Hi, I have 4 agents in a single environment learning to drive in lane. For GAIL, do i have to deactivate all agents except one for recording demonstrations? Must I have demonstrations for each and every agent in this case since all the agents take different actions(opposite lane hence different direction and actions) but share the same behaviour of keeping in lane? Thanks
I don't know what are you trying to achieve, but GAIL and all that stuff gives poor results, reinforcements learning alorgithm ppo that they say state of art even is weak, you 'd better think about inventing a knew algo, Unity suffer from AI