Air Hockey Environment

1. Set up

It's a simple air hockey environment where you can train your own AI using self-play technic and Unity ML-Agents library. To start your own training, you need to follow these steps:

1. Clone this repo
2. Add this project to your Unity Hub
3. Install ML-Agents

2. Usage

In this environment all scripts are stored in the ./Assets/Environment/Scripts directory. You should construct your reward function in the AirhockerTrainer.cs script. You can change any scripts. Don't change the agent's properties, such as observations, actions, and e.t.c, it may cause incorrect behavior in our inference when you upload your model.

You can also move the handles:

• Blue Handle

  • W - Up
  • A - Left
  • D - Right
  • S - Down

• Red Handle

  • Up Arrow - Up
  • Left Arrow - Left
  • Right Arrow - Right
  • Down Arrow - Down