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’re making changes to the Unity Runtime Fee pricing policy that we announced on September 12th. Access our latest thread for more information!
    Dismiss Notice
  3. Dismiss Notice

Question Can anyone recommend a course or other resources?

Discussion in 'ML-Agents' started by LiterallyCoder, Apr 7, 2023.

  1. LiterallyCoder


    Jul 25, 2020
    Hello everyone,

    I've just started learning ML-Agents, but I'm having a very hard time setting up something that actually works. I am familiar with the basics, including collecting observations and giving rewards/punishments. I've even tried using curiosity and imitation learning. Recently I've been working on a simple platformer-like project where I train the agent to move to a randomly-spawning platform and collect a food object on that platform. The agent can move and jump. However, no matter what I did, it almost never worked as intended. I would provide more context, but the project is a complete mess at this point and it even confuses me.

    I went back and forth between designing simple and complex reward functions, collecting more and less observations, using demonstration data, curiosity, you name it. However, the agent never really learned to do anything. It kept moving randomly even after millions of steps of training. Right now, I've ditched that project and I want to start something new. I am going to work on a new project where the agent will try to navigate through an obstacle course to some kind of goal. Could you guys please give me some tips on how to properly plan an ML Agent project so it doesn't result in a confusing mess of weird unfixable behaviour? I'll be happy with any advice at this point, I'm completely lost. There is a surprisingly low amount of tutorials and resources on ML Agents. I followed a few courses on YouTube but they're stupidly simple and don't involve any complex environments like obstacle courses, mazes, etc...