An author-centric approach to procedural content generation

Abstract

This paper describes an alternative approach for videogame procedural content generation focused on providing authors direct control on what gameplay ensues from the generated content. An architecture is proposed that allows designers to define, beforehand, target gameplay indicators, and then generates content for an existing base-design that achieves those same indicators in actual gameplay sessions with human players. Besides providing a description of this architecture, a trial intent on giving evidence of the approach's feasibility is presented. This experiment used an altered version of 'InfiniteMario Bros.' level generator, built to evolve design parameters so as to achieve 3 target gameplay indicators. Employing a Genetic Algorithm in generation of new parameter values, and using 25 players to test the end results, the platform was able to generate parameters that achieved, with precision, the values for those indicators. This result provides evidence of the approach's feasibility, hinting at its potential use for real-life design processes. © Springer International Publishing 2013.