Procedural Content Generation (PCG) has become one of the hottest topics in Computational Intelligence and Artificial Intelligence (AI) game research in the past few years. PCG is the process of automatically creating content for video games, rather than by hand, and can offer great benefits for video games companies by helping to bring costs down and quality up. By guiding the process with AI it can be enhanced further and even be made to personalize content for target players. Among the current research into PCG, search-based approaches overwhelmingly dominate. While search-based algorithms have been shown to have great promise and produce several success stories there are a number of open challenges remaining. In this thesis, we present the Learning-Based Procedural Content Generation (LBPCG) framework, which is an alternative, novel approach designed to address some of these challenges. The major difference between the LBPCG framework and contemporary approaches is that the LBPCG is designed to learn about the problem space, freeing itself from the necessity for hard-coded information by the game developers. In this thesis we apply the LBPCG to a concrete example, the classic first-person shooter Quake, and present results showing the potential of the framework in generating quality content.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:618033 |
| Date | January 2014 |
| Creators | Roberts, Jonathan Ralph |
| Contributors | Chen, Ke |
| Publisher | University of Manchester |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | https://www.research.manchester.ac.uk/portal/en/theses/learningbased-procedural-content-generation(1af8d23d-8ceb-416b-b4ba-d7a2970b47ef).html |
Page generated in 0.002 seconds