Evolving spaceships with a hybrid L-system constrained optimisation evolutionary algorithm
Roberto Gallotta, Kai Arulkumaran, Lisa B. Soros · Proceedings of the Genetic and Evolutionary Computation Conference Companion · 2022
Content generation for video games is a time-consuming process that impacts both game development and gameplay in the case of sandbox games. One approach to improve this process is to use procedural content generation, which can automatically generate diverse content for the game. Lindenmayer systems (L-systems) with handcrafted rules have successfully been used to generate content for games, but rely on undirected expansion of the rules. In contrast, by combining L-systems with an evolutionary algorithm for constrained optimisation, we are able to efficiently utilise simple L-system rules with directed search to find functionally-viable and appealing spaceship designs within the Space Engineers game.