gym_acnportal.gym_acnsim.envs.observation

Module containing definition of a gym_acnsim observation and factory functions for different builtin observations.

See the SimObservation docstring for more information on the SimObservation class.

Each factory function takes no arguments and returns an instance of type SimObservation. Each factory function defines a space_function and and an obs_function with the following signatures:

space_function: Callable[[GymInterface], spaces.Space] obs_function: Callable[[GymInterface], np.ndarray]

The space_function gives a gym space for a given observation type. The obs_function gives a gym observation for a given observation type. The observation returned by obs_function is a point in the space returned by space_function.

Module Contents

Classes

SimObservation

Class representing an OpenAI Gym observation of an ACN-Sim

Functions

_ev_observation(attribute_function: Callable[([GymTrainedInterface, EV], float)], name: str) → gym_acnportal.gym_acnsim.envs.observation.SimObservation
arrival_observation() → gym_acnportal.gym_acnsim.envs.observation.SimObservation	Generates a SimObservation instance that wraps functions to
departure_observation() → gym_acnportal.gym_acnsim.envs.observation.SimObservation	Generates a SimObservation instance that wraps functions to
remaining_demand_observation() → gym_acnportal.gym_acnsim.envs.observation.SimObservation	Generates a SimObservation instance that wraps functions to
_constraints_observation(attribute: str, name: str) → gym_acnportal.gym_acnsim.envs.observation.SimObservation
constraint_matrix_observation() → gym_acnportal.gym_acnsim.envs.observation.SimObservation	Generates a SimObservation instance that wraps functions to
magnitudes_observation() → gym_acnportal.gym_acnsim.envs.observation.SimObservation	Generates a SimObservation instance that wraps functions to
phases_observation() → gym_acnportal.gym_acnsim.envs.observation.SimObservation	Generates a SimObservation instance that wraps functions to
timestep_observation() → gym_acnportal.gym_acnsim.envs.observation.SimObservation	Generates a SimObservation instance that wraps functions to

class gym_acnportal.gym_acnsim.envs.observation.SimObservation(space_function: Callable[[GymTrainedInterface], spaces.Space], obs_function: Callable[[GymTrainedInterface], np.ndarray], name: str)

Class representing an OpenAI Gym observation of an ACN-Sim simulation.

An instance of SimObservation contains a space_function, which generates a gym space from an input Interface using attributes and functions of the input Interface, and an obs_function, which generates a gym observation from an input Interface using attributes and functions of the input Interface. Each instance also requires a name (given as a string).

This class enables Simulation environments with customizable observations, as a SimObservation object with user-defined or built in space and obs functions can be input to a BaseSimEnv-like object to enable a new observation without creating a new environment.

Each type of observation is the same type of object, but the details of the space and obs functions are different. This was done because space and obs functions are static, as observations of a specific type do not have any attributes. However, each observation type requires both a space and observation generating function, so a wrapping data structure is required.

Attributes:

_space_function (Callable[[GymInterface], spaces.Space]):

Function that accepts a GymInterface and generates a gym space in which all observations for this instance exist.

_obs_function (Callable[[GymInterface], np.ndarray]): Function

that accepts a GymInterface and generates a gym observation based on the input interface.

name (str): Name of this observation. This attribute allows an

environment to distinguish between different types of observation.

_space_function :Callable[[GymTrainedInterface], spaces.Space]

_obs_function :Callable[[GymTrainedInterface], np.ndarray]

name :str

get_space(self, interface: gym_acnportal.gym_acnsim.interfaces.GymTrainedInterface) → gym.spaces.Space

Returns the gym space in which all observations for this observation type exist. The characteristics of the interface (for example, number of EVSEs if station demands are observed) may change the dimensions of the returned space, so this method requires a GymInterface as input.

Args:

interface (GymTrainedInterface): Interface to an ACN-Sim Simulation

that contains details of and functions to generate details about the current Simulation.

Returns:

spaces.Space: A gym space in which all observations for this

observation type exist.

get_obs(self, interface: gym_acnportal.gym_acnsim.interfaces.GymTrainedInterface) → numpy.ndarray

Returns a gym observation for the state of the simulation given by interface. The exact observation depends on both the input interface and the observation generating function obs_func with which this object was initialized.

Args:

interface (GymTrainedInterface): Interface to an ACN-Sim Simulation

that contains details of and functions to generate details about the current Simulation.

Returns:

np.ndarray: A gym observation generated by _obs_function

with this interface.

gym_acnportal.gym_acnsim.envs.observation._ev_observation(attribute_function: Callable[[GymTrainedInterface, EV], float], name: str) → gym_acnportal.gym_acnsim.envs.observation.SimObservation

gym_acnportal.gym_acnsim.envs.observation.arrival_observation() → gym_acnportal.gym_acnsim.envs.observation.SimObservation

Generates a SimObservation instance that wraps functions to observe active EV arrivals.

Zeros in the output observation array indicate no EV is plugged in; as such, all observations are shifted up by 1.

gym_acnportal.gym_acnsim.envs.observation.departure_observation() → gym_acnportal.gym_acnsim.envs.observation.SimObservation

Generates a SimObservation instance that wraps functions to observe active EV departures.

Zeros in the output observation array indicate no EV is plugged in; as such, all observations are shifted up by 1.

gym_acnportal.gym_acnsim.envs.observation.remaining_demand_observation() → gym_acnportal.gym_acnsim.envs.observation.SimObservation

Generates a SimObservation instance that wraps functions to observe active EV remaining energy demands in amp periods.

Zeros in the output observation array indicate no EV is plugged in; as such, all observations are shifted up by 1.

gym_acnportal.gym_acnsim.envs.observation._constraints_observation(attribute: str, name: str) → gym_acnportal.gym_acnsim.envs.observation.SimObservation

gym_acnportal.gym_acnsim.envs.observation.constraint_matrix_observation() → gym_acnportal.gym_acnsim.envs.observation.SimObservation

Generates a SimObservation instance that wraps functions to observe the network constraint matrix.

gym_acnportal.gym_acnsim.envs.observation.magnitudes_observation() → gym_acnportal.gym_acnsim.envs.observation.SimObservation

Generates a SimObservation instance that wraps functions to observe the network limiting current magnitudes in amps.

gym_acnportal.gym_acnsim.envs.observation.phases_observation() → gym_acnportal.gym_acnsim.envs.observation.SimObservation

Generates a SimObservation instance that wraps functions to observe the network phases.

gym_acnportal.gym_acnsim.envs.observation.timestep_observation() → gym_acnportal.gym_acnsim.envs.observation.SimObservation

Generates a SimObservation instance that wraps functions to observe the current timestep of the simulation, in periods.

To comply with the timesteps returned by arrival and departure observations, the observed timestep is one greater than than that returned by the simulation. Simulations thus start at timestep 1 from an RL agent’s perspective.