RewardAggregatorInterface¶
-
class
maze.core.env.reward.
RewardAggregatorInterface
¶ Event aggregation object for reward customization and shaping.
-
abstract classmethod
to_scalar_reward
(reward: Any) → float¶ Aggregate sub-rewards to scalar reward.
This method is useful for example in a multi-agent setting where we could sum over multiple actors to assign a joint reward.
- Param
reward: The aggregated reward (e.g. per-agent reward for multi-agent RL settings).
- Returns
The scalar reward returned by the environment.
-
abstract classmethod