Generating Entertaining Platform Game Levels
Nelson André Amaral de Oliveira · Open Repository of the University of Porto (University of Porto) · 2014
Platform games have become less popular with the evolution of the video game industry. Nevertheless, with the recent growth in popularity of mobile devices, they become yet again popular, due to their simplicity in mechanics and game play. Since their popularity is increased, they can be made even simpler, far more addicting and entertaining by utilizing techniques, consequently reducing development costs. This is where procedural content generation comes in play, since creating levels manually is time and money-consuming. These techniques will allow games to have their longevity increased, and become far more entertaining. This dissertation consisted in the usage of machine learning techniques to generate platform levels for the known platform game Super Mario Bros., having the fun factor severely emphasized. For that to happen, several procedural content generation techniques have been utilized, from the simplest level portion to the complex genetic algorithm based tournament selection. These levels are also parametrised by each user’s likings, throughout certain features such as enemy density, the selection of far more common platforms or the difficulty itself, which is based on each player’s real performance. Two social experiments have been performed where majority (90%) of participants considered these levels similar to manually designed ones, augmenting the credibility of said generated levels. These experiments were conducted with players as participants, therefore also boosting the credibility of the content generator.