Computer Generated Mega Man Levels

Generative Adversarial Networks (GANs) can generate level segments for video games by learning from a collection of existing levels. This paper focuses on combining GAN-generated segments in a snaking pattern to create levels for Mega Man, a platforming game where adjacent level segments can be above, below, or to the side of each other. Levels link different types of segments to make cohesive levels, including vertical, horizontal, and corner segments. To pick appropriate segments, multiple GANs were trained on different types of segments to ensure that adjacent segments align properly. Then, the levels were evolved to maximize the length of each level's solution path. Using multiple GANs to represent different types of segments results in significantly longer solution paths than using one GAN for all segment types, and a human subject study verifies that these levels are more fun and have more human-like design than levels produced by one GAN.