Hello,
I am trying to figure out which tool I should use for a machine learning application.
In my case, I want to perform reinforcement learning on multi-agent scenarios, with both collaboration and competition taking place. (basically, the equivalent of a football game, with each agent playing a player)
I am therefore considering Unity ml-agents, as it gives me a full 3D environment with physics, etc...
However, when looking at the unity-gym part (because I would like to have my environment running as a gym environment), the readme says I can only have 1 brain per environment.
Does it make it impossible to train multiple different agents in a multi-agent environment ? And if so, where does this limitation come from ?
Hello,
I am trying to figure out which tool I should use for a machine learning application.
In my case, I want to perform reinforcement learning on multi-agent scenarios, with both collaboration and competition taking place. (basically, the equivalent of a football game, with each agent playing a player)
I am therefore considering Unity ml-agents, as it gives me a full 3D environment with physics, etc...
However, when looking at the unity-gym part (because I would like to have my environment running as a gym environment), the readme says I can only have 1 brain per environment.
Does it make it impossible to train multiple different agents in a multi-agent environment ? And if so, where does this limitation come from ?