Bases: object

A class for keeping trajectories that consist of a sequence of state-action pairs, the features and a clip path that keeps a video visualization of the trajectory.

This class supports indexing, such that t^th index returns the state-action pair at time step t. However, indices cannot be assigned, i.e., a specific state-action pair cannot be changed, because that would enable infeasible trajectories.

Parameters

env (Environment) – The environment object that generated this trajectory.
trajectory (List[Tuple[numpy.array, numpy.array]]) – The sequence of state-action pairs.
clip_path (str) – The path to the video clip that keeps the visualization of the trajectory.

trajectory¶

The sequence of state-action pairs.

Type: List[Tuple[numpy.array, numpy.array]]

features¶

Features of the trajectory.

Type: numpy.array

clip_path¶

The path to the video clip that keeps the visualization of the trajectory.

Type: str

property length: int¶: The length of the trajectory, i.e., the number of time steps in the trajectory.

visualize()¶

Visualizes the trajectory with a video if the clip exists. Otherwise, prints the trajectory information.

Note: FPS is fixed at 25 for video visualizations.

class aprel.basics.trajectory.TrajectorySet(trajectories: List[aprel.basics.trajectory.Trajectory])¶

Bases: object

A class for keeping a set of trajectories, i.e. Trajectory objects.

This class supports indexing, such that t^th index returns the t^th trajectory in the set. Similarly, t^th trajectory in the set can be replaced with a new trajectory using indexing. Only for reading trajectories with indexing, list indices are also allowed.

Parameters: trajectories (List[Trajectory]) – The list of trajectories to be stored in the set.

trajectories¶

The list of trajectories in the set.

Type: List[Trajectory]

features_matrix¶

n x d array of features where each row consists of the d features of the corresponding trajectory.

Type: numpy.array

append(new_trajectory: aprel.basics.trajectory.Trajectory)¶: Appends a new trajectory to the set.

property size: int¶: The number of trajectories in the set.