Connect Python with Unity for reinforcement learning!

Video:

• see examples

You need to install both into Unity project, and into your Python environment.

• In Unity, go to Window | Package Manager
  • change the dropdown in top left to 'Packages: In Project'
  • check that Newtonsoft Json appears
  • if Newtonsoft Json is not already in your project:
    • click on the '+' in the top left, and choose 'Add package by name...'
    • put the name com.unity.nuget.newtonsoft-json, then click 'Add'
• In Unity, in your project's "Assets" folder, create a "Plugins" folder, if it doesn't already exist
• First install AustinHarris.JsonRPC:
  • Download https://www.nuget.org/api/v2/package/AustinHarris.JsonRpc/1.2.3
  • rename to have filename suffix be .zip (you might need to turn on options to see all file extensions)
  • unzip the resulting zip file
  • copy lib/netstandard2.1/AustinHarris.JsonRpc.dll into your Plugins folder
  • select the file, in your Plugins, and in 'Inspector' unselect 'validate references', and click 'Apply'
• Click on 'Releases', on the right of the github page
  • in the most recent release, download PeacefulPie.dll
• Copy PeacefulPie.dll into your Unity project's Plugins folder
• If on Mac silicon:
  • in Unity, goto Plugins, click on PeacefulPie.dll
  • in Inspector, change 'CPU' setting from 'Intel' to 'Any CPU', then click 'Apply'
  • do the same for AustinHarris.JsonRpc.dll

You should be good to go :)

pip install -U peaceful-pie

• currently tested with:
  • python 3.10
  • Unity 2021.3.16.f1
• support Python version: >=3.8.0, <4.0.0
• please create an issue if your preferred platform is not supported (I'm guessing I might need to downgrade Python a little? Let me know!)

I'm usng Visual Studio for Mac to write this.