Search Unity

  1. We are migrating the Unity Forums to Unity Discussions. On July 12, the Unity Forums will become read-only.

    Please, do not make any changes to your username or email addresses at during this transition time.

    It's still possible to reply to existing private message conversations during the migration, but any new replies you post will be missing after the main migration is complete. We'll do our best to migrate these messages in a follow-up step.

    On July 15, Unity Discussions will become read-only until July 18, when the new design and the migrated forum contents will go live.

    Read our full announcement for more information and let us know if you have any questions.

Question Limit possible actions & deep action

Discussion in 'ML-Agents' started by vooood69, Apr 11, 2023.

  1. vooood69


    Jul 20, 2013
    I have two questions if it's possible to do with ML-Agents.

    1) Is it possible to limit what actions an agent can take at a time? I know I can give it negative score for an action taken when it shouldn't be but I would like to train an agent to do setup and play where setup has a completely different set of actions, while in play possible actions might be limited by rules or environment and I would like to skip extra cycles to teach the agent to avoid actions at a specific phase of the game or even it to decide to take a forbidden action because cheating will lead it to victory.

    2) Is it possible to have some kind of "deep" actions where an action taken by an AI then requires additional decisions. Think of this as there is a unit with multiple elements in it and the AI decides to take a "move" action for the unit. Now it has to move a unit but each element inside the unit has to be moved on it's own and formation is not kept.

    (I am aware that 2) might be handled if 1) is supported.)

    Thank you in advance for your help.
    Last edited: Apr 11, 2023
  2. kokimitsunami


    Sep 2, 2021
    vooood69 likes this.