eprllib.Agents.Rewards.BaseReward

Reward Function

This module contains the base class for defining reward functions.

It is preferred to use the infos dictionary and not the observation, since the latter is a numpy array and cannot be called by key values, which is prone to errors when developing the program and indexing an array may change.

The terminated and truncated flags are arguments in the reward function get_reward method to allow implementations with dispersed reward. This flags allow return the final reward when the episode ends.

Classes

BaseReward(agent_name[, reward_fn_config])

This class is the base class for defining reward functions.