Hello
I am looking into multi agent cooperative navigation in dynamic environment using MAPOCA.
I have read the paper on absorbing states and MAPOCA but ran into some confusions.
Does the MAPOCA framework operate on centralised critic and decentralised actor method like MADDPG and MAPPO?
If so, is it able to perform all the general multi agent reinforcement learning functions like MADDPG and MAPPO where the agents don’t expire? or is it strictly for cases where agents will spawn or respawn?
And also, is it possible to import MADDPG/MAPPO in unity using mlagents plugin?