Generating maps using Markov chains

Abstract

In this paper we outline a method of procedurally generating maps using Markov Chains. Our method attempts to learn what makes a "good" map from a set of given human-authored maps, and then uses those learned patterns to generate new maps. We present an empirical evaluation using the game Super Mario Bros., showing encouraging results. Copyright © 2013, Association for the Advancement of Artificial Intelligence. All rights reserved.