Backend

Classes for composing observations for a satellite.

class ObservationBuilder(satellite: Satellite, obs_type: type = <class 'numpy.ndarray'>, dtype: ~numpy.dtype = <class 'numpy.float64'>)[source]

Bases: Resetable

Satellite subclass for composing observations.

Parameters:

satellite (Satellite) – Satellite to observe
obs_type (type) – Datatype of satellite’s returned observation. Can be np.ndarray (default), dict, or list.
dtype (dtype) – Data type for observation np vectors.

reset_overwrite_previous() → None[source]

Perform any once-per-episode setup.

Return type:: None

reset_pre_sim_init() → None[source]

Perform any once-per-episode setup.

Return type:: None

reset_during_sim_init() → None[source]

Perform any once-per-episode setup.

Return type:: None

reset_post_sim_init() → None[source]

Perform any once-per-episode setup.

Return type:: None

obs_dict() → dict[str, Any][source]

Human-readable observation format.

Cached so only computed once per timestep.

Return type:: dict[str, Any]

obs_ndarray() → ndarray[source]

Numpy vector observation format.

Return type:: ndarray

obs_array_keys() → list[str][source]

Get the keys of the obs_ndarray.

Return type:: list[str]

obs_list() → list[source]

List observation format.

Return type:: list

get_obs() → dict | ndarray | list[source]

Update the observation.

Return type:: dict | ndarray | list

property observation_space: Space: Space of the observation.

property observation_description: Any: Human-interpretable description of observation space.