You can use environments that i've developed when you train a reinforcement learning agent.
- Unity Editor: 2021.3.6f1
- Unity ML-Agents Toolkit: Release 19
- Python: 3.7.13
Install the mlagents Python package
conda install pytorch==1.11.0 torchvision==0.12.0 torchaudio==0.11.0 cudatoolkit=11.3 -c pytorch
python -m pip install mlagents==0.28.0
First of all, Open the Unity ML-Agents Project project. After it, You can see the Assets/Environments directory which includes all RL environments. Go to the Assets/Environments/<Environment Name>/Scenes folder, if you want to use an environment. Open the scene file of the environment. You can use it as two ways to train.
- build an environment as an executable file
- directly interact with Unity Editor
You can see descriptions of environments in Wiki.
- config - Training configuration files
- Unity ML-Agents Project - Unity ML-Agents project