Learning-based game generators in first-person shooter

This research project aims at a novel procedural content generation method to generate high quality first-person shooter (FPS) maps in real time. As a result, we have proposed a learning-based method to produce constructive primitives (or building blocks) in FPS. In this approach, active learning by encoding a game designer's knowledge implicitly is applied to obtain those building blocks. Thus, real-time map generation and content adaptation can be done by integrating those blocks with a variety of criteria.

For demonstration, we have applied our approach to a classic FPS game named Cube 2: Sauerbraten for a proof of concept. In this demo, two automatic content generators are implemented: one is a real-time game generator that can generate high quality games via controllable parameters, and the other is an adaptable generator which adapts games to match challenges that a game player can cope with. Screenshots of our procedural games are as follows.

The demo prototype on two scenarios including executable files are available [here]. After downloading and unzipping the generators.zip file, one should read the tutorial document first. If there are any issues with the executable file and tutorial provided, please contact me (shipa@cs.man.ac.uk).