Procedural content generation of puzzle games using conditional generative adversarial networks
This work addresses level generation for a specific puzzle game, but it is incremental as it builds on existing GAN methods without major breakthroughs.
The authors tackled procedural content generation for the puzzle game Lily's Garden using conditional GANs, achieving good performance in approximating map shape but struggling with piece distribution, with no concrete numerical results reported.
In this article, we present an experimental approach to using parameterized Generative Adversarial Networks (GANs) to produce levels for the puzzle game Lily's Garden. We extract two condition vectors from the real levels in an effort to control the details of the GAN's outputs. While the GANs perform well in approximating the first condition (map shape), they struggle to approximate the second condition (piece distribution). We hypothesize that this might be improved by trying out alternative architectures for both the Generator and Discriminator of the GANs.