Geração procedural de biomas em mapas de jogos bidimensionais

Abstract

The electronic game industry increasingly demands the creation of vast and complex worlds, rendering Procedural Content Generation (PCG) an essential tool for ensuring replayability and development efficiency. This work explores PCG techniques for the creation of two-dimensional maps in platform and exploration (sandbox) games, utilizing the game Terraria as a case study for topological and structural validation. The main objective was to develop a system capable of orchestrating multiple algorithms to reproduce the geological complexity of a virtual world. The adopted methodology combined Perlin Noise for macro-topography and biomes, Cellular Automata for simulating erosion in deep caves, and Random Walk algorithms for excavating tunnels and architectural structures, such as dungeons. The results demonstrated that the hybrid and stratified approach is effective in generating coherent maps, allowing for the faithful reproduction of soil layers, caverns, biomes and specific structures of the reference game. The final system yielded deterministic maps based on seeds, validated through direct visual comparison with the original generation of the case study.