Procedural content generation (PCG) is the automatic creation of content. In games, for example, the content refers to the environments, the tools within the game and even the textures. Creating game content by hand is expensive and time-consuming. If meaningful content can be generated programmatically, it can clearly save a lot of time and money for developers. PCG can have some problems though.
My research includes trying to classify the difficulty of game content (such as levels in puzzle games), to allow certain PCG methods to generate content with a specific level of difficulty.
Search-based procedural content generation is a PCG method that uses a generation function, an evaluation function and a search method that together generates content, evaluates it, and then alters the generation function to generate better content. Specifically, I aim to create an evaluation function that can be used in such a PCG method, where better content means that it is closer to the specified level of difficulty.
I intend to use Monte Carlo tree search (often used to create players for the game of Go) to simulate the playing of generated puzzles. I will then analyze such simulated playthroughs and attempt to identify features that correspond to the difficulty of the puzzles. These features can then potentially be used to judge the difficulty of content for the purposes of generating meaningful game content.My contact details: email@example.com