This was a study to help determine how useful PCGML (Procedural Content Generation via Machine Learning) techniques are when used in an editor for creating platforming levels for a simplified version of SuperTux (which is an open source game similar to Super Mario Brothers). To maintain privacy, no personal information was collected.
We do have plans on posting the source code for both the game and the editor to a github repo and to post a version of the game and editor on litetux.ca We are hoping to have this work done by the end of August, so if you are interested in the source or playing the game then please check back again in September for updates.
For those who did not participate in the study, below is a screenshot from the game. Notice that it is a fairly simple platforming game.
The mixed-initiative editor is going to be a bit trickier to port. We will likely have local save support and possibly the ability to share levels via links in the future. The source code will contain information on how to build your own generators for anyone interested in continuing this research with other generators. Below is a screenshot from the editor used to create levels for the game.
SuperTux is an open source game that has similar gameplay mechanics to Super Mario Brothers (SMB) which is the game related researchers used. LiteTux is a simplified clone of it due to SuperTux’s complexity and wanting the game and editor to be used in a browser.
Essentially this is a way of getting a computer to create things such as images, music, screenplays, or even video game levels. There is a lot of research in this area with several different ways of generating content, with the focus of our study being deep learning approaches.
The basic idea is to have both a human and an AI work together to create a level. This allows for the strength of both be combined hopefully resulting in better results than each could achieve independently. Being able to produce good quality levels faster is the goal.
Artificial Intelligence is the attempt to think and learn like humans do. The field of AI is very large with this project using different subfields of this field, in particular machine learning and deep learning.
Not for a long time, if ever. It should, however, make it easier for people to create levels which hopefully will result in more people able to make good quality levels, and allow for the more talented designers to create even more spectacular levels.