Will it support Multi-Agent RL in the future? about parl HOT 3 CLOSED

Yes! We will open-source a multi-agent example in the next release!

For centralized Multi-agent reinforcement learning(MARL) algorithm, its implementation is similar to the single-agent algorithm.
For decentralized MARL algorithm, we can synchronize parameters among different agents by using xparl, the parallel computation tool in PARL.

from parl.

thx，I can't wait for it.

from parl.

Thanks for your interest in PARL. We are now working on the state-of-the-art algorithms on continuous spaces such as TD3, SAC. For other algorithms, we still need more time to implement.

from parl.