Using Genetic Algorithm for Wide yet Even Scattering of Game Objects: Applications on Irregular Levels and Involving Multiple Objects
Pratama Wirya Atmaja, Sugiarto Sugiarto · 2022
The rising costs of game development necessitate algorithmic and automatic methods for creating game content, also known as procedural content generation or PCG. Among many types of content, game levels are increasingly generated with PCG. The literature has recorded various level generation methods for specific games or contexts, yet no general-purpose one. We have previously addressed this gap by proposing a genetic algorithm-based method that scatters objects widely yet evenly on a level. The algorithm satisfies the player’s needs for exploration. Our previous work acquired good results when applying the method on a simple and rectangular level. This paper presents two follow-up case studies that reflect real games: one with irregular levels and another with multiple object types. Our results show that our method fits the first case study; however, the irregular level’s layout may yield unexpected results, such as the tendency of objects to appear along its edge. Meanwhile, we also find that objects of different types are best scattered separately, each under a specific parameter, to optimize their wide yet even distributions. We then discuss our results’ implications for PCG research and how to generalize our method to generate other types of game content.