Procedural content generation in tabletop gaming faces significant hurdles due to the rigid physical constraints of modular components and the necessity for balanced, solvable gameplay. To address these challenges, an evolutionary algorithm has been developed to automate the creation of scenarios for the board game Gloomhaven. By leveraging existing, handcrafted scenarios as a foundational dataset, the system evolves genotypes that encompass map layouts, monster placement, and treasure distribution. This approach moves beyond simple randomization, ensuring that every generated dungeon adheres to the specific spatial and thematic requirements of the game’s physical components.
The system employs a sophisticated fitness function that evaluates potential scenarios based on difficulty, structural complexity, and thematic coherence. Crucially, the algorithm integrates graph-based reachability analysis and physical constraint verification to guarantee that every generated dungeon is both solvable and playable. Experimental results confirm the efficiency of this methodology, with the system capable of producing high-scoring, valid scenarios in approximately one second. This rapid generation capability demonstrates a viable path toward automating content creation that maintains the high design standards typically associated with manual development.
While the current implementation successfully manages the technical and structural requirements of scenario design, the research highlights ongoing opportunities for refinement. Future iterations aim to enhance the human-centric aspects of these dungeons, specifically through improved visualization tools and the integration of large language models to provide richer narrative context. By bridging the gap between algorithmic efficiency and creative design, this framework establishes a robust foundation for the future of automated tabletop content generation, offering a scalable solution for expanding complex game systems without compromising gameplay integrity.
