[rllib] Accessing the reward of other agents in custom multi-agent algorithm

[ericzhao28]

What is your question?

My goal is to implement a custom multi-agent algorithm. However, to update the policy for any individual agent requires access to the rewards of other agents. Is there any way to access the rewards issued to other agents? I imagine this may be similar to the problem of accessing the observation of other agents, which seems to be an open problem (see #7341, #4748).

[sven1977]

Yeah, we are not fully there yet, but are working on a generic API to enable easy set up of things like this in any RLlib agent. The project is called "Trajectory View API" and we have some solutions for LSTM/framestacking, but not multi-agent yet (this will be the last thing we'll add to this API). A rough time estimate, on when this will be done is this Q3, but can't make promises.

[stale]

Hi, I'm a bot from the Ray team :)

To help human contributors to focus on more relevant issues, I will automatically add the stale label to issues that have had no activity for more than 4 months.

If there is no further activity in the 14 days, the issue will be closed!

• If you'd like to keep the issue open, just leave any comment, and the stale label will be removed!
• If you'd like to get more attention to the issue, please tag one of Ray's contributors.

You can always ask for help on our discussion forum or Ray's public slack channel.

[stale]

Hi again! The issue will be closed because there has been no more activity in the 14 days since the last message.

Please feel free to reopen or open a new issue if you'd still like it to be addressed.

Again, you can always ask for help on our discussion forum or Ray's public slack channel.

Thanks again for opening the issue!

[MiladInk]

Is this possible in rllib now?