Incremental Inductive Learning of Answer Set Programs for Maps Generation Problems

In game development, Procedural Content Generation is an approach that replaces the designer's task in creating contents of games, e.g., game maps. We introduce an incremental learning process that utilizes Inductive Learning (IL) of Answer Set Programs (ASP) to automate solving maps generation problems rather than to explicitly specify the characteristics of the maps. In an incremental learning process, a complex learning task is divided into a sequence of learning iterations, where each iteration consists of a set of smaller learning tasks to learn a set of rules. In order to speed up the learning process, each task in the same iteration is solved asynchronously. Our experiments show that IL of ASP successfully learns an answer set program. That is, it provides a set of rules for generating a collection of game maps that possess the same characteristics as the maps referred in the learning scenario.