Abstract

With the release of Pokémon GO, location-based games (LBG) have become mainstream. However, the individual gaming experience may vary depending on the location from which users play. This is because content creation is a significant challenge for LBG development and points of interest are often heterogeneously distributed in certain areas. Manually selecting POIs is time-consuming and difficult to scale, while crowd-sourced approaches require a large community. Procedural content generation, on the other hand, can select POIs automatically with minimal human input. This thesis addresses the problem of creating scalable and engaging game spaces for location-based games. It aims to implement and analyze an algorithm capable of automatically projecting real-world POIs onto a virtual map for LBGs. First, POIs are fetched from OpenStreetMap, classified by relevance using their meta-data, and stored in a graph database. Subsequently, the earth is subdivided into cells using S2Geometry, for which the algorithm calculates content based on the POI data in the graph database. Using cells leads to a homogeneous distribution of points. In addition, the classification of POIs ensures the usage of high-quality POIs where possible and lower-quality ones where necessary. In summary, the algorithm can generate well-distributed, static game spaces for the entire world but needs substantial computing power. Further work may enhance the algorithm to provide more dynamic game spaces that adapt based on player behavior and real-world events. Some resulting world structures also require players to cross borders like rivers or highways. This still needs to be improved.

Results/Implementation/Project Description

Conclusion

[ PDF (optional) ] 

[ Slides Kickoff/Final (optional)]