Orchestrating and Adapting of Dungeon Levels, Locked-door Missions, and Enemies

• Ano de defesa: 2022
• Autor(a) principal: Viana, Breno Mauricio de Freitas
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Adaptive generation; Content orchestration; Enemy generation; Geração adaptativa; Geração de inimigos; Geração de níveis; Level generation; MAP-Elites; Orquestração de conteúdo
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/55/55134/tde-19072022-164759/

Procedural Content Generation (PCG) techniques can be used to automatically generate game content or increase the designers creativity and productivity. Besides, PCG can work as a game feature by providing diverse and targeted content for players. In this context, we tackle the problem of adaptive content orchestration, specifically by exploring how coordinate the generation of levels, missions, and enemies for an Action-Adventure game and different types of players. Thus, the present masters thesis proposes a PCG system to provide adaptive gameplay experiences for different players. Our system is focused on three different game facets, dungeon levels, narratives (missions), and rules (enemies), and it comprises three modules, orchestrator, classifier, and game prototype. The orchestrator module coordinates two algorithms for generating levels and enemies; both apply MAP-Elites to maintain a variety of solutions without losing quality. The level generation approach creates dungeons with enemies (levels facet) and locked-door missions (narratives facet). Next, the enemy generation approach creates enemies with different attributes and behaviors (rules facet). The classifier module receives the players answers to a brief questionnaire regarding their gameplay preferences to categorize players profiles. To adapt the contents, we defined different goals of each generator for each player type. Based on the player type, the orchestrator module appropriately combines the previously generated levels and enemies. We designed the orchestrator to filter and select coherent and good enemies to place in the levels rooms. The game prototype module is where we validate the contents generated by our system and collect data from the players. Our results show that the two MAP-Elites algorithms accurately converge almost the whole population with many executions and cases. The players feedbacks show that they enjoyed the levels played and the enemies faced. Besides, most of them could not indicate that an algorithm created the levels or the enemies. Our system presented positive results for delivering adaptive content properly for different types of players through a simple player profiling process. Thus, we can conclude that our PCG system can generate levels and enemies to entertain different players.