Asking about states and actions

[A S]

Hi all,

I have a question regarding multi agent reinforcement learning.

I have agents with states and actions and a Gridworld with world states. Given that the world is big and the number of states is large. And since I'm usig q-learning with joint states and actions I decided that instead of using the world states I can use the agents state in my program. Is it possible?

[Gautam]

Would you not also need the state of the world around? If you need to record episodes, it will store the state, which includes world state and agent state.