Implementing reinforcement learning into mesa[GSOC - 2024]

[ai-naymul]

while implementing reinforcement learning into mesa. I think Openai gym will be a great choice for us beacuse regarding mesa's python based enviroment and the need for flexibality and and ease of integration.Gym is widely used and it provides variaty of enviroment that serves as benchmark for integrating RL into mesa.

Looking forward for your feedbacks.
Thanks :)

[ai-naymul]

Hey @rht @EwoutH @Corvince @tpike3 @wang-boyu,

I have a made a minimal example of mesa RL using OpenAI Gym Liabry. Here is the refernce : https://github.com/ai-naymul/mesaRL

[rht]

I know the API is the same, but shouldn't it be Gymnasium, successor of Gym?

[ai-naymul]

Yes, you're correct. While the API remains largely the same, the successor to OpenAI Gym is indeed Gymnasium. Gymnasium continues the development of the original Gym API, addressing issues and adding features that may not be present in the older versions of Gym. For those integrating with Mesa or any other project, it's worth considering the switch to Gymnasium for the most up-to-date features and improvements in reinforcement learning environments.