Search Unity

  1. Unity 6 Preview is now available. To find out what's new, have a look at our Unity 6 Preview blog post.
    Dismiss Notice
  2. Unity is excited to announce that we will be collaborating with TheXPlace for a summer game jam from June 13 - June 19. Learn more.
    Dismiss Notice
  3. Dismiss Notice

Question Creating an FSM agent in the dodgeball environment

Discussion in 'ML-Agents' started by Hallahallan, Mar 13, 2023.

  1. Hallahallan


    Jan 16, 2023

    I would like to create a simpler agent for benchmark purposes. I am using NEAT to create new agents for my masters thesis, but i need an agent that is similar to a standard game NPC to measure the results from the NEAT agent against.

    The environment i want to make the agent for is the ml-agents dodgeball environment. I want to note that I have manipulated the environment to be a 1v1 environment instead of the 4v4 elimination mode that is included in the dodgeball environment. Ideally I just want an agent that is able to play the game against a human opponent.

    I am not well versed in unity, so i was wondering if there was a simple and time efficient way to do this. If anyone knows of any frameworks/previously created FSM or rule based agents/general conventional approach to this problem, it would help tons!

    Here is the original Dodgeball environment:
    Here is my repo:

    Answers or discussions around the topic are appreciated.
  2. cecil_kellaway


    Sep 4, 2023
    Hi there,

    It's great to hear about your master's thesis project involving NEAT agents and the ml-agents dodgeball environment.
    Creating a simplified agent for benchmarking purposes in your NEAT-based master's thesis project is a wise approach. To achieve this, considering your specific requirements for the ml-agents dodgeball environment, you might opt for a rule-based agent. You can start by defining basic decision rules and strategies using Unity's scripting capabilities. Additionally, explore Unity assets and community-created frameworks like "Hutong Games' Playmaker" or "NodeCanvas" to streamline the implementation process.

    As you delve into this exciting project, don't forget to consult the ml-agents dodgeball environment documentation for any integration specifics. Best of luck with your thesis work; I hope this guidance proves valuable in your research! ️