ray.rllib.env.multi_agent_episode.MultiAgentEpisode.get_return

MultiAgentEpisode.get_return(include_hanging_rewards: bool = False) → float

Returns all-agent return.

Parameters:

include_hanging_rewards – Whether we should also consider hanging rewards wehn calculating the overall return. Agents might have received partial rewards, i.e. rewards without an observation. These are stored in the “hanging” caches (begin and end) for each agent and added up until the next observation is received by that agent.

Returns:

The sum of all single-agents’ returns (maybe including the hanging rewards per agent).