Search Unity

  1. Welcome to the Unity Forums! Please take the time to read our Code of Conduct to familiarize yourself with the forum rules and how to post constructively.
  2. We have updated the language to the Editor Terms based on feedback from our employees and community. Learn more.
    Dismiss Notice

Question Multi-agent ( asymmetric game )

Discussion in 'ML-Agents' started by yasinmasmoudi, Jan 4, 2023.

  1. yasinmasmoudi

    yasinmasmoudi

    Joined:
    Nov 15, 2021
    Posts:
    6
    Hello,

    Disclaimer : I am new to ml-agents

    So I want to train 3 different agents that each control a different part of a machine in order to get a certain task done. From what I understood from this post :https://github.com/Unity-Technologi...sign-Agents.md#defining-multi-agent-scenarios, I need to give each agent a different Behavior Names and different team IDs, yet when making a SimpleMultiAgentGroup I need to give them the same behavior name and Behavior Parameters.
    So what should I do ?
     
  2. hughperkins

    hughperkins

    Joined:
    Dec 3, 2022
    Posts:
    191
    Depends how you want to train it. If each agent has a different behavior, then I believe it acts completely independently. This will be fairly unstable, but will learn. The SimpleMultiAgentGroup apparently does not support agents with differing behavior. One other option is to have a *single* mlagents Agent in your game, that controls all three machine parts. This might learn faster than three independent agents, since it means you basically have a single AI controlling the three machine parts in a co-ordinated way, a little like playing chess.