Search Unity

few brains for one agent

Discussion in 'ML-Agents' started by AndrewGri, Apr 13, 2020.

  1. AndrewGri

    AndrewGri

    Joined:
    Jan 30, 2020
    Posts:
    12
    Let's assume I have one agent and I trained it for different behaviors separately and as result I have few models. How can I combine all of them together in Unity when running inference? I need agent to switch between brains by some condition.
     
  2. andrewcoh_unity

    andrewcoh_unity

    Unity Technologies

    Joined:
    Sep 5, 2019
    Posts:
    162
    Currently, this type of hierarchical use of policies is not explicitly supported by ML-Agents but it is an interesting possible feature. I think you'd have to do some significant work to make this happen i.e. create a separate agent with each model and then a meta controller that decides which agent is acting.
     
  3. AndrewGri

    AndrewGri

    Joined:
    Jan 30, 2020
    Posts:
    12
    OK, thanks will try that. And yes, as you mentioned, I think it would be great feature for future.