This graduate thesis takes form of a 2D spaceship fighting game in which teams of dozens of AI ships explore the vast map and fight to increase score.
The ships have modular loadouts which change how they play, which are adjusted and balanced during live gameplay thanks to the underlying genetic algorithm.

This demo is powered by a generic genetic algorithm library I wrote to support any sort of demo I created, as the GA idea came before the space game to show as representation. This means it can be applied to multiple projects with little changes, and quickly test & balance gameplay values.

The data from the thesis demo supported the idea that realtime genetic algorithms can be used to hone in gameplay and discover bugs at a substantially faster rate than a human could, but they may require game modification to run during live gameplay rather than in looping parallel genetic algorithms.