When creating an environment i want to be able to tell for every agent what its observation space is.
First one should select the wanted state representation: Symbolic (a list of objects) or feature planes or image pixels
Then one should be able to select the MDP representation (the whole environment is visible) or a POMDP representation
where one can select the view radius.
When creating an environment i want to be able to tell for every agent what its observation space is.
First one should select the wanted state representation: Symbolic (a list of objects) or feature planes or image pixels Then one should be able to select the MDP representation (the whole environment is visible) or a POMDP representation where one can select the view radius.