JOSS: Functionality documentation

[abhiramm7]

Hi, I commend the authors on thorough documentation. In the JOSS paper submission, you mention that Abmarl integration with open ai gym and MultiAgentEnv. Having some documentation or a simple example on how users can interface with the two packages might be helpful.

[rusu24edward]

HI @abhiramm7

Thank you for your insightful review. Regarding integration with MultiAgentEnv and gym.Env, we do have this documented in the API spec here and here. Additionally, we the MultiCorridor tutorial and PredatorPrey tutorial show how to use the MultiAgentWrapper to treat the simulation as a MultiAgentEnv.

Does the documentation need to be clarified? Should I add another example using the GymWrapper?

[abhiramm7]

I see that the MultiCorridor uses gym.spaces. I was under the impression that abmarl would be able to wrap around an existing open ai gym environment. I apologize if I am mis-interpreting it. I was thinking an example that uses an existing open ai gym environment with abmarl would be helpful to users.

[rusu24edward]

Thanks for the clarification! I agree, this would be helpful. I'll add a page on running Abmarl with already-existing gym environments.

[rusu24edward]

@abhiramm7 Here is a brief demo for training a gym environment with abmarl: https://abmarl.readthedocs.io/en/latest/tutorials/gym.html