reward_manager

module reward_manager specifies a class that handles the rewards awarded by the environment.

class reward_manager.RewardManager(bounds: tuple, out_of_max_bound_reward: float, out_of_min_bound_reward: float, in_bounds_reward: float, punish_factor: float, min_distortions: Any, max_distortions: Any)

The RewardManager class

__init__(bounds: tuple, out_of_max_bound_reward: float, out_of_min_bound_reward: float, in_bounds_reward: float, punish_factor: float, min_distortions: Any, max_distortions: Any) → None

get_reward_for_state(total_distortion: float, current_state: State, next_state: State, min_dist_bins: Any, **options) → float

Returns a user specified reward signal depending on the state and the options given

Parameters

• state –

• options –