A computer system capable of watching gameplay videos of Super Mario Brothers on YouTube, and then creating hundreds of new playable sections of the game has been built by Georgia Tech scientists.
Zeroing in on the gaming terrain instead of the player avatar, as well as the positioning between elements onscreen, the system findss the required relationship or level design rule.
Pipes featured in the Mario games tend to stick out of the ground, for example, so the system learns this and prevents any pipes from being level with grassy surfaces. It also avoids “breaks” by using spatial analysis, like making sure there are no impossibly long jumps for the character.
Matthew Guzdial, lead researcher and a PhD student in computer science at Georgia Tech, said:
“An initial evaluation of our approach indicates an ability to produce level sections that are both playable and close to the original without hand coding any design criteria.”
Observing the players in action to see where they actually spend most of their time in the game is central to the process. After recording on-screen locations of sprites, Georgia Tech’s algorithms establish high-interaction areas (spots where players spend more time to collect bonus items or master a challenge).
The automatic-level designer targets these areas specially, to gather design information. The system is then able to build a new level section, element by element.
“Our system creates a model or template, and it’s able to produce level sections that have never been seen before, do not appear random and can be traversed by the player,” says Mark Riedl, associate professor of interactive computing. “One could say that the system ‘studies’ the design of Mario levels until it is able to create new playable areas.”
The system created 151 unique level sections from 17 samples in the original game. Productivity increased to 334 level sections as the system lowered the constraints. The new levels can be played easily by porting them into a game engine.