shasta.base_experiment

Module Contents

Classes

BaseExperiment

class BaseExperiment(config, core, experiment_config=None, *args, **kargs)

Bases: with_metaclass(ABCMeta, object)

reset()

Called at the beginning and each time the simulation is reset

abstract get_action_space()

Returns the action space

abstract get_observation_space()

Returns the observation space

abstract get_actions()

Returns the actions

abstract apply_actions(actions, core)

Given the action, applies the action to group or individual actor

Parameters:

action – value outputted by the policy

abstract get_observation(observation, core)

Function to do all the post processing of observations (sensor data).

Parameters:

sensor_data – dictionary {sensor_name: sensor_data}

Should return a tuple or list with two items, the processed observations, as well as a variable with additional information about such observation. The information variable can be empty

get_done_status(observation, core)

Returns whether or not the experiment has to end

compute_reward(observation, core)

Computes the reward