Bluecap is a simple automated game designer, built in Unity. It includes a little bit of everything to help get you started making AI systems that design games.
Bluecap was designed to pair with my 2020 tutorial, "Getting Started in Automated Game Design", which I gave at the IEEE Conference on Games. You can watch the tutorial on YouTube here:
Download or clone the repository, and then open in Unity 2019 LTS, available here: https://unity.com/releases/2019-lts
I recommend using the Unity Hub to organise your Unity installations and projects. You can get the Unity Hub here: https://unity3d.com/get-unity/download
Details on how to upgrade this to Unity 2020 are coming soon!
Bluecap contains:
- A simple game framework for making two-player, 2D abstract strategy games.
- A design language that can be used to generate random game designs.
- An interpreter that can read game descriptions and let you play games against AI or other people, as well as export them as standalone games.
- Three different AI agents that can play games (Random, Greedy and MCTS).
- An evaluation system that uses the AI agents to test games by playing and analysing them.
- A little (and simplified) "game generator" that randomly generates and evaluates games.
It also comes with several scenes to let you get started immediately with playing games, watching AI play games, and watching games get generated and evaluated!
I've tried my best to comment the code liberally, but it still probably needs more comments and documentation in places. I will endeavour to expand this as soon as I can.
A Bluecap is a creature from English folklore that would help miners with their work - as long as it was paid a fair wage! I like the idea of naming creative AI systems after mythological creatures like this, that can help us as well as hinder us, and that ask to be treated fairly and like equals. Bluecap isn't a fancy or sophisticated system, but if you treat it well, I'm sure you'll both find valuable things in the research mines :)
Things I'd like to add in the near future (but may not have time):
- A -secret thing- in collaboration with Mads Johansen.
- An example of an evolutionary algorithm that develops new game designs.
- An expanded design language with lots of new mechanics.
I'll do my best! I hope that what's here right now is useful to you though. If you ever want to chat or ask questions, please don't be shy, I'd love to help if I can: @mtrc on Twitter or mike@possibilityspace.org
- 11 August 2020