I have a custom multi-agent environment with multiple groups of agents, where I would like each group to utilize the same policy. I configured this by setting AlgorithmConfig.multi_agent(policies={...}, policy_mapping_fn=...). Having the policy remember a history past states and actions would be ver…

Use LSTM model for policy gradient multi-agent with different recurrent hidden states per agent

kimhyun1018 October 21, 2025, 1:53am 2

Is there any update on this? I’m facing a similar issue too.

Topic		Replies	Views
Multi-Agent PPO rollout stalls + “duplicate agent at same env-t” assertion under Connector-v2 (Ray 2.49.2) RLlib	3	37	October 17, 2025
Multi agent partial parameter sharing RLlib	2	442	November 30, 2023
Sharing an LSTM cell between policies RLlib	2	412	July 1, 2021
Decentralized multi agent reinforcement learning RLlib	4	208	November 2, 2024
Multi agent checkpoints - KeyError: 'default_policy' RLlib	1	608	October 30, 2021