Machine Learning for Sonic the Hedgehog Video Games

Dion Watts-Evans


Supervised by Federico Liberatore; Moderated by Charith Perera

My proposal is a Machine Learning project to complete levels in the classic sega megadrive sonic games.

Creating an Ai for these games provide several interesting challenges for me to overcome such as:

- developing a meaningful heuristic for the games' often sprawling level design often requiring the player to move backwards to progress

- creating proper tools to allow the Ai to visualise the playspace

In addition to completing levels in the true spirit of the games I would like to make my Ai prioritise completing the level as fast as possible. allowing for some opportunities in designing my program.

Initial Plan (05/02/2021) [Zip Archive]

Final Report (27/05/2021) [Zip Archive]

Publication Form